When humans and agents form a collective: a framework for collective intelligence in persistent organizational human-AI teams

1 Introduction

Artificial systems increasingly participate in organizational work as recurrent contributors to interdependent activity. Existing research illuminates complementary focal levels: artificial intelligence (AI) studies foreground system capabilities; adoption and reliance research examines human responses; experiments often assess dyadic or episodic performance; and team research offers constructs developed primarily for human groups. Together, these streams provide complementary foundations from which a common theoretical architecture for studying the persistent collective as a unit can be developed.

When human and artificial contributions are organized through roles, decision rights, and repeated interdependence, the collective they constitute becomes the relevant theoretical object. Analysis therefore shifts from isolated components or episodes to the organization of interdependent contributions within the collective (Cooke et al., 2013; Larson and DeChurch, 2020; Naikar et al., 2023).

The organismic human-artificial intelligence (OHAI) framework defines this object as the persistent organizational human-AI team: a bounded collective in which distinguishable human and artificial contributors engage in interdependent activity across successive cycles and prior collective activity can affect later artificial participation. This definition excludes one-shot, independently reset, and statically aggregated interaction while making no assumption of psychological symmetry between contributors.

Here, organismic denotes an analytical focus on the bounded collective as an interdependent, history-sensitive organization of heterogeneous contributions; it does not imply biological equivalence or ontological unity.

OHAI takes collective intelligence (CI) as its primary theoretical target because its focal question is whether and under what conditions an organized human-AI collective exhibits a capacity that generalizes beyond particular task episodes. Consistent with established research, CI is understood as a general group-level capacity manifested across diverse tasks (Woolley et al., 2010; Riedl et al., 2021). The scope of this generalization must be specified in relation to an appropriate universe of work-relevant tasks and empirically established rather than assumed. Persistence makes this question consequential because prior collective activity can shape how contributions are organized and used across subsequent tasks. This focus does not subsume team effectiveness, decision quality, adaptation, collective learning, resilience, or complementarity, which address related but distinct questions. Its priority identifies the focal research problem of OHAI, not a required target for every study of an OHAI-relevant team.

OHAI is a bounded integrative theoretical framework that organizes inquiry around three analytically distinct objects: system configuration, collective functioning, and CI. This architecture provides a grammar for specifying and testing relations among these objects.

Persistence makes those relations history-sensitive by allowing prior collective activity to shape later interaction (Ilgen et al., 2005; Marks et al., 2001; Gorman et al., 2017; Ramos-Villagrasa et al., 2018). The direction and consequences of that history, including learning or improvement, remain empirical questions. Human-only teams serve a secondary comparative role for examining selected cross-substrate correspondences. Causal relations are introduced through local theories and empirical tests, while evidence accumulates through staged tests of bounded claims.

The payoff is a more precise organizational agenda: how are human and artificial contributions configured through roles, decision rights, authority, responsibility, and other organizational arrangements; which configurations are associated with which forms of collective functioning; and how are CI and its variation over time associated with consequences for workers, teams, and organizations?

Table 1 summarizes OHAI's core analytical elements, their concise definitions, and the principal analytical or empirical clarification associated with each element.

Table 1. Core analytical elements of the OHAI framework.

OHAI builds on adjacent traditions that locate team cognition in coordinated interaction (Cooke et al., 2013), conceptualize team processes as recurrent and history-sensitive (Marks et al., 2001; Ilgen et al., 2005), and distribute cognition and work across people, technologies, representations, and organizational arrangements (Larson and DeChurch, 2020; Naikar et al., 2023; Sidji et al., 2024). It applies these commitments to a more tightly bounded object: the persistent organizational human-AI team.

More directly, Gupta and Woolley (2021) conceptualize AI within transactive systems, Aggarwal et al. (2025) develop a theory-generating extension connecting transactive memory and collective identification to human-AI collective intelligence, and Gonzalez et al. (2026) focus on conditions for human-AI complementarity.

OHAI differs in three respects: it defines the persistent organizational human-AI team through interdependence, organizational boundedness, and functional continuity; it separates system configuration from enacted collective functioning and CI, preventing component-to-team inference; and it treats collective history as constitutive, making recursive updating, reconfiguration, and CI trajectories explicit empirical objects. Taken together, these distinctions enable researchers to specify and test whether and how distributed human and artificial contributions become, or fail to become, collective phenomena within organizationally configured, history-sensitive teams composed of psychologically asymmetric contributors.

Table 2 summarizes OHAI's positioning relative to these adjacent theoretical frameworks and traditions.

Table 2. Positioning OHAI relative to adjacent theoretical frameworks.

2 Defining the OHAI domain

OHAI addresses a deliberately bounded organizational phenomenon. Its unit of analysis is the persistent organizational human-AI team: a collective in which distinguishable human and artificial contributors participate in interdependent organizational activity across successive cycles. The unit includes the contributors, the activity that connects them, their functional roles and interdependencies, and the arrangements through which collective boundaries, authority, and responsibility are defined. This focus is consistent with interactional and sociotechnical accounts of teamwork (Cooke et al., 2013; Larson and DeChurch, 2020; Naikar et al., 2023). The unit is therefore not the artificial system alone, an individual user's response to it, or a single human-AI exchange.

The term team denotes the organization of interdependencies among contributions. It does not imply that human and artificial contributors possess equivalent psychological capacities or experience the collective in the same way.

