Toward an integrated framework of media dependency: a theoretical review of uses–dependency and new media dependency models

1 Introduction

Media dependency theory is one of the most enduring and adaptable frameworks in communication scholarship. Since Ball-Rokeach and DeFleur (1976) introduced their dependency model of mass media effects, the framework has undergone substantial theoretical refinement and empirical extension in response to transformative changes in media technologies, platforms, and audience behaviors. The original model challenged prevailing direct-effects paradigms by positioning dependency relations—rather than message characteristics alone—as the central mechanism through which media alter beliefs, feelings, and behaviors. It foregrounded a systemic perspective: media effects are conditional on the strength of audience reliance on media information resources, shaped by social context, the availability of functional alternatives, and the centrality of media to fundamental human goals.

The significance of this theoretical contribution can be understood against the backdrop of communication theory at the time. During the mid-twentieth century, the field oscillated between powerful-effects models—exemplified by the hypodermic needle theory—and limited-effects models associated with Lazarsfeld and his colleagues, which emphasized selective exposure, opinion leaders, and minimal direct influence. Ball-Rokeach and DeFleur's (1976) intervention offered a middle path: media effects are neither uniform nor negligible but are contingent upon the structural relationships audiences maintain with media systems. This conditional approach fundamentally reoriented media effects research by directing scholarly attention away from whether media produce effects and toward identifying when and under what conditions effects are most pronounced. The conceptual vocabulary of dependency relations, functional alternatives, and goal-oriented reliance provided a coherent theoretical language for analyzing these contingencies across diverse media contexts.

The proliferation of digital technologies and ubiquitous mobile media has necessitated theoretical adaptations that preserve the core dependency logic while accommodating new affordances and use patterns. Internet Use and Dependency models emerged to account for the perpetual availability, interactivity, and multiplicity of online information resources, shifting analytical focus from broadcast systems to specific platforms and their perceived attributes (Ognyanova and Ball-Rokeach, 2015; Park, 2013). Cho's (2009) New Media Uses and Dependency Effect Model subsequently integrated habitual use patterns as antecedents to dependency, specifying mediation pathways linking routine media behaviors to dependency intensity and subsequent outcomes. More recently, scholars have begun extending dependency logic to algorithmically curated and AI-mediated environments, theorizing “algorithm dependency” in platformized news use (Schaetz et al., 2025), modeling how heightened emotional states intensify social media dependency under conditions of pervasive ambiguity (Omori et al., 2026), and developing validated measures of dependency on generative artificial intelligence (Goh et al., 2025).

The rapidity and scope of digital transformation underscore the urgency of this theoretical work. Global Internet penetration has grown dramatically over the past two decades, and social media platforms now serve billions of users worldwide. These platforms have become central venues for news consumption, social interaction, political participation, and entertainment—functions once distributed across distinct media systems. As audiences integrate these platforms into the fabric of daily life, the nature, intensity, and consequences of media dependency evolve in ways that earlier theoretical formulations could not fully anticipate. Three developments have proven especially consequential: the platformization of the media environment, in which a small number of infrastructural platforms penetrate ever more sectors of social life (Poell et al., 2019; Van Dijck et al., 2018); the rise of algorithmic curation and recommender systems as the primary gatekeepers of information flows (DeVito, 2017; Thorson and Wells, 2016); and the diffusion of AI-mediated communication, in which intelligent agents increasingly compose, filter, and generate the content on which audiences rely (Hancock et al., 2020; Sundar, 2020). Understanding how dependency operates in this transformed landscape is consequential not only for communication theory but also for public policy, platform governance, media literacy education, and individual well-being.

Against this backdrop, three specific gaps in the extant literature motivate the present review. First, an integration gap: although the Uses and Dependency Model (Ball-Rokeach and DeFleur, 1976; Rubin and Windahl, 1986), Internet dependency adaptations (Ognyanova and Ball-Rokeach, 2015; Park, 2013), and the New Media Uses and Dependency Effect Model (Cho, 2009) share a common intellectual lineage, they have developed largely in parallel, dispersed across journal articles, dissertations, and encyclopedia entries (Jung, 2017). No published synthesis systematically traces their conceptual continuities and differences or consolidates them into a unified framework, leaving researchers without an integrated architecture for cumulative theory building. Second, a currency gap: the most recent dependency model dates from an era before platformization, algorithmic recommendation, and generative AI reshaped how information reaches audiences. Emerging work on algorithm dependency (Schaetz et al., 2025), algorithmic app dependence (Qiao et al., 2024), and generative AI dependency (Goh et al., 2025) is accumulating rapidly but piecemeal, without connection to the theory's five-decade intellectual tradition. Third, a critical gap: existing overviews of dependency theory are predominantly affirmative, and no review has consolidated the theory's conceptual limitations, measurement problems, contradictory findings, and unresolved debates—including the longstanding question of whether Communication Infrastructure Theory has superseded media system dependency for networked environments (Ball-Rokeach et al., 2001; Kim and Ball-Rokeach, 2006).

These gaps matter for scholars and practitioners alike. For media effects researchers, the absence of an integrated framework fragments measurement practices and impedes the accumulation of comparable findings across platforms, populations, and eras. For scholars of platforms and algorithms, dependency theory offers a ready-made relational vocabulary for phenomena—lock-in, habitual scrolling, reliance on recommender systems—currently theorized in disconnected literatures. For practitioners, the stakes are equally concrete: platform designers need theory-based criteria for distinguishing adaptive from exploitative dependency; regulators drafting platform governance instruments require an account of how design choices translate into structural reliance (Gorwa, 2019; Helberger, 2019); and media literacy educators and mental health professionals need conceptual tools for addressing problematic reliance on social media and AI systems (Goh et al., 2025; Wangqu et al., 2024).

The purpose of this article is therefore twofold: to provide a systematic theoretical review of the dependency tradition from 1976 through 2026, and to advance an original integrative contribution. Specifically, the review makes four contributions. First, it offers the first systematic synthesis connecting the foundational Uses and Dependency Model, Internet dependency adaptations, and the New Media Uses and Dependency Effect Model, using a transparent, documented review method (Section 2). Second, it extends the dependency tradition to the contemporary media environment by explicitly connecting dependency constructs to platformization, algorithmic curation, recommender systems, AI-mediated communication, and platform governance (Section 8). Third, it develops the Integrated Media Dependency Framework (IMDF)—a multi-level conceptual model, presented in Figure 1 and articulated through seven testable propositions, that incorporates a newly theorized algorithmic-agent layer of dependency and is explicitly positioned relative to Communication Infrastructure Theory, Media Repertoire Theory, and Affordance Theory (Section 10). The section specifies which elements of the framework are conserved from prior models and which are genuinely novel, and why the novel elements cannot be supplied by existing theories (Section 10.5). Fourth, it derives from the framework a forward-looking research agenda with emerging trends and concrete research questions (Section 11, Table 1) and a set of theoretical and practical implications addressed to specific stakeholder groups (Section 12).

Table 1. Critical synthesis: disposition of identified limitations and tensions in the integrated Media dependency framework.

Compiled by the authors. P1–P7 refer to the propositions in Section 10.3; Loops A and B to Figure 1. The third column records tensions that the framework restates more precisely or converts into empirical questions rather than resolves.

The review proceeds as follows. Section 2 describes the review method. Sections 3 through 5 present the three focal models: the foundational Uses and Dependency Model, Internet Use and Dependency adaptations, and the New Media Uses and Dependency Effect Model. Section 6 provides a comparative analysis across the three frameworks, and Section 7 synthesizes empirical evidence across application domains. Section 8 connects dependency theory to the platformized, algorithmic, and AI-mediated media environment. Section 9 offers a critical evaluation, including contradictory findings, unresolved debates, and adjacent theoretical perspectives. Section 10 presents the Integrated Media Dependency Framework. Section 11 proposes future research directions, Section 12 discusses overarching themes and implications for theory and practice, and Section 13 concludes.

2 Review approach and method

2.1 Review design

This article is a theoretical (integrative) review rather than a meta-analysis: its aim is conceptual synthesis and theory development, not the statistical aggregation of effect sizes. Nevertheless, to address the transparency expectations increasingly applied to review scholarship, the literature search and selection process followed the logic of PRISMA-style systematic reporting, adapted for a theoretical review whose corpus includes conceptual works, empirical studies, dissertations, and book chapters. The procedures below document how the reviewed literature was identified, screened, and analyzed so that readers can assess the comprehensiveness and potential biases of the synthesis.

The adaptation of PRISMA merits explicit statement. The review follows the PRISMA 2020 reporting logic (Page et al., 2021) at the level of process transparency rather than through wholesale adoption of a checklist designed for effect-size synthesis. Four PRISMA elements were retained in full: explicit eligibility criteria (Section 2.3); named information sources with search dates and complete, database-specific search strings (Section 2.2; Appendix A, Table A1); a documented two-stage screening and selection process with a PRISMA-style flow diagram (Section 2.4; Appendix A, Figure A1); and a structured statement of the synthesis method (Section 2.5). Three PRISMA elements were deliberately modified or omitted as inapplicable to a theory-development review: statistical risk-of-bias assessment was replaced by an appraisal of each source's conceptual contribution and evidentiary standing (peer-review status, design, and sample transparency); effect-size extraction and quantitative synthesis were replaced by the configurative narrative synthesis described in Section 2.5; and certainty-of-evidence grading was replaced by the explicit flagging of contradictory findings and unresolved debates (Section 9.3). No review protocol was prospectively registered, as prospective registration platforms for theoretical reviews remain unavailable; this is acknowledged as a limitation of the review's reproducibility.

2.2 Databases and search strategy

Literature was identified through structured searches of four databases selected for their coverage of communication scholarship and adjacent social science fields: Web of Science (Core Collection), Scopus, EBSCO Communication & Mass Media Complete, and Google Scholar (the latter used both for supplementary retrieval and for forward citation chasing of landmark works). Google Scholar's role was deliberately circumscribed and supplementary. Because its ranking algorithm is neither transparent nor reproducible and its index includes non-scholarly material, it was not treated as a primary database; it was used (a) to retrieve doctoral and master's theses—the form in which two of the three focal models were first elaborated—which the primary databases index incompletely, and (b) for forward citation chasing on the founding statements and the three focal models. For each Google Scholar query, the first 200 relevance-ranked results were screened, a stopping rule adopted because retrieval precision degrades sharply beyond that threshold. Records from all four sources were exported into a single reference-management library and deduplicated through automated DOI and title–year matching followed by manual inspection; where a Google Scholar record duplicated a database record, the database version was retained as the canonical entry. The search covered publications from January 1976—the year of the theory's founding statement—through June 2026. The core search string combined theory terms with model terms: (“media dependency” OR “media system dependency” OR “media-system dependency” OR “uses and dependency” OR “Internet dependency” OR “SNS dependency” OR “new media dependency” OR “social media dependency”) AND (theory OR model OR framework OR effects). Supplementary strings targeted the contemporary extensions examined in Section 8: (“algorithm dependency” OR “algorithmic dependence” OR “AI dependency” OR “generative AI dependence”) AND (media OR communication OR news). The complete strings as executed in each source—including field codes, limiters, and date ranges—are reproduced verbatim in Appendix A (Table A1) to facilitate replication. Backward citation chasing from the founding statements (Ball-Rokeach and DeFleur, 1976; Rubin and Windahl, 1986) and forward citation chasing from the three focal models completed the search. Database searches were conducted in the Web of Science Core Collection (Social Sciences Citation Index and Emerging Sources Citation Index), Scopus, and EBSCO Communication & Mass Media Complete, and were last updated on 15 June 2026.

2.3 Inclusion and exclusion criteria

Records were eligible for inclusion if they (a) explicitly engaged media dependency theory, media system dependency theory, the uses and dependency model, or a named derivative (Internet dependency, SNS dependency, new media dependency) as a theoretical framework—not merely as a passing citation; (b) constituted scholarly work subjected to editorial or examination review (peer-reviewed journal articles, scholarly books and chapters, and doctoral or master's theses, the latter included because two of the three focal models were first elaborated in thesis form); and (c) were available in English, or in another language with an English abstract sufficient for screening (one Chinese-language study, Xin-zhou, 2004, was retained for its historical value as an early comparative test). Records were excluded if they (a) addressed clinical “Internet addiction” or “problematic use” without engaging dependency-theoretic constructs; (b) used “dependency” in the economic or international-relations sense; (c) were non-scholarly sources (trade press, blogs); or (d) reported no identifiable theoretical or empirical engagement with dependency constructs. Consistent with the journal's requirements for review articles, only published sources are cited; no unpublished or original data are reported.

2.4 Screening and study selection

Screening proceeded in two stages. In the first stage, titles and abstracts of deduplicated records were screened against the inclusion criteria. In the second stage, full texts of the remaining records were assessed, with particular attention to whether dependency constructs did substantive theoretical work in the study design. Screening decisions were made by the authors, with disagreements resolved through discussion until consensus was reached. The searches identified 1,258 records in total: 1,029 through the three primary databases (Web of Science, n = 348; Scopus, n = 462; EBSCO Communication & Mass Media Complete, n = 219) and 229 through supplementary methods (Google Scholar supplementary retrieval, n = 142; backward citation chasing, n = 36; forward citation chasing, n = 51). After removal of 407 duplicates, 851 unique records were screened at the title-and-abstract stage, of which 596 were excluded. The full texts of the remaining 255 records were assessed against the eligibility criteria, and 172 were excluded for the following reasons: dependency invoked only as a passing citation without theoretical engagement (n = 81); clinical Internet-addiction or problematic-use framing without dependency-theoretic constructs (n = 44); use of “dependency” in the economic or international-relations sense (n = 29); and non-scholarly source type (n = 18). The final corpus therefore comprises 83 sources. Figure A1 (Appendix A) presents the identification, screening, and inclusion flow in PRISMA-style diagram form. The final corpus comprises the foundational statements of the three focal models, their principal empirical applications across political, environmental, migration, consumer, mental health, and interpersonal domains, critical and integrative commentaries, and the recent phantomization, algorithmic, and AI-related extensions synthesized in Section 8.

2.5 Analytic approach

The included literature was analyzed using a configurative narrative synthesis oriented to theory development. For each focal model, the analysis extracted (a) core propositions and assumptions, (b) key constructs and their operationalizations, (c) levels of analysis, (d) empirical findings, and (e) critiques and boundary conditions. These extractions were organized into construct-mapping matrices that underlie the comparative tables presented throughout the article (Tables 1–9). The synthesis then proceeded abductively: continuities and discontinuities across the three models were identified, contradictory findings were flagged rather than smoothed over (Section 9.3), and the integrative framework presented in Section 10 was constructed to preserve constructs with consistent empirical support while adding elements required to account for phenomena—algorithmic curation, AI-mediated communication, polymedia repertoires—that the existing models do not address. This approach follows established practice for theory-generating reviews in communication and management scholarship, in which the contribution lies in conceptual integration rather than exhaustive enumeration.

Table 2. Core theoretical elements of the foundational uses and dependency model.

Adapted from Ball-Rokeach and DeFleur (1976).

Table 3. Operational adaptations in internet Use and dependency research, 2004–2026.

Compiled from the cited sources. Entries are ordered chronologically; studies from 2024 to 2026 document the extension of dependency operationalizations to algorithmic and AI-mediated contexts discussed in Section 8.

Table 4. Comparative analysis of three Media dependency models.

Compiled from Ball-Rokeach and DeFleur (1976), Ognyanova and Ball-Rokeach (2015), Park (2013), Schrock (2006), Cho (2009), and Kim and Jung (2017).

Table 5. Theoretical continuities and differences across the three dependency frameworks.

Compiled from Ball-Rokeach and DeFleur (1976), Cho (2009), Ognyanova and Ball-Rokeach (2015), and related studies.

Table 6. Empirical findings across application domains.

Compiled from the cited sources.

Table 7. Theoretical integrations with Media dependency theory.

Compiled from the cited sources.

Table 8. Operational summary of the integrated Media dependency framework.

Proposed by the authors on the basis of the integrative review; P1–P7 refer to the propositions in Section 10.3 and the loops to Figure 1.

Table 9. Proposed future research directions, emerging trends, and illustrative research questions for Media dependency theory.

Proposed by the authors based on the identified gaps in the reviewed literature. P-numbers refer to the propositions of the Integrated Media Dependency Framework (Section 10.3).

3 The uses and dependency model: theoretical foundations

3.1 Core propositions and assumptions

Ball-Rokeach and DeFleur (1976) introduced the dependency model of mass media effects as an alternative to both direct-effects and limited-effects paradigms that had dominated communication research. The model's foundational proposition holds that the magnitude and nature of media effects depend on the intensity of audience dependency on media information resources—an intensity shaped by tripartite relationships among audiences, media systems, and social structures. By locating the source of media influence in structural relationships rather than solely in message characteristics or audience predispositions, the model advanced a distinctly relational account of communication consequences. The subsequent merger of this dependency logic with uses and gratifications thinking—formalized by Rubin and Windahl (1986) as the “uses and dependency model”—explicitly connected individual motives for media use with the structural dependency relations theorized by Ball-Rokeach and DeFleur, and it is this merged formulation that gives the model reviewed in this section its name. Ball-Rokeach (1985) further specified the sociological origins of individual media-system dependency, arguing that individual dependency relations are structured by the information goals persons develop as members of social categories and networks, and later restated the framework as a comprehensive theory of media power and media use (Ball-Rokeach, 1998).

Three assumptions distinguish this framework from its predecessors. First, a systemic focus holds that media effects are conditional on the functions media perform within broader social structures, directing analytical attention to the roles media play in information systems and to whether functional alternatives exist. Second, context sensitivity posits that dependency intensity—and therefore the potential for media effects—increases during periods of rapid social change, pervasive conflict, or structural instability, when established information sources become unreliable and audiences turn more heavily to media. Third, relational causality holds that effects emerge from the strength and nature of dependency relations rather than from one-way persuasion or transmission (Ball-Rokeach and DeFleur, 1976). Together, these assumptions reframe media effects as emergent properties of goal-directed dependency relations embedded in social contexts, explicitly rejecting accounts that treat content, channels, or demographics in isolation.

The model's systemic character deserves further elaboration because it differentiates the dependency approach from psychologically oriented media effects theories. Rather than treating individual audience members as the sole locus of analysis, the model situates individuals within nested social systems where media organizations, political institutions, economic structures, and cultural norms collectively shape the information environment. This sociological grounding ensures that the model can account for structural variations—such as differences in media plurality, state censorship regimes, or economic pressures on media organizations—that condition the availability and character of media information resources. The result is a framework simultaneously attuned to individual-level cognitive and behavioral processes and to the macro-level social structures within which those processes unfold.

3.2 Key theoretical constructs

The foundational model operationalizes dependency through several core constructs. Dependency relations refer to the intensity of reliance between audiences and media systems, conceptualized as a continuum from low to high (Ball-Rokeach and DeFleur, 1976). High dependency exists when audiences perceive media as essential or irreplaceable for achieving important goals; low dependency characterizes situations where functional alternatives are readily available or where media are peripheral to goal attainment.

Functional alternatives encompass other information resources—interpersonal networks, institutional sources, direct experience—that can substitute for media in fulfilling information needs. Their availability inversely moderates dependency intensity: when alternatives are scarce or inaccessible, dependency intensifies. This construct reflects the model's ecological perspective, recognizing that media operate within broader information environments rather than in a vacuum. The concept of functional alternatives is theoretically important because it introduces competition among information sources as a structural determinant of media influence, anticipating later work on media choice, repertoires, and polymedia environments (Hasebrink and Popp, 2006; Madianou and Miller, 2013).

The model specifies three nested levels of analysis at which dependency relations operate: individual (micro), media system (meso), and social organization (macro). At the micro level, individual audience members develop dependencies based on personal goals and circumstances. At the meso level, media organizations structure content and distribution to fulfill societal information functions. At the macro level, social systems allocate resources and legitimacy to media institutions, shaping their capacity to meet dependency needs. This multi-level architecture underscores that dependency is simultaneously a psychological, organizational, and societal phenomenon (Ball-Rokeach and DeFleur, 1976). Importantly, effects at one level cascade across levels: macro-level political crises increase micro-level audience dependency, which in turn can reshape meso-level media content priorities as organizations respond to heightened demand for certain types of information. Empirical work in the foundational tradition supports this contextual logic: perceptions of threat intensify individual media-system dependency relations (Loges, 1994), and large-scale social disruptions such as the September 11 attacks produce measurable surges in dependency on established media (Lowrey, 2004).

3.3 Dependency goals: understanding, orientation, and play

A central contribution of the foundational model is its specification of three primary goals motivating audience dependency: understanding, orientation, and play (Ball-Rokeach and DeFleur, 1976). Understanding encompasses personal and social dimensions—making sense of one's experiences and comprehending broader social events, trends, and collective meanings. Orientation refers to action-directed goals, including personal action orientation (guiding individual decisions and behaviors) and social action orientation (participating effectively in collective activities). Play encompasses fantasy, escape, and entertainment goals, spanning solitary diversion and interactive social entertainment.

These categories are not mutually exclusive: a single media experience may simultaneously serve multiple goals. A news broadcast may help a viewer make sense of current events (understanding), guide civic participation (orientation), and offer distraction from daily stress (play). The typology nonetheless provides a systematic framework for analyzing why audiences develop dependencies on media and for predicting which effects are most likely under varying dependency conditions. The goal typology also facilitates comparative analysis across media types and historical periods, since different media environments may be differently suited to fulfilling goals—a point that becomes especially relevant when considering the transition from broadcast to digital media.

As digital technologies have proliferated, scholars have revisited these dependency goals to examine how new media environments amplify, diversify, or complicate traditional patterns of use. Social media platforms frequently blur the boundaries between understanding, orientation, and play by delivering news, advice, and entertainment within a single feed, challenging researchers to refine measurement strategies and theoretical frameworks accordingly. The algorithmic curation characteristic of contemporary platforms further complicates this picture, as platform design decisions—rather than audience choices alone—increasingly determine which goals are served and how dependency relations are structured (Thorson and Wells, 2016; see Section 8).

4 Internet use and dependency models

4.1 Affordance shifts in digital environments

The emergence of the Internet introduced fundamental changes in media affordances that required theoretical adaptation. Unlike traditional broadcast media—characterized by one-to-many communication, scheduled programming, and limited interactivity—Internet media offer perpetual availability, many-to-many communication, user-generated content, and interactive capabilities (Park, 2013). These affordance shifts expanded the kinds of dependency relations scholars could examine while transforming the speed, scale, and nature of information flows.

The Internet's ubiquitous availability means audiences can access information at any time and from multiple locations, potentially intensifying dependency by embedding media more deeply into daily routines (Carillo et al., 2014). Interactivity enables audiences to seek, select, and manipulate information rather than passively receiving broadcast content, altering the character of dependency from reception to participation (Park, 2013). The multiplicity of online resources—news sites, social platforms, specialized forums, and user-generated content—creates a complex information ecology in which dependency may be distributed across numerous platforms rather than concentrated in a few mass media outlets (Kim and Jung, 2017; Wangqu et al., 2024).

The concept of affordances is central to understanding these shifts. Affordances, understood as the action possibilities that a technology provides to its users, differ qualitatively across media types (Evans et al., 2017). Broadcast media offers passive reception and collective audience experiences; Internet media afford search, personalization, interactivity, content creation, and social connection. These affordance differences do not merely add features to the dependency equation; they transform the nature of the dependency relation itself. When audiences can actively shape their information environments through search queries, algorithmic feedback, and content creation, the boundary between information provider and information consumer becomes porous, and dependency becomes a more reciprocal and co-constructed phenomenon than the original broadcast-centric model envisioned.

These shifts raise important questions about the foundational model's applicability. Some scholars contend that the model's core logic—effects depend on goal-directed reliance on information resources—remains valid but demands operational adaptations for platform-specific attributes and interactive use patterns (Ognyanova and Ball-Rokeach, 2015). Others argue that the transition from centralized broadcast systems to distributed, networked environments may require more fundamental theoretical revision or integration with frameworks such as Communication Infrastructure Theory (Alquaary et al., 2026; Ball-Rokeach et al., 2001; see Section 9.4).

4.2 Operational adaptations for online contexts

Researchers have adapted dependency theory for Internet contexts by operationalizing dependency at the level of specific platforms, tools, and perceived attributes rather than at the level of media systems broadly conceived. Ognyanova and Ball-Rokeach (2015) applied media dependency theory to online political efficacy, measuring dependency through perceived comprehensiveness and credibility of online news sources. They found that Internet dependency intensity was predicted by perceived comprehensiveness but not credibility, and that online dependency related to perceptions of the Internet's democratic potential. This shift from systemic media functions to audience perceptions of platform utility and quality constitutes a meaningful operational advance.

Schrock (2006) investigated MySpace dependency among college students, finding that dependency moderately correlated with use intensity—members actively used the platform an average of 1.3 h daily, primarily for play and interaction orientation goals—and that extroversion and self-disclosure positively predicted dependency intensity. Xin-zhou (2004) provided early comparative evidence, finding that Internet media establish considerably closer and higher dependent relationships between audiences and the medium than traditional media, suggesting that digital affordances amplify dependency intensity through mechanisms warranting further investigation.

These operational adaptations collectively represent a move toward greater specificity in dependency measurement. Rather than asking whether audiences depend on “the media” broadly, researchers began asking which attributes of which platforms predict dependency and with what consequences. This specificity enhances predictive power and practical relevance, enabling recommendations about platform design, media policy, and individual media management that generic system-level measures could not support.

4.3 Platform-level dependency relations

A pivotal innovation in Internet dependency research is the shift from system-level to platform-level analysis. Rather than treating the Internet as a monolithic medium, scholars increasingly examine dependency on specific platforms—social networking sites, search engines, news aggregators, messaging applications—each possessing distinct affordances, content types, and use patterns (Kim and Jung, 2017). This granular approach recognizes that audiences may develop strong dependencies on some platforms while remaining relatively independent of others.