An OHAI-relevant configuration must satisfy several conditions jointly. Human and artificial contributions must be distinguishable and directed toward shared organizational work. Their relation must be interdependent, such that what one contributor provides affects what others can interpret, decide, or do, rather than consisting of outputs produced independently and combined only afterward. The collective must also have recognizable functional roles and boundaries, with identifiable allocations of decision authority and responsibility. Finally, participation must continue across successive cycles. These requirements establish a domain threshold, not a scale of technological sophistication or team effectiveness.

Persistence is defined by functional continuity. Some state, representation, or interaction history generated through prior collective activity must be available to the artificial contributor and capable of affecting its later outputs or actions. The location of that state is immaterial to the definition. It may reside within an agent, in external memory, in a retrieval system, in an orchestration layer, in a shared knowledge infrastructure, or in another architecture that makes relevant history usable during later participation. A technically stateless base model can therefore belong to a persistent OHAI configuration when the deployed system can access and use externally maintained state. Conversely, storing records is insufficient if they cannot enter subsequent interaction. The relevant criterion is whether the collective's prior activity can become functionally consequential for later artificial participation.

One-shot interaction and recurrent prompting in which every episode functionally resets the artificial contribution therefore lie outside the domain. Static aggregation of separately produced human and AI outputs likewise lacks the required interdependence. Configurations in which only human members retain and use the shared history also fall outside the focal domain: humans may learn about the system, but the artificial contribution does not participate through that history. Such arrangements remain relevant to the broader study of AI use, but they do not instantiate the persistent collective that OHAI takes as its focal unit.

Artificial participation is accordingly defined functionally rather than ontologically or by technology label. An artificial contributor is a distinguishable deployed system whose task-relevant outputs or actions enter and influence collective activity. Whether it is called a tool, advisor, teammate, autonomous agent, embodied agent, or multi-agent subsystem does not determine eligibility. A system described as a tool may participate in an OHAI-relevant configuration, while a system marketed as a teammate may not.

Autonomy, embodiment, specific memory architectures, adaptation, and partner-modeling capabilities describe features of a configuration; none is individually necessary or sufficient for domain membership.

Human and artificial contributions can be functionally integrated while remaining psychologically asymmetric. The functional descriptions used here therefore do not imply consciousness, subjective experience, intentions, belonging, trust, or other psychological states.

This domain boundary determines the evidential status of studies. Direct OHAI evidence comes from configurations that satisfy the preceding criteria and examines the persistent hybrid collective at the relevant team or CI level. Repeated AI use is not enough if either the configuration or the phenomenon studied remains individual or episodic. Component evidence instead isolates a capacity, response, interaction, or process relevant to OHAI without instantiating the complete domain. Short human-AI dyads, bounded task experiments, one-shot studies, and tests of individual system functions can establish feasibility, refine measures, or motivate focused questions. They cannot by themselves demonstrate collective functioning or collective intelligence in a persistent organizational human-AI team.

Human-team comparator evidence supplies constructs and reference phenomena that can motivate questions about hybrid collectives, but it neither establishes that the same phenomena occur there nor makes human teams a normative baseline. Boundary or analogical evidence from fully artificial multi-agent or other out-of-domain systems can clarify computational possibilities, constraints, or contrasts but is not direct OHAI evidence. These categories indicate inferential proximity to the OHAI domain rather than evidential quality. Study type, including empirical investigations, reviews, conceptual contributions, and computational demonstrations, and publication status are distinct from domain proximity. Sources should be interpreted according to their design, scope, and review status.

3 OHAI organizing architecture

OHAI makes persistent organizational human-AI teams theoretically tractable by distinguishing three analytical objects: system configuration, collective functioning, and collective intelligence, while treating downstream consequences as analytically separate criteria.

The distinction is the framework's central organizing architecture. It specifies what kind of phenomenon is being described at each point in an explanation and what evidence would be needed to relate one kind to another.

Local theories can specify relations among these analytical objects, and empirical tests can evaluate those claims. Their ordering in OHAI serves analytical organization, while causal roles are defined within the particular theory under examination. OHAI therefore provides a grammar for formulating and testing explanations.

System configuration comprises the pre-specifiable human, artificial, task, and organizational features of the hybrid system under study. It is a structured, potentially time-indexed description rather than a latent construct, global score, or measure of technological sophistication. Human features may include expertise, experience, and task-relevant capacities; artificial features include relevant capabilities, specialization, reliability, information access, available state, updating functions, interaction modalities, autonomy, and output or action repertoire. Task features include complexity, uncertainty, information distribution, and interdependence, while organizational features include roles, authority, decision rights, responsibility, procedures, communication arrangements, and technical or knowledge infrastructures. Temporal features such as duration, pacing, interaction frequency, membership continuity, and access to prior activity may also vary within the persistence boundary. Which features matter depends on the theoretical claim, and no study must measure the full profile or reduce it to a common scale (Larson and DeChurch, 2020; Naikar et al., 2023).