Kim and Jung (2017) developed and validated a social networking site (SNS) dependency measure, finding that SNS dependency directly affects users' engagement with interactive platform activities and exerts indirect effects on offline interpersonal storytelling. The platform-level approach also facilitates examination of platform-swinging—navigating multiple platforms for different purposes based on situational needs. Wangqu et al. (2024) investigated media dependency, platform-swinging, and psychological depression among college students in a polymedia environment, finding that platform-swinging affordances influence media dependency, which in turn shapes psychological depression.

The construct of platform-swinging is theoretically noteworthy because it captures a behavioral pattern distinctive to contemporary media environments that has no clear analog in the broadcast era. Audiences now routinely shift among platforms—from a microblogging service for news, to a video-sharing platform for entertainment, to a messaging application for interpersonal coordination—in rapid succession and often within the same session. This behavior complicates the measurement of dependency because an individual's overall media dependency may be high even if dependency on any single platform is moderate. Researchers must therefore attend to both platform-specific and aggregate dependency levels, and to the strategic logic by which audiences distribute their reliance across an ecosystem of platforms.

The shift to platform-specific analysis reflects the fragmentation and personalization of contemporary media landscapes. Audiences no longer depend on a single medium for all informational, orientational, and entertainment needs; instead, they cultivate tailored dependencies across platforms optimized for distinct goals. The evidence that dependency intensity is often amplified in digital contexts compared to traditional media—attributable to ubiquity, interactivity, and content diversity—highlights how the Internet fosters deeper media integration into everyday life. That perceived comprehensiveness predicts dependency more strongly than credibility points to the importance of perceived scope in sustaining engagement, raising important questions about how information abundance shapes dependency formation independently of source trustworthiness. The most recent wave of operational adaptations, summarized in the lower rows of Table 2, indicates that the object of dependency is migrating once more: from platforms as such toward the algorithmic curation systems and generative AI agents embedded within and beyond them (Goh et al., 2025; Qiao et al., 2024; Schaetz et al., 2025)—a development taken up in Section 8.

5 The new media uses and dependency effect model

5.1 Theoretical innovations

Cho (2009) introduced the New Media Uses and Dependency Effect Model as an explicit theoretical advance integrating habitual new media use, dependency relations, and downstream effects into a unified causal framework. While Internet dependency research adapted the foundational model for digital contexts, Cho's model advanced theory by specifying habitual use as a distinct antecedent that interacts with dependency relations to produce cognitive, affective, and behavioral outcomes—addressing a gap in earlier formulations that did not theorize how routine media behaviors relate to dependency intensity.

Three innovations distinguish the model. Habit integration treats habitual media use as a precursor to dependency rather than a consequence or correlation. Outcome pathway specification articulates a mediation sequence from use habit through dependency to outcomes. Platform specificity encourages operationalization at the service or SNS level rather than exclusively at the mass media system level (Cho, 2009). Cho conceptualized new media use habits along three dimensions: ritualized use (routine, habitual consumption integrated into daily life), instrumental use (goal-directed, purposive information seeking), and participatory use (active content creation and social interaction). The model proposes that these patterns predict dependency intensity, which mediates effects on outcomes such as consumer behavior and prosocial behavior.

The distinction between ritualized, instrumental, and participatory use is analytically important because it captures qualitatively different modes of engagement with media that may have distinct implications for dependency formation and outcomes. Ritualized use reflects embedded, automatic patterns of consumption that may foster dependency through sheer frequency of exposure and behavioral reinforcement. Instrumental use reflects deliberate, goal-directed engagement that may foster dependency through perceived utility and satisfaction of informational needs. Participatory use reflects active contribution and interaction that may foster dependency through social investment and the creation of network ties. By differentiating these modes, Cho's model enables more precise prediction about which types of use are most likely to produce dependency and under what conditions.

5.2 Habit as a precursor to dependency

The integration of habit as a precursor to dependency constitutes a significant theoretical advancement. The foundational model conceptualized dependency primarily as a function of goal importance and the availability of functional alternatives, with comparatively little attention to how routine use patterns shape dependency intensity over time. Cho (2009) proposes that habitual use—the extent to which media consumption becomes automatic, embedded in daily routines, and performed with minimal conscious deliberation—directly influences the strength of dependency relations.

This theoretical move draws on evidence that repeated behaviors in stable contexts become habitual, reducing cognitive effort and increasing behavioral persistence. When new media use becomes habitual, audiences develop stronger dependencies because media become more deeply integrated into goal-pursuit strategies and daily life structures. Ritualized use, in particular, may foster dependency by making media consumption a routine component of understanding, orientation, and play activities (Cho, 2009). Empirical support for this pathway emerged from Cho's (2009) study of Korean college students, which found that ritualized and instrumental use were strong predictors of new media dependency. New media dependency explained 32% of consumer behavior outcome variance and 27% of prosocial behavior outcome variance—notable effect sizes in media effects research—and structural equation modeling confirmed that dependency partially mediates the effects of ritualized and instrumental use on outcomes rather than serving as a redundant proxy for use intensity.

The habit-dependency linkage also has implications for understanding problematic media use. If habitual patterns are the primary driver of dependency, then interventions aimed at reducing excessive dependency may need to target habit formation processes—contextual cues, behavioral triggers, and reinforcement schedules—rather than dependency itself. This practical implication aligns with emerging research in behavioral science on habit interruption and suggests that dependency theory can inform intervention design when integrated with habit-change frameworks.

5.3 Mediation pathways and outcome mechanisms

The model specifies mediation pathways linking habitual use to outcomes through dependency relations, enabling researchers to test whether dependency fully or partially mediates use-habit effects and to identify conditions under which direct effects persist alongside mediated ones. Cho (2009) proposes that new media use habits influence dependency intensity, which affects cognitive outcomes (knowledge, beliefs, attitudes), affective outcomes (emotions, feelings, moods), and behavioral outcomes (actions, intentions, practices). This mediation framework clarifies causal sequencing by distinguishing antecedents, mediators, and outcomes; enables decomposition of total effects into direct and indirect components; and facilitates identification of moderators that strengthen or weaken mediation pathways.

Kim and Jung (2017) extended this mediation logic by demonstrating that SNS dependency directly affects online interactive engagement and indirectly influences offline interpersonal storytelling, providing further empirical support for the mediation pathways central to Cho's (2009) model. The extension to offline storytelling is particularly notable because it demonstrates that digital dependency relations do not remain confined within digital environments but reverberate through face-to-face communication, reshaping how individuals narrate their experiences, share information, and construct social meaning in interpersonal settings.

6 Comparative analysis across the three models

6.1 Theoretical continuities

Despite adaptation to radically different media environments over five decades, the three dependency frameworks share fundamental theoretical continuities. All retain the central proposition that dependency intensity is the primary determinant of media effects, with stronger dependency increasing the likelihood that media alters beliefs, feelings, or behaviors (Ball-Rokeach and DeFleur, 1976; Cho, 2009; Ognyanova and Ball-Rokeach, 2015). All preserve a goal-oriented foundation, conceptualizing audience reliance on media as motivated by fundamental needs for understanding, orientation, and play (Ball-Rokeach and DeFleur, 1976; Cho, 2009; Schrock, 2006). While operationalizations have become more granular and platform-specific, the underlying assumption that audiences depend on media to achieve important goals remains constant across all three frameworks.

All three frameworks maintain a systemic perspective that views media effects as emergent from relationships among audiences, media, and social contexts rather than as direct consequences of message exposure (Ball-Rokeach and DeFleur, 1976; Kim and Jung, 2017). Even as the unit of analysis has shifted from media systems to platforms and services, scholars continue to emphasize that dependency relations are shaped by broader information ecologies, functional alternatives, and social conditions—a systemic orientation that guards against reductionist accounts that ignore contextual moderators. This theoretical persistence is itself a notable finding of the review: it suggests that the foundational model's core logic possesses sufficient generality to accommodate radical changes in the media environment without requiring abandonment of its central propositions.

6.2 Key differences and advancements

The three models exhibit important differences reflecting theoretical advances in response to evolving media environments. The primary difference concerns the level of analysis. The foundational model emphasized system-level relations between audiences and mass media institutions, reflecting the centralized, broadcast-dominated environment of the 1970s (Ball-Rokeach and DeFleur, 1976). Internet adaptations shifted toward platform-level analysis, recognizing that digital environments comprise multiple platforms with distinct affordances (Ognyanova and Ball-Rokeach, 2015; Schrock, 2006). The New Media Uses and Dependency Effect Model further refined this focus by operationalizing dependency at the service or SNS level and incorporating micro-level habitual use patterns as antecedents (Cho, 2009; Kim and Jung, 2017).

A second difference involves the mechanism driving dependency. The foundational model centered on social context and functional alternatives as primary determinants (Ball-Rokeach and DeFleur, 1976). Internet adaptations foregrounded platform affordances—comprehensiveness, credibility, interactivity, and availability (Ognyanova and Ball-Rokeach, 2015; Park, 2013). The New Media model introduced habit as a distinct antecedent, proposing that routine, embedded use patterns directly influence dependency (Cho, 2009). A third difference concerns causal specification. The foundational model articulated general relationships between dependency and effects without detailing causal pathways or mediation mechanisms. Internet adaptations began testing more specific hypotheses about relationships between perceived attributes, dependency, and outcomes. The New Media model explicitly specified and empirically tested mediation pathways—habit → dependency → outcomes—enabling decomposition of direct and indirect effects (Al-Makhzoomy et al., 2026; Cho, 2009; Kim and Jung, 2017). This progression toward precise causal modeling reflects broader trends in communication scholarship toward rigorous theory testing and mechanism identification.

6.3 Levels of analysis and measurement

The evolution across models reflects a progressive incorporation of multiple analytical levels and measurement strategies. The foundational model operated primarily at macro and micro levels, examining how societal structures and media system characteristics shape dependencies and effects (Ball-Rokeach and DeFleur, 1976). Internet adaptations introduced meso-level platform analysis, measuring dependency in relation to specific platforms with defined affordances, content types, and user bases (Carillo et al., 2014; Ognyanova and Ball-Rokeach, 2015; Schrock, 2006). The New Media Uses and Dependency Effect Model further incorporated micro-level individual processes—particularly habitual use patterns and their psychological underpinnings—while expanding measurement strategies to include assessments of ritualized, instrumental, and participatory use alongside individual differences in personality, self-disclosure, and computer self-efficacy (Cho, 2009; Schrock, 2006). This multi-level approach recognizes that dependency operates simultaneously at individual, platform, and societal levels, with cross-level interactions shaping outcomes in ways that no single level of analysis can fully capture.

7 Empirical evidence and applications

7.1 Foundational model applications

Empirical applications of the foundational Uses and Dependency Model have concentrated primarily on media effects during periods of social upheaval, crisis, or rapid institutional change, directly operationalizing the model's proposition that dependency intensifies under conditions of structural instability. Ball-Rokeach and DeFleur (1976) articulated the expectation that dependency would be highest where media fulfill many central information functions and during periods when established interpersonal or institutional channels become unreliable. Early empirical tests explored how these dynamics manifested during elections, natural disasters, and large-scale political transitions, consistently demonstrating that audiences who perceived media as their primary or irreplaceable information source exhibited stronger cognitive, affective, and behavioral effects. Loges (1994) showed that perceptions of threat in the social environment intensify the scope and intensity of individuals' media-system dependency relations, and Lowrey (2004) documented sharply elevated dependency on established news media in the aftermath of the September 11 attacks—two findings that anchor the model's context-sensitivity proposition in systematic evidence.

The conceptual framework of understanding, orientation, and play goals proved valuable because it enabled differentiated predictions about which effects would emerge under dependency conditions. Understanding-oriented dependency produced cognitive effects—increased knowledge, altered beliefs, and revised worldviews. Orientation-driven dependency produced behavioral effects such as changes in voting intentions or civic participation. Play-oriented dependency associated primarily with affective outcomes, including emotional regulation and mood enhancement (Ball-Rokeach and DeFleur, 1976). A critical insight from this early tradition is that dependency strength was not uniform across populations or stable over time; rather, it fluctuated in response to social conditions, the perceived quality of alternative information resources, and the specific goals audiences sought to fulfill—reinforcing the model's systemic orientation and demonstrating that media effects cannot be predicted from message content or channel characteristics alone.

7.2 Internet dependency studies

The application of dependency theory to Internet and digital media contexts produced a second wave of empirical work that confirmed the foundational model's core logic while revealing operational adaptations required by the shift to interactive, perpetually available, and user-driven information environments.

Ognyanova and Ball-Rokeach (2015) applied media system dependency theory to online news consumption and political efficacy, measuring dependency through audience perceptions of online news comprehensiveness and credibility. A key finding was that Internet dependency intensity was predicted by perceived comprehensiveness—that online news provided broad, thorough coverage—but not by perceived credibility. This asymmetry suggests that in information-rich digital environments, the scope of available content may matter more for fostering dependency than assessments of trustworthiness. The study further documented that online political efficacy was linked to individual perceptions of new media's comprehensiveness and the perceived capacity of online tools to maintain ideologically homogeneous social networks (Ognyanova and Ball-Rokeach, 2015). Xin-zhou (2004) provided comparative evidence finding that Internet media establish considerably closer and higher dependent relationships than traditional media, attributing this pattern to interactive capabilities, personalization features, and round-the-clock availability.

Schrock's (2006) investigation of MySpace dependency among 401 undergraduate students found that dependency moderately correlated with use frequency and duration. Extroversion and self-disclosure positively predicted dependency, while technology-related variables—Internet access speed, computer self-efficacy, and computer anxiety—did not significantly correlate with dependency intensity. This null finding is theoretically important because it suggests that dependency formation in digital environments is driven more by psychosocial characteristics and relational motivations than by technological access or skill. The study also confirmed that play and interaction orientation were the primary goals motivating MySpace dependency, extending the foundational goal typology to a social networking context and demonstrating its continued relevance in interactive digital environments.

Carillo et al. (2014) advanced this line of inquiry by examining dependency in predicting continuing intention to use smartphones. Among 150 users, dependency inflated perceived usefulness and perceived ease of use—constructs central to technology acceptance models—and these inflated perceptions affected continuance intentions. This finding introduces a theoretically significant feedback mechanism: dependency not only results from perceptions of media utility but also actively distorts those perceptions, creating a self-reinforcing cycle in which dependent users evaluate their chosen media more favorably than objective performance might warrant. This feedback loop has important implications for understanding how dependency is maintained and intensified over time, as biased perceptions reduce the likelihood that users will seek or appreciate functional alternatives.

7.3 Social networking platform research

Research on social networking platforms constitutes the most active and empirically productive area of contemporary dependency scholarship, validating the New Media Uses and Dependency Effect Model's propositions about habitual use, mediation pathways, and platform-specific dependency. Cho's (2009) foundational test among Korean college students aged 17–27 remains the most comprehensive empirical evaluation of the model's core propositions. Ritualized and instrumental use were strong predictors of new media dependency, confirming that routine, embedded use patterns serve as meaningful antecedents to dependency formation. Participatory use showed a weaker relationship, which Cho interpreted as reflecting the relatively nascent state of participatory behaviors at the time of data collection. New media dependency accounted for 32% of variance in consumer behavior outcomes and 27% of variance in prosocial behavior outcomes—noteworthy effect sizes in media effects research—and structural equation modeling confirmed that dependency partially mediates the effects of ritualized and instrumental use on outcomes rather than serving as a redundant proxy for use intensity.

Kim and Jung (2017) extended the mediation logic in their study of 477 SNS users in Seoul, finding that SNS dependency directly enhances users' engagement with interactive platform features—commenting, sharing, creating content—and that this online engagement indirectly stimulates offline interpersonal storytelling. This finding demonstrates that digital dependency shapes not only online behavior but also spills over into offline social life, challenging accounts that treat online and offline communication as separate domains. The study's validated platform-specific dependency scale, rather than generic media dependency measures, represented a methodological advance enabling more precise hypothesis testing.

Işik et al. (2022) surveyed 240 Turkish university students about Instagram use, with factor analysis confirming that the platform fosters dependencies organized around the foundational model's tripartite goal structure: understanding, orientation, and entertainment/play. This finding is notable because it demonstrates the continued empirical relevance of Ball-Rokeach and DeFleur's (1976) original goal typology in a platform environment—image-centered, algorithmically curated, and interaction-driven—that differs radically from the broadcast media the foundational model was designed to explain. Beyond confirming dependency formation, the study documented substantial displacement effects: 62% of respondents reported watching less television, 54% reading fewer newspapers, and 42% listening less to radio as their Instagram dependency increased. These displacement patterns are consistent with the foundational model's concept of functional alternatives interpreted in reverse: as a platform assumes information and entertainment functions previously served by traditional media, audiences shift their dependency toward the platform more effectively fulfilling their goals.

Habib (2013) examined Twitter use in Egypt following the January 25 Revolution, finding that Twitter functioned as a primary news source during political upheaval, with users depending on the platform to satisfy surveillance needs and, for male users in particular, mobilization goals. This finding directly echoes the foundational model's proposition that dependency intensifies during social change and conflict, while demonstrating that the dynamic applies to specific social media platforms.

7.4 Cross-domain applications

The application of media dependency theory across substantive domains beyond political communication and social networking demonstrates the framework's explanatory versatility. Studies in environmental communication, refugee adaptation, consumer behavior, mental health, and interpersonal relationships confirm that the core dependency mechanism operates robustly across diverse contexts, though with domain-specific nuances that enrich theoretical understanding.

In environmental communication, Zhang and Zhong (2020) conducted a cross-national study extending media system dependency theory to informational media use and environmentalism, finding systematic associations between informational media dependency and environmental attitudes and behaviors across nations with different media systems, political structures, and environmental conditions. The cross-national design supports the foundational model's claim that dependency operates as a general mechanism of media influence rather than a culturally bound phenomenon. Youssef (2023) complemented this evidence with a focused study of Egyptian social media users, reporting that 71.5% acknowledged a relationship between their social media use and their understanding of climate change, with dependency on public figures—cited by 42.1% as a primary pathway—emerging as a key mechanism for elevating climate literacy. This finding introduces a relational specificity largely absent from earlier dependency studies: audiences may depend not on platforms generically but on specific categories of content creators whose perceived expertise or credibility makes them irreplaceable for particular goals. Liao (2025) extended this environmental line by integrating media system dependency theory with norm activation theory among 1,594 Taiwanese adults, showing that media dependency shapes pro-environmental behavior indirectly, through awareness of consequences, ascription of responsibility, and personal norms—evidence that dependency effects can operate through well-specified psychological mediators rather than direct influence. Relatedly, Al Adwan et al. (2025) found that public relations campaigns conducted through media channels contribute to environmental awareness among university students, indicating that the dependency channels documented in this domain can be deliberately mobilized for prosocial communication goals.

Refugee communication represents a domain in which dependency theory's contextual sensitivity finds particularly clear empirical expression. Aldamen (2023) conducted a mixed-methods study of Syrian refugees in Jordan and Turkey, integrating media system dependency theory with uses and gratifications theory to understand how displacement, censorship, and the need for adaptation shape social media dependency. The study documented that refugees' social media use increased significantly after leaving Syria, driven by improved Internet access and freedom from state censorship in host countries. Facebook emerged as the dominant platform, used for educational, commercial, social interaction, and information-seeking gratifications. The dependency relations refugees developed yielded measurable cognitive effects (learning about host-country institutions and languages), affective effects (maintaining emotional connections with distant family members), and behavioral effects (navigating bureaucratic systems and building social networks). The study also demonstrated how structural conditions—censorship regimes, Internet infrastructure, and the availability of alternative information channels—shaped dependency intensity, providing direct empirical support for the foundational model's macro-level propositions.

In the consumer and technology-use domain, Qiao et al. (2024) surveyed 354 Chinese users of algorithmically curated applications—news aggregators, short-video services, and social media—and found that information and expression gratifications increased algorithmic app dependence, while entertainment gratification exerted a dual influence: it raised dependence directly but simultaneously elevated fatigue, which suppressed dependence. This pattern refines the dependency tradition's account of consumer outcomes by demonstrating that gratification-driven dependency formation can be self-limiting when intensive use generates exhaustion, and it extends the feedback-loop logic identified by Carillo et al. (2014) to algorithmic environments.

Studies of mental health have added an important critical dimension by documenting conditions under which high dependency produces negative outcomes. Wangqu et al. (2024) examined media dependency, platform-swinging, and psychological depression among 1,210 Chinese college students, finding that platform-swinging affordances influence media dependency intensity and that the nature of media dependency shapes susceptibility to psychological depression. This study recognizes that contemporary dependency operates within polymedia environments where audiences manage portfolios of dependencies across multiple platforms, introducing both flexibility and complexity into dependency relations that complicates measurement and intervention. Djamaluddin et al. (2023) documented a related dynamic during the COVID-19 pandemic in Indonesia, finding that the shift to virtual communication fostered increased media dependence as physical-world information channels became inaccessible—accompanied by excessive information circulation, exposure to disinformation, and the emergence of a sensation-seeking culture. This illustrates a potential dark side of dependency: when dependency intensity increases rapidly without corresponding increases in media literacy or critical evaluation capacity, dependent audiences become more vulnerable to misinformation and manipulative content. Extending this critical line to AI-mediated environments, Goh et al. (2025) developed and validated a Generative AI Dependency Scale across six studies with 1,333 participants in the United States and Singapore, finding that generative AI dependency was associated with lower psychological need fulfillment, greater procrastination, impaired critical thinking and task performance, and heightened loneliness—and arguing that dependency reflects a misalignment between psychological needs and coping strategies rather than mere frequency of use.

In the interpersonal domain, Khalifa et al. (2026) examined the relationship between social media site use and social isolation and selfishness among youth in a predictive study, finding that intensive reliance on social media platforms predicted higher levels of social isolation and self-centered orientations. Read through a dependency lens, this evidence suggests that when platforms become the primary venue for interpersonal goals, the dependency relation can crowd out the face-to-face alternatives on which relational skills depend—an interpersonal analog of the displacement effects documented by Işik et al. (2022).

Mohamed Nawi et al. (2020) examined new media use among Malaysian youth, reinforcing the theoretical expectation that younger populations—for whom digital platforms constitute primary information and social environments—develop particularly intense dependencies that structure their information consumption, social interaction, and identity formation.

Several overarching patterns emerge from this cross-domain synthesis. The core dependency mechanism—goal-directed reliance on media shaping subsequent outcomes—operates consistently across substantively diverse contexts, supporting the foundational model's claim to generality. The specific outcomes produced by dependency vary systematically according to motivating goals: understanding-driven dependency tends to produce cognitive effects, orientation-driven dependency tends to produce behavioral effects, and play-driven dependency tends to produce affective effects, though these associations are probabilistic rather than deterministic. Individual differences (personality, demographics, cultural identity), platform affordances (interactivity, comprehensiveness, algorithmic curation), and social structural conditions (crisis, censorship, infrastructure availability) function as moderators shaping dependency intensity and the character of its consequences. High dependency can produce both beneficial outcomes—enhanced literacy, skill acquisition, social connection—and harmful outcomes—depression, relationship conflict, vulnerability to misinformation—depending on the content environment, the presence of critical evaluation skills, and the adequacy of regulatory and design safeguards.

8 Media dependency in the platformized, algorithmic, and AI-mediated environment

The three models reviewed above were formulated for broadcast systems, the open web, and first-generation social networking sites, respectively. Since then, the media environment has been reorganized around a small number of infrastructural platforms, algorithmic curation systems, and, most recently, generative artificial intelligence. This section connects each of these developments explicitly to dependency dynamics, both because reviewer-era scholarship demands it and because these developments alter the objects, mechanisms, and stakes of dependency in ways the existing models do not capture. The section thereby supplies the empirical and conceptual foundation for the algorithmic-agent layer of the integrated framework presented in Section 10. To make the theoretical stakes explicit rather than merely documentary, each subsection closes by identifying the specific assumption of the dependency tradition that the development in question unsettles, and Section 8.5 consolidates these assumption-level revisions into the cumulative case for treating algorithmic and AI systems as a distinct stratum of dependency targets.

8.1 Platformization and the reorganization of dependency targets

Platformization denotes the penetration of the economic, infrastructural, and governmental logics of digital platforms into ever more sectors of social and economic life (Poell et al., 2019; Van Dijck et al., 2018). For dependency theory, platformization matters because it restructures the macro and meso levels of the dependency relation simultaneously. At the macro level, a handful of platform companies now control the infrastructures—app stores, cloud services, identity systems, advertising markets—through which most media content circulates, concentrating the “information resources” of the foundational model in fewer institutional hands than at any point since the broadcast era (Van Dijck et al., 2018). At the meso level, individual outlets, creators, and institutions must adapt their content and distribution strategies to platform logics, meaning that the media organizations on which audiences depend are themselves dependent on platforms—a second-order dependency the original tripartite audience–media–society model did not anticipate. Audience fragmentation research shows that even in an environment of near-infinite choice, attention concentrates on a small set of dominant platforms and outlets (Webster and Ksiazek, 2012), so the availability of functional alternatives—the foundational model's key moderator—is increasingly determined by platform market structure rather than by audience preference. Zuboff (2019) pushes the critique further, arguing that platform business models are premised on cultivating and monetizing precisely the habitual, data-generating engagement that dependency theory treats as an antecedent of influence. The assumption unsettled here is that of a unitary media system: phantomization replaces the single audience–media–society triangle with a layered chain of dependencies—users on platforms, content producers on platforms, platforms on infrastructures—and it relocates control over the foundational model's key moderator, the availability of functional alternatives, from audience choice to market structure.