Collective functioning concerns what occurs when configured contributors engage in interdependent activity. It directs attention from available resources to their realized organization and use: how expertise and information are accessed, distributed, communicated, evaluated, combined, and updated; how dependencies and roles are coordinated; and how contributors detect errors, provide feedback, repair breakdowns, or adjust their activity. Interaction patterns, role differentiation, representational alignment, and partner-related modeling belong here only insofar as they are manifested in or consequential for collective activity. The same requirement applies to emergent states and team processes, which must be established as genuinely team-level phenomena rather than inferred from an individual response or an artificial feature. This emphasis follows Interactive Team Cognition in treating coordinated interaction as a principal locus of team cognition (Cooke et al., 2013). Collective functioning is not a single macro-construct or global index. It is an analytical domain containing distinguishable processes, states, patterns, and failures whose relevance depends on the question under study. Coordination, for example, describes how collective work is organized; it is neither equivalent to CI nor evidence of effective performance by definition (Marks et al., 2001).

This level discipline is consistent with multilevel team-effectiveness research, which distinguishes member and compositional characteristics from team processes, emergent phenomena, and collective outcomes, while treating emergence and adaptation as phenomena that must be established at the appropriate analytical level (Kozlowski and Ilgen, 2006). For OHAI, this distinction is particularly important because properties implemented in an artificial component can otherwise be mistaken for processes or states realized by the team.

Collective intelligence denotes a general group-level capacity manifested across diverse tasks (Woolley et al., 2010; Riedl et al., 2021). Its scope of generalization must be qualified relative to the universe of tasks represented in a study. Episodic task performances are analytically distinct from the capacity inferred from their common pattern, although they may provide indicators for its estimation. Other performance measures, including new tasks or task families excluded from that estimation, may serve as external criteria. Informational synergy captures a different property: information available from contributions jointly that is unavailable from the components considered separately (Rosas et al., 2020; Mediano et al., 2022). Coordination, representational alignment, role differentiation, and particular emergent states likewise describe distinct aspects of collective functioning that may be examined in relation to CI. Designating CI as OHAI's primary theoretical target establishes the priority of a research question rather than assigning CI a fixed causal position. Its role as an outcome, predictor, or time-varying phenomenon depends on the local theory and empirical design.

Broader employee, team, and organizational criteria, such as learning, agency, wellbeing, innovation, or organizational performance, remain analytically distinct from CI. Local theories may position them as conditions, correlates, mechanisms, or consequences. OHAI does not presume that higher CI produces them uniformly, directly, or beneficially.

Socially minded intelligence provides a further adjacent but distinct construct. Bingley et al. (2026) conceptualize it in terms of flexible perception, thought, and action at individual, subgroup, and collective levels in relation to superordinate goals. OHAI does not use socially minded intelligence as a synonym for CI: the former concerns orientation across levels of social agency, whereas CI here remains a general group-level capacity manifested across diverse tasks. Socially minded processes may therefore be examined in relation to collective functioning or CI without being incorporated into the definition of CI itself.

OHAI's fundamental inferential rule is that component properties do not by themselves establish team-level phenomena. For CI, this rule entails a further distinction.

Evidence of cross-task generality does not by itself establish that the observed capacity depends on the organization of interdependent contributions rather than predominantly on a highly capable component. OHAI therefore distinguishes evidence for a general capacity in the performance of the observed team from evidence concerning the collective organization that supports that capacity.

Claims about the latter require evidence showing how differentiated contributions are organized and made consequential in collective activity. Attributing a causal contribution to that organization additionally requires a design capable of supporting the specified causal inference. This requirement does not define CI as superiority to the best-performing component. Advantage relative to an individual contributor, an artificial system, or an aggregation baseline is a further empirical claim whose meaning depends on the comparison specified.

The inferential rule also operates in the opposite direction: a team-level function need not depend on one particular artificial architecture. Similar collective functions may be realized through different combinations of human practices, artificial features, task structures, organizational arrangements, and shared infrastructures, while nominally similar configurations may produce different forms of functioning. OHAI therefore treats multiple realization as an empirical possibility and specifies no uniform AI capability or architecture as necessary or sufficient for a collective function or for CI.

This architecture supports local tests of relations between configuration and collective functioning, between collective functioning and CI or task performance, and between CI and broader consequences. Such relations may be absent, conditional, nonlinear, or reciprocal; their status depends on the theoretical claim and empirical design rather than on their position in the framework.

Figure 1 summarizes OHAI's organizing architecture and its history-sensitive extension across successive observations of the same persistent team.

4 Collective functioning in persistent human-AI teams

Collective functioning concerns how distributed contributions become organized and used in interdependent activity.

OHAI examines this domain through three selective lenses: theory of mind (ToM)-related partner modeling, transactive memory system (TMS)-related coordination, and shared mental model (SMM)-related representational alignment. These lenses were selected because they capture three complementary functional problems in distributed collective cognition: how contributors anticipate one another, how a team organizes and uses distributed expertise, and how task-relevant representations become sufficiently compatible for coordinated activity. Together, they illustrate distinct ways in which team-level functions may be realized across psychologically asymmetric human and artificial contributors. They serve as focused analytical entry points into collective functioning, while broader causal relations remain to be specified through local theory and empirical test.

Other forms of collective functioning may also matter in particular configurations. Human-team evidence illustrates emergent role differentiation (Goldstone et al., 2024), while fully artificial research provides boundary evidence concerning coordination and joint planning (Agashe et al., 2025). Communication, information routing, and breakdown repair are further candidate forms of collective functioning within OHAI when they are realized at the team level.