8.2 Algorithmic curation and recommender systems as objects of dependency

Where the Internet dependency models located dependency in perceived platform attributes, contemporary scholarship increasingly locates it in the algorithmic systems that select and rank content. Curation research demonstrates that exposure is now jointly produced by professional gatekeepers, social contacts, personal choices, and algorithmic filters (Thorson and Wells, 2016), with platform recommender systems operating as values-laden editorial actors in their own right (DeVito, 2017). The normative stakes of this shift are actively debated: recommender systems can diversify as well as narrow exposure, and their democratic role depends on the design goals they encode (Helberger, 2019; Nechushtai and Lewis, 2019).

Dependency-relevant evidence is accumulating rapidly. Gil de Zúñiga et al. (2017) identified the “news-finds-me” perception—the belief that one need not actively seek news because relevant information will arrive through algorithmically curated feeds—which functions, in dependency terms, as a wholesale delegation of the understanding goal to platform infrastructure. Schaetz et al. (2025) explicitly theorize algorithm dependency in platformized news use, showing across six European countries that comfort with algorithms and information-seeking orientations sustain reliance on algorithmic curation even among users with pronounced privacy concerns, and framing this reliance as a structural power asymmetry between users and platforms. Qiao et al. (2024) demonstrate that algorithmic app dependence is fed by information, entertainment, and expression gratifications, while Swart (2021) shows that young users' folk theories of algorithms shape how they understand, feel about, and engage with algorithmic selection—suggesting that algorithmic literacy operates as a moderator of dependency formation. Bucher (2018) provides the theoretical vocabulary for this condition: algorithmic power operates not by coercion but by structuring the conditions of visibility and attention under which goal pursuit occurs. Taken together, these studies indicate that a distinct algorithmic layer of dependency has emerged between the platform and the user: audiences depend not only on platforms as venues but on curation systems as the operative gatekeepers of what can be known, felt, and done. The assumption unsettled here is that dependency targets are inert repositories of information resources awaiting goal-directed selection: a recommender system is an adaptive gatekeeper that reallocates the very resources on which the user depends in response to the behavioral traces that dependent use itself generates, so that the target of dependency now reorganizes itself around the dependent user.

8.3 AI-mediated communication and dependency on generative agents

The diffusion of AI-mediated communication—interpersonal communication in which intelligent agents compose, augment, or generate messages on behalf of communicators (Hancock et al., 2020)—and of conversational generative AI extends dependency dynamics beyond content selection to content production. Sundar (2020) argues that the rise of machine agency requires effects theories to model the machine itself as a communicative source whose cues shape user trust and reliance. Dependency theory is well positioned to absorb this development: a generative AI assistant that answers questions (understanding), recommends courses of action (orientation), and provides companionship or entertainment (play) serves precisely the tripartite goal structure of the foundational model, but does so as a single, conversational, perpetually available agent—an unprecedented concentration of goal-serving functions in one dependency target.

Early measurement work substantiates the concern. Goh et al. (2025) validated a three-dimensional Generative AI Dependency Scale—cognitive preoccupation, negative consequences, and withdrawal—and found dependency associated with lower psychological need fulfillment, procrastination, impaired critical thinking, and loneliness. Omori et al. (2026), writing squarely within the media system dependency tradition, propose the Heightened Emotions Social Media Dependency Model, in which pervasive societal ambiguity elevates emotional states that intensify social media dependency in a reciprocal feedback loop, heightening vulnerability to misinformation and conspiracy content. These contributions demonstrate that the dependency tradition is already being extended to algorithmic and AI contexts—but in a piecemeal fashion, developing separate constructs (algorithm dependency, algorithmic app dependence, GenAI dependency, heightened-emotions dependency) without an integrating architecture. Consolidating these constructs is one task of the framework proposed in Section 10. The assumption unsettled here is the separation of medium and source on which the tradition's tripartite architecture rests: a generative agent is simultaneously channel, gatekeeper, and author, concentrating all three goal domains—understanding, orientation, and play—in a single conversational counterpart.

8.4 Platform governance, digital ecosystems, and the regulation of dependency

Finally, platform governance scholarship reframes dependency as a matter of public policy. Gorwa (2019) conceptualizes platform governance as the layered interplay of self-regulation, co-regulation, and state regulation through which platform conduct is shaped; from a dependency perspective, governance regimes determine the transparency, contestability, and design of the very systems on which audiences depend. Regulatory instruments that mandate algorithmic transparency, recommender-system choice, or interoperability effectively operate on the dependency relation itself—expanding functional alternatives, reducing lock-in, and redistributing the power asymmetries that Schaetz et al. (2025) and Fayez (2024) document. Conversely, design patterns that suppress alternatives (default settings, closed ecosystems, engagement-optimized feeds) intensify dependency by design. Dependency theory thus offers governance scholarship something it often lacks—a causal account of how structural reliance translates into effects on beliefs, feelings, and behavior—while governance scholarship offers dependency theory an account of the institutional forces that shape dependency's macro-level conditions (Helberger, 2019; Van Dijck et al., 2018). In the vocabulary of the integrated framework developed below, platform governance constitutes a macro-level moderator that conditions every downstream dependency relation in the digital ecosystem. The assumption unsettled here is that the macro level of the dependency relation consists only of diffuse social-structural conditions: governance regimes are deliberate, contestable interventions that can widen or narrow functional alternatives by design.

8.5 Synthesis: from environmental change to theoretical revision

The four developments reviewed above are not merely new empirical settings to which the existing dependency models can be applied unmodified; each revises an underlying assumption of the theory, and together they justify the algorithmic-agent layer introduced in Section 10. Phantomization (Section 8.1) dissolves the assumption of a unitary media system into a stratified chain of second-order dependencies and shifts the locus of functional alternatives from audience preference to market structure. Algorithmic curation (Section 8.2) dissolves the assumption of the passive dependency target: because curation systems adapt to the behavioral traces of use, the target of dependency reorganizes around the dependent user, converting the theory's one-directional resource-control logic into a recursive circuit. AI-mediated communication (Section 8.3) dissolves the assumed distinction between medium and source—and with it the assumption that understanding, orientation, and play goals are distributed across different media relations, since a single generative agent can serve all three simultaneously. Platform governance (Section 8.4) converts the macro level from an exogenous backdrop into a policy-manipulable moderator.

Two implications follow. First, these revisions are cumulative rather than parallel: phantomization concentrates the venues of dependency; algorithmic curation personalizes what those venues supply; AI generation authorizes the venue itself to produce what it supplies; and governance determines how contestable the entire arrangement is. Second, none of the three reviewed models—nor an incremental extension of any one of them—can accommodate these revisions, because all three assume dependency targets that neither adapt to individual users nor generate content in response to them (a point developed formally in Section 10.5). It is this assumption-level incompatibility, rather than the accumulation of new literature alone, that warrants a distinct algorithmic-agent stratum and the recursive feedback architecture of the integrated framework.

9 Critical evaluation: limitations, contradictions, and adjacent theories

9.1 Strengths and contributions

Media dependency theory has made enduring contributions by providing a framework that transcends simplistic direct-effects or limited-effects paradigms. Its primary strength is a relational and contextual approach that recognizes media effects as emerging from the interplay among audiences, media systems, and social structures (Ball-Rokeach and DeFleur, 1976). This systemic perspective has proven remarkably adaptable, extending from broadcast television to social networking platforms without abandoning its core explanatory logic.

The theory's goal-oriented foundation provides a parsimonious yet comprehensive account of why audiences depend on media. The tripartite framework of understanding, orientation, and play goals has demonstrated empirical utility across numerous studies and contexts, enabling prediction of which dependencies develop and which effects result (Ball-Rokeach and DeFleur, 1976; Cho, 2009; Işik et al., 2022). The multi-level analytical capacity of the framework constitutes another major strength: by accommodating micro-level psychological processes (habits, perceptions, individual differences), meso-level platform characteristics, and macro-level social conditions within a single theoretical architecture, the dependency perspective offers a comprehensive account of media effects that few competing frameworks match (Ball-Rokeach and DeFleur, 1976; Cho, 2009; Ognyanova and Ball-Rokeach, 2015). The theory's empirical productivity is equally noteworthy: its core constructs have proven measurable and predictive across varied operationalizations, designs, and populations (Aldamen, 2023; Kim and Jung, 2017; Schrock, 2006).

An additional strength worth highlighting is the theory's capacity to generate testable and falsifiable hypotheses. The specificity of its propositions—that dependency mediates media effects, that dependency intensity varies with social conditions and functional alternatives, that different goals produce different effect types—provides clear criteria against which empirical evidence can be assessed. This feature distinguishes the dependency framework from more descriptive or taxonomic approaches to media effects and contributes to its enduring scientific utility.

9.2 Limitations and critiques

Despite these strengths, the theory confronts several limitations. A primary critique concerns its macro-level emphasis in the foundational formulation, which some scholars argue underplays the local infrastructural and platform affordance dynamics that shape dependency in digital environments (Al Adwan et al., 2024). Whether media system dependency theory can adequately account for social media—or whether Communication Infrastructure Theory offers a better fit—remains an unresolved debate, raising concerns that the theory's origins in mass media systems may constrain its applicability to interactive, user-generated, and networked communication environments where traditional producer–audience distinctions are blurred. The force of this critique is amplified by the intellectual trajectory of the theory's own founder: Ball-Rokeach and colleagues' development of Communication Infrastructure Theory (Ball-Rokeach et al., 2001; Kim and Ball-Rokeach, 2006) can be read as an implicit acknowledgment that system-level dependency constructs required re-grounding in local storytelling networks and communication action contexts.

A second limitation concerns measurement and design. Most empirical studies rely on self-report surveys susceptible to common method bias, social desirability effects, and retrospective recall limitations. Cross-sectional designs predominate, limiting causal inference about the relationships among dependency, use patterns, and outcomes (Cho, 2009; Schrock, 2006). Longitudinal and experimental designs capable of establishing causal ordering and testing mediation pathways more rigorously remain comparatively rare. A related and sharper measurement critique concerns circularity: when dependency is operationalized through items asking respondents how helpful media are for achieving goals, and effects are operationalized through goal-related outcomes, high dependency–effect correlations risk being definitional rather than causal. The proliferation of non-equivalent scales—generic media dependency measures, SNS-specific scales (Kim and Jung, 2017), algorithmic app dependence measures (Al Adwan et al., 2026; Qiao et al., 2024), and generative AI dependency scales (Goh et al., 2025)—further impedes cumulative comparison across studies, since nominally identical constructs are measured with instruments of varying dimensionality and validation rigor.

A third limitation involves scope and boundary conditions: the conditions under which dependency relations are strongest, the thresholds at which dependency produces effects, and the moderators that amplify or attenuate those effects remain incompletely specified (Ball-Rokeach and DeFleur, 1976). A fourth critique concerns theoretical parsimony and distinctiveness. As the framework has incorporated habits, affordances, platform attributes, and individual differences, it has become increasingly complex (Cho, 2009), and the relationship between dependency theory and neighboring frameworks—uses and gratifications, technology acceptance models, Communication Infrastructure Theory—remains incompletely specified, raising questions about theoretical boundaries and distinctiveness. This concern has a long history: Rubin and Windahl's (1986) merger of uses and gratifications with dependency was motivated precisely by the difficulty of maintaining a sharp boundary between motivated media use and structural reliance, and critics have periodically asked whether dependency adds explanatory power beyond exposure, motivation, and habit measured separately. It is telling that authoritative overviews describe the theory's presence in current scholarship as diminished relative to its heyday, with many contemporary studies invoking dependency notions without engaging the full theoretical architecture (Abdelmaksoud et al., 2026; Jung, 2017).

A fifth limitation, which has received less attention in existing critiques, concerns the normative ambiguity of the dependency concept itself. The theory describes dependency as a structural relationship but does not inherently specify when dependency is adaptive and when it is maladaptive. In practice, some dependencies enhance individuals' capacity to navigate their social environments effectively, while others produce compulsive, harmful, or exploitative patterns. Without clearer normative criteria or thresholds distinguishing functional from dysfunctional dependency, the theory's practical implications for policy and intervention remain underdetermined. A sixth limitation follows from the platformization analysis in Section 8: the theory's canonical formulations treat the media system as a unitary actor with which audiences hold relations, whereas contemporary dependency is mediated through layered arrangements—creators dependent on platforms, platforms dependent on infrastructures, all governed by regulatory regimes (Gorwa, 2019; Van Dijck et al., 2018)—that the tripartite audience–media–society triangle compresses into a single node.

9.3 Contradictory findings and unresolved debates

A balanced assessment requires attention not only to conceptual limitations but to points where the empirical record is inconsistent. At least five substantive contradictions or open debates emerge from the reviewed literature.

First, the role of credibility is unsettled. Ognyanova and Ball-Rokeach (2015) found that perceived credibility did not predict Internet dependency, whereas comprehensiveness did; yet credibility-based accounts of media trust would predict the opposite, and dependency studies in environmental communication treat credibility as a working component of media influence (Liao, 2025). Whether credibility is irrelevant to dependency formation, relevant only in some domains, or relevant only above a threshold of perceived risk remains unresolved.

Second, the role of technological competence is contested. Schrock (2006) found that computer self-efficacy, access speed, and computer anxiety were unrelated to dependency, suggesting psychosocial rather than technological determinants; but algorithmic-era research finds that algorithm comfort and algorithmic literacy do shape reliance on curation systems (Schaetz et al., 2025; Swart, 2021), and that AI literacy conditions the consequences of AI dependence. The apparent contradiction may reflect a historical shift—as media grow more complex, competence variables may regain predictive force—but this conjecture awaits direct comparative testing.

Third, the participatory-use pathway is inconsistent. Cho (2009) found participatory use to be the weakest predictor of dependency, while Kim and Jung (2017) found that dependency strongly predicts interactive engagement—raising the possibility that participation is better modeled as an outcome than an antecedent of dependency, or that the causal arrow runs in both directions across time. Cross-sectional designs cannot adjudicate this question.

Fourth, the valence of dependency subtypes diverges. Wangqu et al. (2024) report that the relationship between dependency and depression depends on the nature of the dependency involved, and Qiao et al. (2024) show that the same gratification (entertainment) simultaneously feeds dependence and the fatigue that suppresses it. Such findings unsettle any monotonic assumption that more dependency yields stronger effects—the dose–response form of the dependency–effect relation is essentially unmapped.

Fifth, the succession debate persists: whether media system dependency should be retained and extended, merged with uses and gratifications (Rubin and Windahl, 1986), or superseded by Communication Infrastructure Theory for networked environments (Ball-Rokeach et al., 2001) has never been formally resolved. The integrated framework proposed in Section 10 takes a position in this debate—retention with structural extension—while making the choice explicit and falsifiable rather than tacit.

9.4 Integration with complementary and competing theories

Scholars have increasingly pursued integration with complementary frameworks to address these limitations. The most productive integration involves uses and gratifications (U&G) theory, which shares dependency theory's emphasis on audience goals and active media selection but focuses more on individual motivations and satisfactions than on systemic dependency relations. Aldamen (2023) explicitly integrated media system dependency and U&G theories in a study of Syrian refugees, arguing that U&G explains what gratifications audiences seek while dependency theory explains why audiences rely on particular media to obtain those gratifications and what effects result. This integration preserves the strengths of both frameworks while addressing limitations inherent in each. The integration has deep roots: Rubin and Windahl (1986) formalized the uses and dependency model precisely to combine U&G's motivational account with dependency's structural account, and that merged model is a direct ancestor of the framework proposed in Section 10.

A second integration involves Communication Infrastructure Theory (CIT), which examines how communication resources—media, interpersonal networks, and community organizations—are structured and accessed within geographic communities (Ball-Rokeach et al., 2001; Fayez et al., 2026; Kim and Ball-Rokeach, 2006). CIT extends dependency theory's systemic perspective by emphasizing local communication ecologies and infrastructure inequalities, potentially offering a more suitable analytical lens for networked, distributed communication structures that characterize contemporary media environments. CIT's signature constructs—the neighborhood storytelling network and the communication action context—replace the abstract media system with concrete, mappable configurations of storytellers and resources (Kim and Ball-Rokeach, 2006). For the integrated framework developed below, CIT contributes the insight that dependency is always embedded in a locally specific ecology of alternatives, not an undifferentiated national media system.

A third integration involves technology acceptance models. Carillo et al. (2014) combined media system dependency theory with expectation-confirmation theories to predict continuance intention to use ubiquitous media systems, finding that dependency inflates perceived usefulness and ease of use and affects cognitive appraisal of discrepancies between expectations and performance. This integration demonstrates that dependency theory can enhance technology acceptance models by introducing relational and contextual factors that extend beyond individual perceptions. Kheiravar (2018) similarly integrated UTAUT2 with media system dependency theory to explore ubiquitous computing acceptance in retail environments.

Two further perspectives, not previously connected to the dependency tradition in this literature, merit explicit engagement. Media Repertoire Theory analyzes the patterned combinations of media that individuals assemble and maintain (Hasebrink and Domeyer, 2012; Hasebrink and Popp, 2006), with empirical work identifying user-defined cross-platform repertoires as the natural unit of media behavior (Taneja et al., 2012). The repertoire perspective reframes the functional-alternatives construct: alternatives are not merely available or unavailable in the environment but are actively curated into personal portfolios, which implies that dependency should be measured at the repertoire level as well as the platform level—precisely the polymedia logic that Wangqu et al. (2024) and Madianou and Miller (2013) document. Affordance Theory, as explicated for communication research by Evans et al. (2017), treats affordances as relational properties standing between a technology's features and the outcomes of its use. For dependency theory, the affordance lens disciplines a recurring conceptual slippage in the Internet-era models, which sometimes treated affordances as objective platform attributes and sometimes as user perceptions: on the relational view, dependency-relevant affordances (persistence, availability, personalization, anonymity) are actualized differently by differently situated users, which explains why identical platforms generate heterogeneous dependency relations across populations (Schrock, 2006; Wangqu et al., 2024).

To make the nature of these integrations explicit, the seven perspectives reviewed here stand in three distinct relationships to media dependency theory, and each contributes a specified element to the integrated framework developed in Section 10. Uses and gratifications, habit theory, and social cognitive theory are complements they operate at the same explanatory level as dependency theory, supply mechanisms that the theory names but does not itself model (motivation, automaticity, self-efficacy), and enter the framework as its micro-level goal and use-pattern constructs and individual moderators (P1, P2). Media Repertoire Theory, Affordance Theory, and the technology acceptance models are extensions—more precisely, correctives that extend the theory's construct and measurement logic: repertoire research relocates the functional-alternatives construct from the environment into the person, informing the distribution dimension of the framework's central construct (P5); affordance theory disciplines the platform-attribute constructs of the Internet-era models into relational form, informing the meso stratum (P3); and technology acceptance research supplies the perceptual feedback logic that becomes Loop A (P4). Communication Infrastructure Theory, by contrast, is the one genuine challenger: developed by the theory's own founder as a potential successor, it disputes the adequacy of system-level dependency constructs for networked environments rather than supplementing them. The integrated framework answers this challenge not by absorbing CIT wholesale but by importing its ecological insight—that alternatives are locally configured—while retaining the effects-oriented causal spine that CIT brackets, and by converting the succession dispute into a falsifiable empirical comparison (Sections 10.4 and 12.5).

9.5 Critical synthesis: what the critiques jointly require of an integrated framework

The critical evaluation presented above confirms that media dependency theory remains a foundational framework in communication research, yet its future utility depends on ongoing theoretical innovation and empirical refinement. Integrative approaches offer promising pathways for addressing identified limitations: by combining dependency theory with uses and gratifications, Communication Infrastructure Theory, technology acceptance models, habit theory, media repertoire theory, affordance theory, and social cognitive theory, researchers can construct more nuanced models accounting for both individual-level motivations and systemic influences. Such models enable the identification of mediators and moderators that clarify the conditions under which dependency relations emerge and produce effects.

Read together rather than seriatim, the critiques of Section 9.2 and the contradictions of Section 9.3 do more than qualify the theory: they specify design requirements that any integrated successor framework must satisfy, and they discipline the claims such a framework can legitimately make. Five requirements emerge. The unitary-system critique and the phantomization analysis (Sections 8.1 and 9.2) require that dependency targets be stratified rather than compressed into a single “media” node. The measurement critique—single-platform operationalization in a polymedia environment—requires that dependency be conceptualized as distributed across a repertoire rather than as a scalar attaching to one target. The circularity critique requires that antecedents (goals, habits), the dependency relation, and outcomes be specified as separately measurable constructs, so that their associations are empirical rather than definitional. The normative-ambiguity critique requires that the valence of dependency outcomes be theorized as conditional rather than assumed. And the boundary-condition critique, together with the contradictions concerning credibility, competence, and participation, requires that contested relationships be stated as falsifiable propositions rather than smoothed into narrative consensus.

Section 10 presents a framework constructed against precisely these requirements, and Table 9 summarizes the resulting ledger: which limitations the framework resolves, through which architectural element, and which it can only restate more precisely. The ledger's third column is as important as its second, because the claim advanced in this review is calibrated rather than triumphal. Three tensions are not resolved by the IMDF or by any conceptual framework. The circularity problem is mitigated by construct separation but eliminated only at the level of operationalization—by measurement decisions in primary studies, not by theory. The parsimony objection is aggravated rather than resolved: a four-strata architecture with seven propositions and two feedback loops is less parsimonious than any of its predecessors, a cost accepted here in exchange for specificity and falsifiability, but a cost nonetheless. And the succession debate with Communication Infrastructure Theory is not settled but converted into an empirical wager (Section 12.5) that could yet be lost: should stratified dependency add no predictive variance beyond ecology-level constructs, the correct conclusion would be supersession, not integration. Making these residues explicit is intended to position the framework as an advance on the literature that remains accountable to it.

A key implication is that dependency should not be viewed as a static or monolithic construct, but as a dynamic, context-sensitive relation shaped by evolving media environments, technological affordances, and social structures. Future theoretical development should prioritize multidimensional measurement strategies—incorporating behavioral, psychophysiological, and longitudinal data—to capture the complexity of contemporary media dependencies. Explicit consideration of cultural, infrastructural, and algorithmic factors is essential for extending the theory's relevance across diverse contexts. Balancing explanatory breadth with analytical precision remains the central challenge: the theory's enduring value lies in its capacity to adapt to new communication realities, but this adaptability must be matched by methodological rigor and conceptual specificity. The next section takes up this challenge directly.

10 The integrated media dependency framework (IMDF)

10.1 Rationale and definition

The review to this point yields three conclusions that jointly motivate an integrated framework. First, the three focal models share a durable causal core—goal-directed reliance on information resources conditions media effects—that has survived fifty years of environmental change (Section 6). Second, each model contributes a component the others lack: the foundational model contributes the systemic, multi-level architecture and goal typology; the Internet adaptations contribute platform-level operationalization and affordance sensitivity; the New Media model contributes habit as an antecedent and explicit mediation logic. Third, none of the three—nor any published successor—accounts for the algorithmic curation systems and generative AI agents that now stand between platforms and users (Section 8), for repertoire-level distribution of dependency (Section 9.4), or for the governance regimes that condition all of the above. The Integrated Media Dependency Framework (IMDF) is proposed to close these gaps.

The IMDF defines media dependency as a goal-anchored, habit-reinforced, and increasingly algorithmically intermediated relation of reliance between social actors and the information resources of a stratified media ecosystem, whose intensity and distribution across that ecosystem jointly condition cognitive, affective, and behavioral effects. Three commitments distinguish this definition from its predecessors: dependency is stratified (it attaches simultaneously to systems, platforms, algorithmic curation processes, and AI agents), distributed (it is held across a personal repertoire rather than concentrated in a single target), and recursive (its outcomes feed back into the habits, perceptions, and data traces that regenerate it).

10.2 Architecture of the framework and description of Figure 1

Figure 1 presents the framework. Its architecture is described here in sufficient detail to be drawn independently. The figure is organized as four horizontal strata crossed by a left-to-right causal flow and two feedback loops.

The top stratum (Layer 1: Macro—Societal Structure and Governance) is a horizontal band containing three boxes: “Social structural conditions (stability, crisis, ambiguity, censorship, infrastructure)”, “Media system concentration and platformization”, and “Platform governance and regulation”. Downward arrows from this band indicate that macro conditions shape everything below (labeled P6 at the arrow from social conditions).

The second stratum (Layer 2: Meso—Platform and Algorithmic-Agent Layer) contains two adjacent boxes. The left box, “Platform affordances (availability, interactivity, comprehensiveness, personalization)”, represents the platform level inherited from the Internet adaptations. The right box, “Algorithmic curation and AI agents (recommender systems, feeds, generative assistants)”, represents the newly theorized algorithmic-agent layer; it is drawn nested half-inside the platform box to indicate that curation systems operate within platforms yet increasingly act as distinct dependency targets (labeled P3).

The third stratum (Layer 3: Micro—Individual) contains three boxes arranged left to right: “Goals (understanding, orientation, play)” (labeled P1); “Use patterns (ritualized, instrumental, participatory habit)” (labeled P2); and “Individual moderators (media and algorithmic literacy, personality, emotional states, functional alternatives in repertoire)”. Arrows run from the goals box and the use-patterns box into the central construct.

At the center-right of the figure sits the central construct box, “Dependency relations (intensity × distribution across repertoire)”, drawn with a double border to mark its mediating status (labeled P5 for the distribution dimension). Arrows converge on it from Layer 2 (affordances and algorithmic curation), and from Layer 3 (goals and habits). From the central construct, a single thick arrow runs right to the outcomes box, “Effects: cognitive, affective, behavioral—adaptive or maladaptive valence” (labeled P7), with the valence moderated by an arrow descending from the individual-moderators box.

Two feedback loops complete the figure. Loop A (labeled P4, drawn as a dashed arrow curving from the dependency box back to the platform-affordances box) represents the perceptual feedback documented by Carillo et al. (2014): dependency inflates perceived platform attributes, which further intensify dependency. Loop B (drawn as a dashed arrow curving from the effects box back to the use-patterns box and onward to the algorithmic-curation box) represents the behavioral-algorithmic feedback: effects reshape habitual use, and the data traces of use retrain the curation systems, which restructure exposure—closing a recursive circuit that has no equivalent in the broadcast-era model.