4.1 ToM-related partner modeling

ToM-related partner modeling refers broadly to representations of other contributors that matter for interdependent activity. On the human side, these may concern what others know, their capabilities and expertise, likely actions, intentions, or other task-relevant states. Some involve mental-state attribution; others are learned expectations about competence, roles, or behavior. On the artificial side, comparable task functions may be supported by a user or team-member model, state tracking, latent-variable inference, behavioral prediction, or updating from interaction history. These mechanisms may estimate knowledge, preferences, capabilities, or likely responses. They are computational representations and operations, not evidence of subjective Theory of Mind or psychological understanding.

The OHAI-relevant phenomenon is whether partner-related representations enter collective activity. A human may hold a detailed but inaccurate or unused model of an artificial partner, while an artificial system may predict human behavior without improving interdependent work. Team-level partner modeling becomes visible when such representations help contributors direct information, adapt contributions to others' likely needs, allocate activities, anticipate dependencies, or revise coordination following an unexpected response. Representations need not be identical or mutually accessible. What matters is their enacted role in organizing activity.

Available studies provide indirect but theoretically relevant evidence for partner modeling. Westby and Riedl (2023) combine human-team data with computational modeling of partner-related states, while Kleiman-Weiner et al. (2025) show in fully artificial settings that updated partner representations can condition cooperative action. Together, these studies establish human-team, computational, and boundary evidence that motivates partner modeling as an OHAI-relevant empirical target. Direct evidence of partner modeling as a realized collective function in persistent OHAI teams remains to be established.

Jorge et al. (2024) examine candidate observable cues of human trustworthiness in a bounded human-agent experiment, providing an empirical basis for computational partner modeling.

These cues define candidate inputs for computational partner modeling that could inform subsequent interaction and reliance regulation. Their use in adaptive reliance regulation within persistent OHAI teams remains to be demonstrated and they should be understood as computational assessments rather than interpersonal trust on the AI side.

4.2 TMS-related coordination

Transactive memory systems provide a second lens by focusing on the organization and use of distributed expertise. In Wegner (1987) foundational account, transactive memory arises when members combine differentiated individual knowledge with knowledge about where relevant expertise resides in the group. Subsequent team research conceptualizes TMS as a distributed cognitive architecture commonly characterized through specialization, credibility, and coordination (DeChurch and Mesmer-Magnus, 2010). Specialization concerns the differentiation of knowledge across members; credibility concerns judgments about the reliability of task-relevant sources; and coordination concerns the organized location, retrieval, and use of distributed knowledge. “Who knows what” therefore describes more than an expertise directory: it becomes a team-level phenomenon when differentiated knowledge is accessed and made consequential through transactive activity.

In a persistent human-AI team, three loci must remain distinct. Human members may hold psychological beliefs about where expertise resides and how reliable particular sources are. The artificial contributor may maintain expertise or source representations, retrieval indexes, routing policies, role models, or computational estimates of source reliability. At the team level, TMS-related coordination concerns whether distributed expertise is actually located, evaluated, retrieved, and used. It may be visible when questions are directed to an appropriate contributor, information is routed to where it can be used, responsibilities are differentiated by expertise, or source selection changes following experience. An AI expertise map or retrieval mechanism does not establish this phenomenon. It may support coordination, remain unused, or organize only the artificial component's internal operations.

TMS-related coordination is multiply realizable through different combinations of stable roles, organizational directories, communication routines, human expertise beliefs, shared knowledge infrastructures, and artificial expertise representations. What matters is whether these arrangements make differentiated knowledge accessible and usable through collective activity. Computational source estimates may support this function while remaining psychologically distinct from human credibility judgments.

Existing evidence supports TMS-related coordination as a relevant empirical target. Human-team research links TMS to team processes and performance (DeChurch and Mesmer-Magnus, 2010), while Gupta and Woolley (2021) extend transactive-systems logic conceptually to human-AI collective intelligence through functions such as updating, allocation, and retrieval.

Aggarwal et al. (2025) examine relations among TMS, collective identification, and team effectiveness in longitudinal human-team research; their implications for CI and human-AI collectives remain theory-generating.

More circumscribed evidence comes from simulated clinical tasks involving newly composed human-AI teams (Bienefeld et al., 2023), exploratory cross-cultural self-report measurement at the individual level (Akgun and Toker, 2026), and multimodal modeling of specialization, credibility, and coordination in human-only teams (Biancardi et al., 2024).

Taken together, this literature motivates TMS-related coordination as a team-level empirical target while leaving a psychologically shared TMS in persistent human-AI teams unestablished.

4.3 SMM-related representational alignment

Shared mental models provide a third lens on task-relevant representations. In human teams, SMMs are shared or compatible knowledge structures concerning goals, roles, procedures, constraints, system states, expected actions, and interdependencies. Their significance lies in supporting anticipation and coordination, not in requiring identical representations (Andrews et al., 2023). In hybrid teams, humans may hold psychological representations of the task and team, whereas an artificial contributor may maintain computational task states, role representations, constraint models, or action predictions. An AI-maintained task representation is not a human mental model and does not by itself establish a psychologically shared representation.

The OHAI-relevant target is representational alignment or compatibility in collective activity. It concerns whether human and computational representations support compatible task interpretations, role and dependency expectations, coordination-relevant predictions, and the detection or repair of consequential mismatches. Alignment need not mean uniformity. Differentiated representations may remain compatible, while convergent representations may be jointly inaccurate.

Available evidence remains primarily conceptual, review-based, and component-level. Andrews et al. (2023) show that hybrid SMMs remain theoretically and methodologically unsettled, while Narayanan et al. (2025) emphasize the need to distinguish representations attributed to individual contributors from cognition manifested through interaction.

Adjacent aviation reviews and conceptual work indicate that mismatches in capabilities, roles, task allocation, and situational representations can disrupt coordination and require continuing alignment as conditions change (Korentsides et al., 2026; Li D. et al., 2026). These findings support treating representational compatibility as an evolving property of collective activity rather than inferring a hybrid SMM from similarly labeled internal models.

The team-level claim therefore requires evidence that representations are interpreted, coordinated, challenged, or repaired through interdependent activity.

The three lenses isolate different functions. Partner modeling concerns representations of contributors; TMS-related coordination concerns the organization and use of distributed expertise; and SMM-related alignment concerns the compatibility of task-relevant representations. They may intersect while retaining distinct empirical targets and relationships to collective functioning. Each requires evidence at the team level and can be examined in relation to CI without being treated as constitutive of it.

5 Persistence, recursive updating, and CI trajectories

Persistence changes the theoretical object studied by OHAI because it makes collective history consequential. Once prior collective activity can affect later participation, the team cannot be represented adequately as a sequence of independent encounters. Recursive updating, change across different temporal rates, and possible variation in collective intelligence follow from this history sensitivity. They specify what becomes theoretically and empirically visible when a persistent collective has a consequential past.

5.1 Persistence and recursive updating

When the functional-continuity condition defined in Section 2 is met, outputs, states, representations, roles, routines, expectations, or interaction histories formed through earlier activity may enter the conditions of subsequent activity.

• Recursive updating names this temporal relation. It may affect only some system elements, occur asymmetrically across contributors, and arise without deliberate or central control.
• Recursive updating may occur at distinct loci. Human contributors may revise expectations, expertise beliefs, or interpretations of tasks and roles. Artificial contributors may update or retrieve stored interaction history, task or role representations, source estimates, routing rules, or other computational state. Enacted role allocations, interaction routines, or patterns of coordination may change at the team level, while procedures and shared infrastructures may change organizationally. These updates can affect one another without becoming equivalent or forming a single feedback mechanism. A change in artificial state remains a component or configuration change even when it affects collective activity; a revised human expectation remains an individual state even when it becomes consequential for interdependent functioning.
• Recursion is also broader than regulation. Regulation consists of activities through which a team monitors, evaluates, corrects, plans, or adjusts its functioning. Recursive updating occurs whenever prior activity modifies later conditions, including when nobody deliberately evaluates or corrects the process.

Human-team research situates regulation within broader history-sensitive change. Konradt et al. (2015) distinguish guided reflexivity and feedback interventions from the reflection and adaptation enacted in response, while recurring-phase, input-mediator-output-input (IMOI), dynamical, and complex-adaptive traditions treat team episodes as interdependent and history-sensitive (Marks et al., 2001; Ilgen et al., 2005; Gorman et al., 2017; Ramos-Villagrasa et al., 2018). For OHAI, the key implication is path dependence: recursive histories may support improvement or recovery, but may also stabilize error or ineffective routines. Adjacent evidence illustrates this possibility at different loci. Simón et al. (2024) show organizational history dependence in a four-year human-AI integration case, while Park et al. (2023) show computationally how stored observations can shape later behavior in fully artificial agents. Together, these studies motivate direct tests of team-level recursive updating and CI change in persistent OHAI teams.

5.2 Overlapping timescales of collective change

History-sensitive change can unfold at different temporal rates. Some adjustments occur within an interaction or task episode, while others accumulate across episodes in roles, routines, expertise structures, expectations, procedures, or persistent representations. These rates coexist within the same collective, and the same phenomenon may unfold across more than one timescale. Changes at different rates can also shape one another, as repeated local interactions consolidate more durable routines or persistent structures influence later information use and coordination. The significance of alignment or misalignment across rates depends on the task and configuration.

Human-team evidence shows that process gains and losses can emerge over time rather than uniformly at team formation (Watson et al., 1998), while dynamical-systems perspectives provide concepts such as stability, perturbation, and hysteresis for describing history-dependent change without assuming that persistent human-AI teams already instantiate a validated dynamical model (Gorman et al., 2017; Favela, 2020). Persistent artificial state and accumulated interaction history may connect these rates only when they become consequential for later collective activity.

5.3 CI trajectories

History-sensitive collective functioning makes variation in CI over time an empirical possibility. A CI trajectory is the temporal pattern in repeated estimates of the same collective-intelligence construct across observation periods. It is therefore distinct from repeated single-task performance and from temporal changes in coordination, informational synergy, reliance, representational alignment, or other processes, which may be associated with CI without constituting change in the general collective capacity itself.

This distinction follows from CI's multi-task basis. Performance in one episode can reflect task difficulty, local information distribution, temporary coordination problems, novelty, role mismatch, or stochastic variation in artificial outputs. A change in that score may therefore occur without a change in the broader capacity. Woolley et al. (2010) identified CI through common variation in group performance across diverse tasks, and Riedl et al. (2021) likewise distinguished collective capability from performance in a particular setting. The minimum logic for estimating a CI trajectory is consequently repeated estimation of the same construct through comparable or appropriately linked task batteries, with sufficient task diversity to separate general capacity from episodic performance. This requirement does not prescribe a particular battery or longitudinal protocol.