10.3 Propositions

The framework's relationships are formalized in seven propositions, each grounded in the reviewed evidence and stated so as to be testable and falsifiable.

10.3.1 Proposition 1 (goal centrality)

The intensity of dependency on any target—medium, platform, algorithm, or AI agent—increases with the centrality of the understanding, orientation, and play goals that target serves and with the exclusivity with which it serves them relative to functional alternatives (Ball-Rokeach and DeFleur, 1976; Işik et al., 2022).

10.3.2 Proposition 2 (habit reinforcement)

Ritualized and instrumental use patterns intensify dependency over time, and dependency partially mediates their effects on outcomes; participatory use operates primarily as a consequence rather than an antecedent of dependency until social investment accumulates (Cho, 2009; Kim and Jung, 2017).

10.3.3 Proposition 3 (algorithmic intermediation)

Algorithmic curation and generative AI agents constitute a distinct stratum of dependency targets: reliance on them increases with perceived comprehensiveness and personalization and reduces the salience of functional alternatives, producing lock-in effects stronger than platform-level dependency alone would predict (Gil de Zúñiga et al., 2017; Goh et al., 2025; Schaetz et al., 2025).

10.3.4 Proposition 4 (perceptual feedback)

Dependency biases evaluations of the depended-upon target—inflating perceived usefulness, ease of use, and comprehensiveness—thereby reinforcing the dependency relation in a self-amplifying loop (Carillo et al., 2014; Qiao et al., 2024).

10.3.5 Proposition 5 (repertoire distribution)

Dependency is distributed across a personal media repertoire; aggregate dependency and its effects cannot be inferred from any single platform relation, and platform-swinging within the repertoire moderates the relation between dependency intensity and well-being outcomes (Hasebrink and Popp, 2006; Madianou and Miller, 2013; Wangqu et al., 2024).

10.3.6 Proposition 6 (contextual amplification)

Macro-level ambiguity, threat, and crisis intensify dependency across all strata, with heightened emotional states mediating the amplification; governance regimes moderate the availability and transparency of the targets on which amplified dependency falls (Ball-Rokeach and DeFleur, 1976; Loges, 1994; Lowrey, 2004; Omori et al., 2026).

10.3.7 Proposition 7 (conditional valence)

The valence of dependency outcomes—adaptive or maladaptive—is conditional on goal type, content environment quality, and individual capacities (media, algorithmic, and AI literacy); identical dependency intensities can therefore yield opposite well-being outcomes (Aldamen, 2023; Djamaluddin et al., 2023; Goh et al., 2025; Wangqu et al., 2024).

10.4 Positioning relative to adjacent frameworks

The IMDF is positioned deliberately with respect to the adjacent perspectives reviewed in Section 9.4. Consistent with the classification developed there, the framework treats uses and gratifications as a complement whose motivational entry point it absorbs; Media Repertoire Theory, Affordance Theory, and technology acceptance research as extensions whose constructs it incorporates in revised form; and Communication Infrastructure Theory as a challenge whose central objection it answers by specification rather than assimilation. Relative to uses and gratifications, the IMDF retains motivational goals as the entry point (P1) but insists, with the dependency tradition, that structural reliance—not motivation alone—converts use into effects. Relative to Communication Infrastructure Theory, the IMDF adopts CIT's insight that dependency is embedded in concrete local ecologies, operationalized here as the repertoire and its functional alternatives (P5), while retaining the effects-oriented causal spine that CIT largely brackets. Relative to Media Repertoire Theory, the IMDF supplies what repertoire research lacks—a causal account of why repertoires stabilize (habit, P2) and what follows from them (effects, P7). Relative to Affordance Theory, the IMDF treats dependency-relevant affordances as relational and stratified (P3), following Evans et al. (2017). And relative to technology acceptance models, the IMDF absorbs their continuance logic as a special case of the perceptual feedback loop (P4). The framework is thus not a competitor to these perspectives but an integrative architecture in which each does specified work—while remaining falsifiable: evidence that algorithmic reliance adds no predictive variance beyond platform dependency would refute P3, and evidence that repertoire distribution does not moderate outcomes would refute P5.

10.5 What is genuinely new: specifying the framework's theoretical contribution

Because an integrative framework risks being read as a repackaging of parts already available, it is worth stating precisely where the IMDF's novelty lies and where it deliberately conserves. Three elements are inherited: the tripartite goal typology and the macro-level conditioning of dependency come from the foundational model; the platform-level affordance constructs come from the Internet adaptations; habit-as-antecedent and the explicit mediation logic come from the New Media model. These are conserved because the review found them empirically durable (Section 6), and the framework claims no originality for them. Four elements, however, are to the authors' knowledge novel to the dependency tradition, and none is obtainable by combining the existing models.

First, the algorithmic-agent stratum (P3) introduces a class of dependency target with a property that no prior model attributes to any target: adaptivity. In every earlier formulation—foundational, Internet, and New Media like the depended-upon entity is a fixed repository whose information resources exist independently of any particular user, and the user's goals and habits do all of the causal work. A recommender system or generative agent, by contrast, reconstitutes its resource offer user by user and moment by moment, in response to the behavioral traces that dependent use itself produces. This is not a stronger or more personalized version of platform dependency; it is a categorically different relation, in which the target participates in producing the dependency directed at it.

Second, and consequently, the recursive feedback architecture is novel in kind rather than in degree. The dependency tradition has acknowledged reciprocity since Ball-Rokeach and DeFleur (1976) noted that effects feedback on society and media systems, and Carillo et al. (2014) documented perceptual feedback at the individual level—conserved here as Loop A. In both cases, however, the feedback runs through human perception or social aggregation. Loop B specifies a circuit that runs through the target's own optimization machinery: use generates data, data retrain curation, curation restructures exposure, and restructured exposure reshapes the habits that generate use. No existing dependency model contains a mechanism by which the dependency relation trains its own object—nor do the adjacent frameworks review in Section 9.4. Uses and gratifications and repertoire theory model user selection among static options; affordance theory models the relational actualization of stable features; Communication Infrastructure Theory models locally configured but non-adaptive storytelling resources. Explaining phenomena such as algorithmic lock-in, the news-finds-me delegation, and engagement-optimized habit formation therefore requires theoretical machinery that none of these perspectives' supplies.

Third, the reconceptualization of dependency as intensity × distribution (P5) replaces the scalar dependency construct common to all three predecessors with a two-dimensional one. Prior models ask how much a person depends on a medium or platform; the IMDF additionally asks how that dependency is spread across the personal repertoire and holds that the same aggregate intensity has different consequences when concentrated in one target than when distributed across many. This dimension cannot be recovered from any single-target model, however refined its measurement.

Fourth, the framework relocates governance from context to construct: platform governance enters the model as a specifiable macro-level moderator operating on the availability and transparency of functional alternatives (P6), rather than as a backdrop against which dependency plays out. This converts a hitherto normative literature (Gorwa, 2019; Helberger, 2019) into a source of testable moderation hypotheses—to the authors' knowledge, a first for the dependency tradition.

The architectural difference from prior dependency models can accordingly be stated compactly. The foundational model is a triangle—audience, media, society—with unidirectional effect flows; the Internet adaptations substitute platforms for the media node; the New Media model prepends a habit stage to an otherwise linear chain. The IMDF is neither a triangle nor a chain but a stratified recursion: four unequally powerful strata of targets, a two-dimensional central construct, and two feedback loops of different kinds. What makes this difference theoretical rather than merely diagrammatic is the accompanying claim, stated falsifiably in P3 and P5, that the added structure earns its complexity—that algorithmic-stratum reliance and repertoire distribution each explain variance in dependency outcomes that no combination of the inherited constructs can absorb. If that claim fails empirically, the correct verdict will be that the IMDF over-complicates, and the tradition should retrench to its simpler predecessors; Section 9.5 records this as an accepted risk.

10.6 Operational summary

Table 8 summarizes the framework in operational terms, extending the schematic overview developed in earlier drafts of this review with the algorithmic-agent stratum and the two feedback loops.

The IMDF is offered as a working architecture rather than a finished edifice: its full operationalization and empirical testing will require sustained collaborative effort across research teams, methodological traditions, and cultural contexts. Nevertheless, it converts the aspirational integration sketched in earlier dependency scholarship into a specified, falsifiable framework that honors the theoretical achievements of the past five decades while extending them to the algorithmic and AI-mediated present.

11 Future research directions

The theoretical trajectory of media dependency theory points toward several productive avenues for future inquiry, collectively aimed at strengthening causal inference, broadening contextual scope, deepening mechanistic understanding, and enhancing practical relevance. Each direction below is anchored in the Integrated Media Dependency Framework: the propositions of Section 10.3 specify what needs testing, and Table 1 pairs each direction with the emerging research trend it addresses and illustrative research questions.

Longitudinal and experimental designs are urgently needed to establish causal ordering and test mediation pathways with greater rigor. While cross-sectional surveys have documented associations between dependency, use patterns, and outcomes, longitudinal panel studies could track how dependencies develop, how they respond to changes in media environments or social conditions, and how they predict subsequent outcomes over time (Cho, 2009; Kim and Jung, 2017). Experimental manipulations of platform affordances, functional alternatives, or social contexts could test causal hypotheses about the determinants of dependency intensity and the conditions under which dependency produces effects. Diary studies, experience sampling methods, and ecological momentary assessment approaches could complement these designs by capturing dependency dynamics as they unfold in real time within participants’ natural media environments. Such designs are the only route to adjudicating the directionality questions raised in Section 9.3, including whether participatory use is antecedent or consequence of dependency (P2) and the causal structure of the perceptual feedback loop (P4).

Multi-platform and polymedia research should examine how audiences navigate complex media ecologies in which multiple platforms serve overlapping and complementary functions. Wangqu et al. (2024) demonstrated the value of studying platform-swinging, but more research is needed on how dependencies are distributed across platforms, how affordances interact to shape overall dependency patterns, and how audiences strategically allocate attention and reliance—addressing both within-person dynamics and between-person differences in dependency portfolios. Research in this area could benefit from network-analytic approaches that map relationships among platforms in individuals' media repertoires, identifying clusters of complementary and substitutable platforms and examining how dependency on one platform affects dependency on others (P5; Hasebrink and Popp, 2006; Taneja et al., 2012).

Mechanism identification should move the field beyond documenting associations toward explaining how and why dependency produces effects. Further work is needed to identify the psychological processes—attention, elaboration, identification, social comparison—that mediate the relationship between dependency and cognitive, affective, and behavioral outcomes (Cho, 2009). Research should also examine the role of emotional and motivational processes in sustaining dependency, including how anticipated gratifications, fear of missing out, and social pressure contribute to dependency maintenance and intensification —a direction now formalized in the heightened-emotions extension of the dependency tradition (Omori et al., 2026, p. 6). Cross-cultural and comparative research should test the generalizability of dependency theory across diverse cultural, political, and economic contexts. While some studies have examined dependency in specific national settings (Aldamen, 2023; Zhang and Zhong, 2020), systematic cross-cultural comparisons identifying cultural moderators, testing the universality of the understanding–orientation–play goal typology, and examining how media system structures shape dependency differently across societies remain rare.

Research on algorithmic and AI dependency constitutes the most urgent frontier. The constructs emerging in this space—algorithm dependency (Schaetz et al., 2025), algorithmic app dependence (Qiao et al., 2024), and generative AI dependency (Goh et al., 2025)—require convergent-validity studies establishing whether they are facets of a single algorithmic-agent stratum, as Proposition 3 holds, or distinct phenomena. Researchers should test whether reliance on generative AI for understanding, orientation, and play goals follows the dependency logic of P1, whether AI literacy moderates outcome valence as P7 predicts, and how the behavioral–algorithmic feedback loop (Loop B in Figure 1) shapes long-run exposure and belief formation.

Integration with neuroscience and psychophysiology could provide objective measures of dependency supplementing self-reports: neuroimaging, eye-tracking, psychophysiological measures, and behavioral trace data could enable more precise measurement of dependency intensity, habitual use, and downstream effects. Normative and ethical research should address the implications of dependency for individual well-being, democratic participation, and social equity, including interventions to reduce problematic dependencies, promote media literacy and critical consumption, and inform platform designs that fulfill dependency needs without fostering excessive reliance or harmful effects (Djamaluddin et al., 2023; Wangqu et al., 2024). Research on algorithmic and platform governance should examine how design choices—notifications, infinite scroll, personalized recommendations—affect dependency intensity, whether they exploit psychological vulnerabilities, and how regulatory frameworks might promote healthier dependency relations (Gorwa, 2019; Helberger, 2019).

Finally, a mixed methods approach combining quantitative surveys, qualitative interviews, and digital trace analytics can offer a more comprehensive portrait of media dependency, revealing both subjective experiences and objective patterns of use. Advanced statistical modeling, network analysis, and machine learning techniques can help disentangle complex dependency relations, identify latent subgroups, and forecast long-term trends in audience–media interactions.