In specialized organizational settings, an appropriate multi-task basis need not consist of generic tasks disconnected from the team's work. A study could define a bounded universe of work-relevant tasks in advance and sample substantively different cognitive demands within it, such as diagnosis, forecasting, constrained planning, and the generation and evaluation of alternatives. Multiple instances within each task family would help distinguish performance specific to one family from broader generalization across the declared universe. Surface variations of the same procedure would provide a weaker basis for such a claim.

Across observation periods, studies could repeat the same battery, use alternative forms connected through common tasks or bridge samples, or adopt another justified design that establishes a common measurement reference. Repeating the same tasks preserves their content but does not by itself establish measurement comparability, because familiarity and task-specific learning may alter what performance reflects. Common tasks or bridge samples may provide a reference for relating scores across forms, provided that the relevant measurement relationship remains sufficiently stable for the intended comparison. Shared task labels or scoring rubrics do not by themselves establish comparability. The measurement properties required depend on the intended claim, but when no defensible link can be established, change in a unitary CI score or trajectory should not be inferred. Studies may instead report differentiated task-family profiles for each period, restricting longitudinal comparisons to measures whose comparability is supported.

Adjacent measurement work reinforces the need to define and repeatedly estimate the same CI construct over time. Riedl and Weidmann (2026) separate individual problem-solving ability, collaborative ability, task difficulty, and momentary variation in episodic human-AI performance. Their provisional approach concerns episodic performance rather than longitudinal CI, but it clarifies why changes in collaborative performance can support claims about CI change only when the broader collective capacity has been appropriately defined and repeatedly measured.

Interpretation requires three questions to remain distinct: whether the same collective continues, whether its claim-relevant configuration remains sufficiently stable, and whether the CI measure remains comparable across observation periods. A team may retain a recognizable line of work, interdependence, responsibility, and functionally available history while undergoing material reconfiguration. Reconfiguration therefore does not necessarily create a new collective, but it limits attribution of observed change to development under stable conditions. Changes in the artificial model or version, memory architecture, information access, human membership, authority structure, roles, or organizational procedures should be documented and located within the team's history. When the collective continues and the measure remains appropriately linked, change across a reconfiguration may still be described as change in the capacity of the continuing team. Attributing that change to accumulated experience, learning, or the reconfiguration itself requires a comparison appropriate to the claim. Repeated estimates may reveal improvement, stabilization, deterioration, oscillation, recovery, or persistent variability, all of which remain empirical patterns rather than expected developmental stages.

Persistence thus makes collective history theoretically consequential; recursive updating describes how prior activity enters later conditions; overlapping timescales describe how resulting changes unfold and interact; and CI trajectories provide a way to examine whether the collective capacity itself varies across that history.

6 Cross-substrate correspondence as a comparative program

Cross-substrate correspondence is a secondary comparative program within OHAI that asks whether selected human-team phenomena recur or perform comparable functions in persistent organizational human-AI teams. Human-only teams serve as epistemic references rather than normative standards, and any supported claim is bounded to the phenomenon, representation, conditions, and correspondence form tested.

6.1 Forms of cross-substrate correspondence

Any cross-substrate claim must specify what is being compared, between which systems, under what conditions, and at which analytical level. OHAI distinguishes three distinct, non-cumulative forms of correspondence. Observable-pattern correspondence concerns whether a prespecified behavioral, interactional, temporal, or coordination pattern recurs under sufficiently comparable conditions. Functional correspondence concerns whether differently realized elements or processes perform a prespecified comparable function in collective activity. Similar observable patterns need not perform the same function, and comparable functions need not be realized through similar patterns; evidence for one form therefore does not establish the other.

Relational-pattern correspondence concerns whether a specified team-level organization of relations, such as role dependencies, expertise routing, or information flows among functionally defined elements, is preserved across human-only and hybrid teams despite differences in composition and substrate. OHAI treats this third form only as a limited future methodological possibility: any test would need to prespecify the unit of analysis, relevant elements and relations, level of abstraction, cross-system mapping, and criteria for meaningful correspondence and non-correspondence; similar behavioral patterns or outcomes, comparable functions, analogous failures, or use of a common formalism would not alone provide sufficient evidence.

6.2 Overarching comparative claim and local test families

The scope of the comparative program is expressed by the following overarching claim: selected CI-relevant patterns, functions, and processes may exhibit cross-substrate correspondence between matched human-only and OHAI-relevant hybrid teams.

The terms “selected” and “may exhibit” limit the claim to specified empirical possibilities. It scopes the comparative program rather than serving as an omnibus proposition for a single study.

Four correspondence families (C1–C4) organize separable local empirical questions. Each test must specify the relevant correspondence form and its evaluation criterion. The families may be examined independently and do not combine into a framework-wide correspondence score.

C1 concerns interaction-pattern correspondence. A local test asks whether a pattern reliably observed in a human-only reference condition recurs in a matched OHAI-relevant hybrid condition. The pattern, its representation, metric, matched conditions, and decision criterion must be specified in advance. Support concerns only that pattern under those conditions.

Bellina et al. (2026) provide a provisional boundary illustration: individual multimodal language models aligned systematically with simulated majority responses, but because social influence was implemented through prompted descriptions rather than agent interaction, the finding identifies a candidate observable pattern rather than a team interaction pattern.

C2 concerns cross-substrate correspondence in informational synergy. A local test first establishes, within each system, whether collective interaction produces information unavailable from the relevant components or an appropriate aggregation baseline, using the same or explicitly linked informational representation and a prespecified decision criterion. It then asks whether the resulting synergistic property or profile corresponds across matched human-only and OHAI-relevant hybrid teams. Informational synergy remains distinct from generic performance gain, complementarity, coordination quality, and CI.

Abril-Bermúdez et al. (2026) provide a provisional boundary illustration showing that statistical synergy can arise without direct interaction among components, reinforcing that informational synergy alone does not establish a particular coordination mechanism or form of collective functioning.

C3 concerns failure-process correspondence, defined by a prespecified process signature. Tests should specify the perturbation or relational condition and the relevant process dimensions, such as onset, breakdown, propagation, or recovery. Hidden-profile research illustrates why outcome similarity is insufficient: human groups and fully artificial agents may both fail under distributed information, yet similar final errors do not establish the same underlying process (Stasser and Titus, 1985; Li Y. et al., 2026). Support therefore requires correspondence in the specified process signature.

C4 concerns correspondence in TMS-related coordination, SMM-related representational alignment, or another prespecified team-level coordination structure. Tests should target its enacted organization in collective activity while distinguishing human psychological representations from artificial computational properties, and prespecify the relevant structure, indicators, cross-system linking rule, and decision criterion. Developmental claims also require an appropriate observation period. Similar component labels or representations are insufficient evidence of team-level correspondence, which remains analytically distinct from CI.

Divergence from a human-team reference may reveal substrate-sensitive constraints, multiple realizations of a collective function, or genuinely hybrid forms of organization. The comparative question is therefore which patterns, functions, and processes correspond or diverge under specified conditions, rather than how closely hybrid teams reproduce human organization.

7 A staged empirical program for OHAI

OHAI is empirically usable as a framework for bounded claims, qualified measures, and relations specified at the correct analytical level rather than as an omnibus causal theory. The four stages organize research maturity rather than team development or a compulsory validation ladder: studies may enter at different stages, address only one, or return to qualification work as constructs or configurations change, with evidence accumulating by strengthening or constraining specific claims.

7.1 Stage 1. Configuration and measurement qualification

Strong inference begins by establishing the system and phenomenon under study. Direct OHAI evidence must show that the configuration meets the domain boundary, that the persistent organizational human-AI team is the relevant unit, and that the target is assessed at its proper level. Studies should document the human, artificial, task, and organizational features material to interpretation or replication. Relevant artificial features may include model and version, interaction-relevant prompting, stochastic settings, memory or persistent-state architecture, retrieval and information access, orchestration or routing, updating rules, and available outputs or actions; relevant human, task, and organizational features include expertise, team composition, interdependence and information distribution, roles, authority, procedures, and infrastructure. This profile is claim-dependent. Measurement qualification then requires either defensible cross-condition meaning and examination of equivalence or, when substrate-specific indicators are needed, a specified and validated functional or representational linking rule. Similar labels, summary scores, or implemented AI features are insufficient as evidence of measurement invariance or the corresponding team-level phenomenon.

7.2 Stage 2. Circumscribed local tests

Once the relevant configuration and measures are qualified, studies can test explicit local relations between selected configuration features and collective functioning, between collective functioning and task performance or CI, or through cross-substrate comparisons when a human-only reference is theoretically relevant. Each study should prespecify the local claim, relevant configuration features, target phenomenon and analytical level, indicators or representation, required comparison or baseline, and decision criterion. Cross-substrate tests must additionally specify the correspondence form.

7.3 Stage 3. Longitudinal team-level tests

Longitudinal evidence is required when a claim concerns recursive updating, changing coordination structures, development or deterioration of partner-related representations, change or development in TMS-related coordination or SMM-related representational alignment, or another history-sensitive form of collective functioning.

The central task is to distinguish trajectories of configuration, collective functioning, task performance, and CI. Repeated CI measurement is appropriate when CI is the target and the same general collective-capacity construct is estimated across periods using an adequate multi-task basis. Material changes in the AI model or version, memory, retrieval, information access, human membership, roles, decision authority, or organizational procedures must be documented so that observed change can be interpreted in relation to the continuing collective's configuration history; attribution requires comparisons appropriate to the claim. Improvement, convergence, and adaptation remain empirical interpretations.

7.4 Stage 4. Integrated multi-level studies

After the constructs, representations, indicators, and measures required for a question have been sufficiently qualified, integrated studies may examine selected relations among configuration features, collective functioning, recursive relations across time, CI, task performance, and downstream consequences. Researchers should distinguish hypothesized from exploratory relations and specify which variables serve as conditions, mechanisms, or outcomes. Human-perceived team membership, calibrated human trust or reliance, team regulation, and other processes or emergent states may be included when required by a local theory, but are not obligatory components or mediators and moderators by definition. Integration therefore tests selected relations rather than a universal causal chain or the complete OHAI system.