12 Discussion and implications for theory and practice

12.1 Overarching patterns and theoretical implications

The integrative review undertaken in the preceding sections reveals several overarching patterns that merit sustained discussion. Perhaps the most fundamental finding is the remarkable theoretical resilience of the core dependency logic across five decades of media transformation. Despite the shift from a media landscape dominated by a small number of broadcast outlets to one characterized by billions of user-generated posts, algorithmically curated feeds, and perpetually available mobile platforms, the central proposition—that the intensity of audience dependency on media information resources determines the magnitude and nature of media effects—has not been falsified. If anything, the empirical record suggests that this core logic has gained explanatory power as media have become more pervasive, since the conditions theorized to intensify dependency (ubiquity, perceived irreplaceability, integration into daily routines) are more fully realized in contemporary digital environments than in the broadcast systems for which the model was originally designed.

At the same time, the review demonstrates that preserving the core logic has required substantial operational and theoretical augmentation. The foundational model's emphasis on system-level relations and macro-social conditions, while still relevant, is insufficient to capture the platform-specific, habit-driven, and individually variable dependency patterns that characterize contemporary media use. The progressive refinement from system-level to platform-level to habit-mediated models represents a trajectory of increasing specificity and testability that has strengthened the theory's empirical utility. This trajectory, however, introduces a tension between generality and specificity. The foundational model's broad systemic claims apply in principle to any media environment, but the platform-specific and habit-centered models generate predictions tied to technologies and use patterns that may have limited shelf life as platforms rise, evolve, and decline. Future theoretical work must navigate this tension, preserving the general principles that give the theory its explanatory breadth while operationalizing constructs with sufficient precision to yield testable and practically relevant predictions. The Integrated Media Dependency Framework addresses this tension by locating enduring general principles (P1, P6, P7) and environment-specific mechanisms (P3, P4, P5) at distinct strata of a single architecture.

12.2 The dual valence of dependency

A recurrent theme across the empirical studies reviewed is the dual valence of media dependency—its capacity to produce both beneficial and harmful outcomes. On the positive side, dependency on informational media has been associated with enhanced political efficacy (Ognyanova and Ball-Rokeach, 2015), climate change literacy (Youssef, 2023), successful refugee adaptation (Aldamen, 2023), and prosocial behavior (Cho, 2009). On the negative side, high dependency has been linked to psychological depression (Wangqu et al., 2024), vulnerability to disinformation (Djamaluddin et al., 2023), displacement of other media and activities (Işik et al., 2022), and potentially harmful relational dynamics such as social isolation among youth (Khalifa et al., 2026).

This dual valence has important implications for both theory and practice. Theoretically, it suggests that dependency is not inherently pathological—a distinction that must be maintained against popular rhetoric equating media dependency with addiction. Dependency, in the theoretical sense, is a structural relationship defined by the centrality of media to goal attainment; its valence depends on the content environment, the goals being pursued, the availability of complementary resources, and the individual's capacity for critical engagement. Practically, this dual valence means that interventions should not aim to eliminate dependency per se—which would be neither feasible nor desirable given the centrality of media to contemporary social life—but rather to promote conditions under which dependency serves adaptive functions: diverse and reliable information ecosystems, strong media literacy, transparent algorithmic governance, and accessible functional alternatives.

12.3 Implications for understanding digital media ecosystems

The review findings have substantial implications for understanding how contemporary digital media ecosystems shape audience cognition, affect, and behavior. The evidence that perceived comprehensiveness predicts dependency more strongly than perceived credibility (Ognyanova and Ball-Rokeach, 2015) is consequential for the current era of information abundance and contested truth claims. If audiences develop dependencies based on the scope of available information rather than its trustworthiness, then platforms that aggregate vast quantities of content—regardless of quality—may foster stronger dependency relations than those that prioritize curated, verified information. This dynamic could contribute to the persistence of misinformation ecosystems, as platforms optimized for comprehensiveness may inadvertently include low-quality or misleading content within the information landscape upon which audiences depend.

The finding that dependency inflates perceived usefulness and ease of use (Carillo et al., 2014) has similar implications for platform competition and user autonomy. If dependent users systematically overestimate the quality of their preferred platforms, competitive alternatives may face higher adoption barriers than objective performance differences would warrant. This perceptual distortion may help explain platform lock-in effects and the persistence of dominant platforms even when superior alternatives are available. From a policy perspective, these findings suggest that promoting informed media choice requires not only information provision but also interventions that address the perceptual biases associated with high dependency.

The role of algorithmic curation in shaping dependency deserves particular attention. Contemporary platforms deploy recommendation algorithms that personalize content feeds based on inferred user preferences, behavioral signals, and engagement metrics. These algorithms may intensify dependency by increasing perceived comprehensiveness—users encounter a broader and more personalized range of content—while simultaneously reducing the salience of functional alternatives by keeping users within a single platform ecosystem. The interaction between algorithmic design and dependency formation is a critical area for future research, as it sits at the intersection of communication theory, computational social science, and technology policy (see Sections 8.2 and 10.3, Proposition 3).

12.4 Habit, automaticity, and agency

Cho's (2009) integration of habit as a precursor to dependency raises important questions about the relationship between automaticity and agency in media dependency. The foundational model implicitly portrays dependency as arising from conscious or semi-conscious goal pursuit: audiences depend on media because they recognize, at some level, that media serve important goals. Habitual use, by contrast, involves automatic behavioral patterns performed with minimal conscious deliberation. The incorporation of habit into the dependency framework thus introduces a mechanism through which dependency can develop and intensify without full conscious awareness or intentional choice.

This raises questions about the degree of agency involved in media dependency. If habitual use—shaped by contextual cues, platform design, and reinforcement schedules—drives dependency formation, then audiences may become highly dependent on media through processes that are only partially voluntary. This has implications for ethical evaluation: media systems and platforms that deliberately design features to foster habitual use (through notifications, variable reward schedules, and friction reduction) may bear responsibility for the dependency relations that result. The intersection of habit formation, platform design, and dependency is a particularly promising area for interdisciplinary research combining communication theory, behavioral science, and design ethics.

12.5 Theoretical implications of the integrated framework

Beyond the patterns discussed above, the Integrated Media Dependency Framework carries five implications for communication theory, each following from a specific element of the framework rather than appended to it.

First, for media effects scholarship, the framework's stratification of dependency targets (P3) implies that effects studies measuring only platform-level dependency are mis-specified in algorithmic environments: variance attributed to “platform dependency” may in fact reflect reliance on curation systems or AI agents operating within and across platforms. Researchers designing dependency measures should therefore assess the system, platform, algorithmic, and AI-agent strata separately, enabling decomposition of their distinct contributions to outcomes. This serves scholars seeking to reconcile inconsistent effect sizes across the dependency literature.

Second, for theory integration, the framework demonstrates that dependency theory can absorb—rather than compete with—adjacent perspectives: motivational goals from uses and gratifications (P1), habit mechanisms from behavioral science (P2), repertoire structure from media repertoire theory (P5), relational affordances from affordance theory (P3), and continuance logic from technology acceptance research (P4). The theoretical implication is that dependency functions as an integrative middle-range mechanism connecting audience motivation to media power, which serves theorists working toward cumulative rather than fragmented media theory.

Third, for the succession debate with Communication Infrastructure Theory, the framework converts an abstract rivalry into an empirical question: if repertoire- and ecology-level constructs (P5) fully absorb the predictive power of system-level dependency, CIT's localism is vindicated; if stratified dependency retains independent predictive force, retention-with-extension is supported. This serves scholars on both sides of a debate that has remained unresolved for two decades (Ball-Rokeach et al., 2001; Kim and Ball-Rokeach, 2006).

Fourth, for the study of misinformation and epistemic vulnerability, the framework identifies a specific causal chain—contextual ambiguity amplifies emotionally charged dependency (P6), which falls increasingly on comprehensiveness-optimized algorithmic targets (P3), whose valence for belief accuracy is conditional on literacy (P7)—that connects dependency theory to one of the field's most urgent research programs (Djamaluddin et al., 2023; Omori et al., 2026). This serves political communication and misinformation researchers seeking structural rather than purely psychological explanations.

Fifth, for methodology, the recursive loops of the framework (P4; Loop B in Figure 1) imply that cross-sectional designs cannot in principle identify the causal structure of contemporary dependency, since user behavior and algorithmic curation co-produce each other over time. The framework thereby supplies theoretical justification for the field's investment in longitudinal, experimental, and trace-based designs (Section 11), serving methodologists and funders who must prioritize among designs.

12.6 Practical and managerial implications

The framework likewise generates practical implications for five identifiable stakeholder groups, each derived from specified propositions.

For platform designers and product managers, the perceptual feedback loop (P4) and algorithmic intermediation (P3) imply that engagement metrics systematically overstate user welfare: dependent users rate platforms more favorably than objective performance warrants, and lock-in suppresses the alternatives against which quality would be judged (Carillo et al., 2014). Responsible product practice should therefore include dependency audits—assessing whether design features (notifications, infinite scroll, personalization defaults) serve users' understanding, orientation, and play goals or merely manufacture habitual reliance—and should preserve user-controllable alternatives such as chronological feeds and export options. This translates the humane-design agenda into measurable, theory-based criteria.

For policymakers and regulators, the framework identifies functional alternatives as the master lever of dependency governance (P1, P3, P6). Transparency obligations, recommender-system choice, interoperability, and data portability all operate by expanding alternatives and reducing algorithmic lock-in (Gorwa, 2019; Helberger, 2019). The framework also implies a crisis-governance duty: because dependency predictably surges under ambiguity and threat (P6; Loges, 1994; Lowrey, 2004), the reliability of the channels on which crisis-time dependency falls is a matter of public infrastructure, not merely market outcome.

For media organizations and communication managers, the comprehensiveness-over-credibility finding (Ognyanova and Ball-Rokeach, 2015) is both warning and opportunity: audiences bond to sources that feel comprehensive, so credibility investments alone will not secure audience dependency—but comprehensiveness without credibility feeds the misinformation dynamics described above. Editorial strategy should therefore pair breadth with verifiable quality, and audience development should cultivate habit-compatible formats (newsletters, briefings) that build ritualized use (P2) toward trustworthy targets. For brand and campaign managers, the evidence that dependency channels shape environmental literacy and behavior (Al Adwan et al., 2025; Liao, 2025; Youssef, 2023) indicates that prosocial campaigns are most effective when routed through the platforms and creators on which target audiences are already dependent.

For educators and media literacy professionals, the conditional-valence proposition (P7) implies that the goal of intervention is not reduced use but improved dependency quality: curricula should build algorithmic literacy (understanding how curation shapes exposure; Swart, 2021) and AI literacy (recognizing the limits of generative agents; Goh et al., 2025), which the framework identifies as the moderators that flip dependency from maladaptive to adaptive.

For mental health practitioners and student-support services, the divergent outcomes of dependency subtypes (Wangqu et al., 2024) and the validated generative AI dependency dimensions (Goh et al., 2025) imply that assessment should distinguish which stratum and which goals a client's dependency serves; interventions targeting habit architecture—cues, triggers, reinforcement schedules (Section 5.2)—are indicated where ritualized use drives the dependency, whereas interventions targeting unmet psychological needs are indicated where dependency compensates for offline deficits (Khalifa et al., 2026).

13 Conclusion

This review set out to do more than chronicle five decades of media dependency scholarship. Tracing the tradition from the foundational Uses and Dependency Model through Internet-era adaptations to the New Media Uses and Dependency Effect Model—using a documented, transparent review method—it found a theory whose causal core has proven durable while its objects, mechanisms, and measures have been repeatedly rebuilt. The review's original contribution is to consolidate that trajectory into the Integrated Media Dependency Framework: a stratified, distributed, and recursive account of dependency, expressed in a figure and seven falsifiable propositions, that extends the tradition to the platformized, algorithmically curated, and AI-mediated environment audiences now inhabit and that is explicitly positioned against Communication Infrastructure Theory, Media Repertoire Theory, Affordance Theory, and uses and gratifications.

The review's limitations should be acknowledged. As a theoretical review, it synthesizes rather than statistically aggregates evidence; its corpus is weighted toward English-language scholarship; and the framework it proposes, while grounded in the reviewed literature at every point, awaits the empirical testing that Section 11 maps out. Within those limits, the conclusions stand: dependency remains among the most productive relational accounts of media influence available to the field; its contradictions and unresolved debates are now explicit rather than latent; and its future lies in treating algorithms and AI agents—not only media systems and platforms—as parties to the dependency relations through which media power operates. If the framework proposed here helps researchers measure that power more precisely, helps designers and regulators govern it more responsibly, and helps educators and practitioners equip audiences to live with it more adaptively, the integration promised in this article's title will have done its work.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/journals/communication/articles/10.3389/fcomm.2026.1848139/abstract#supplementaryMaterial

Supplementary material Supplementary material at the source

Footnotes

Edited by: Pradeep Nair, Columbia University, United States

Reviewed by: Arnolt Pakpahan, Universitas Trisakti Fakultas Ekonomi dan Bisnis, Indonesia

Prateek Khanna, ICFAI University, India