Interpretation across the program depends on task validity, construct and level-of-analysis validity, measurement or linking validity, statistical adequacy, specification of relevant AI-system variability, and sufficient observation or stabilization time when the claim requires it.

These safeguards should be specified before results are interpreted and cannot be expanded without limit post hoc to protect claims from non-support. A well-qualified negative result should constrain the local claim, narrow its boundary conditions, or motivate its abandonment.

Table 3 maps different empirical claims to the evidence they require and the uncertainty that remains. These claims do not form a single cumulative hierarchy: evidence about collective functioning, cross-task generality, collective organization, temporal variation, and causal effects addresses distinct inferential questions. A study may contribute valid evidence about configuration, collective functioning, or multidomain performance without estimating CI.

Table 3. Evidence and permissible inference for CI claims in OHAI.

Consider a study of persistent distributed-information teams, each comprising three humans and one artificial contributor across successive organizational tasks. Stage 1 would document the configuration features relevant to the study and define a bounded universe of work-relevant task families that differ in their cognitive demands and distribution of information.

Human and artificial capabilities could be assessed separately on relevant individual tasks as configuration features. Team performance could be evaluated across task families through prespecified, family-appropriate criteria, such as accuracy, forecasting error, or plan feasibility. Stage 2 could examine enacted TMS-related expertise coordination through observable sequences in which requests are directed to contributors with task-relevant expertise and their contributions are subsequently used, checked, or rejected with reasons. These process indicators would remain separate from both component-level expertise representations and the task performances used to estimate CI.

Stage 3 could use appropriately linked task forms, subject to the requirements in Section 5.3. Analysis across teams could examine whether performance supports a common capacity beyond task-family-specific patterns and whether that capacity predicts performance on new cases withheld from its estimation. When the evidence does not support that interpretation, studies would report differentiated task-family profiles rather than a unitary CI score. Material changes in models, memory, information access, membership, roles, authority, or procedures would be documented, while continuity of the collective and comparability of the measure would be assessed separately. For example, teams continuing with the same model could provide a contemporaneous comparison for teams receiving a model update, with comparable task exposure and other material changes documented. Such a comparison could help distinguish experience-related change from change associated with the update, although causal attribution would depend on the design and control of alternative explanations.

A later Stage 4 study could examine selected relations between configuration, enacted expertise coordination, CI, and external criteria.

8 Discussion

OHAI reframes human-AI collaboration as an organizational problem. When an artificial contributor participates persistently in interdependent work, the focal object becomes the organization of the collective: how roles, dependencies, information access, communication, decision rights, authority, responsibility, and accumulated history connect human and artificial contributions. This perspective extends beyond adoption, implementation, acceptance, or task automation and aligns with sociotechnical approaches that locate technology within the wider work system (Larson and DeChurch, 2020; Naikar et al., 2023). Organizations thereby configure the conditions under which AI functions become part of collective functioning.

The distinction between configuration and functioning directs organizational attention to how available capabilities enter everyday work. Their consequences depend on how outputs are interpreted and contested, which contributors can act on them, and how dependencies are coordinated. The same artificial system may therefore function differently across organizational arrangements, while different combinations of technology, human practices, and shared infrastructure may support similar collective functions.

Persistence adds an organizational implication: governance extends beyond initial deployment because retained collective history becomes part of the conditions of future work. Interaction histories, source estimates, role representations, routing patterns, and routines can support continuity and coordination, while also carrying forward errors, obsolete representations, inappropriate reliance, inefficiency, or biased allocation. A mistaken source classification may continue to shape routing, while an earlier procedure may become rigid as tasks or capabilities change. Persistence therefore makes the history carried into later activity an object of governance.

Organizational governance should specify decision rights, human responsibilities, oversight, and arrangements for challenge, override, and escalation. Their adequacy should be examined as roles, capabilities, and accumulated collective history evolve, because these arrangements shape how contributors can act on, question, and revise collective work.

The organizational consequences of hybrid teamwork also depend on how benefits, responsibilities, and demands are distributed among human members. A configuration that improves collective capability or task performance may increase verification work for some members or concentrate decision authority in particular roles. Evaluating these developments therefore requires attention to both collective outcomes and the relationships through which people contribute to and experience their work.

8.1 Collective intelligence and the conditions of working well

Persistent human-AI teams organize both collective problem solving and the relationships through which people experience their work. OHAI therefore opens an organizational question: how does artificial participation reshape the conditions under which human members can contribute, rely on colleagues, receive recognition, and experience belonging? Changes in expertise coordination and role expectations may alter whom members consult, whose judgments carry weight, and how responsibilities are distributed. These developments connect the framework's account of collective functioning to employee wellbeing. Human-team research provides a basis for examining these connections: transactive memory systems have been associated with job satisfaction and team identification (Michinov et al., 2008), while transactive memory and shared mental models have been linked to team engagement and satisfaction (Guchait, 2016). Their relevance to persistent hybrid teams requires examining how artificial participation changes these processes and their consequences for human members.

A multiple-identification perspective adds that changes within a team may be experienced differently through employees' attachments to their team, profession, and organization (Rovetta et al., 2025). A redistribution of expertise may strengthen mutual support while diminishing some members' sense of professional recognition. These relationships may also develop recursively. As changes in collective intelligence become evident through the team's capacity to address diverse tasks, interpretations of that capacity and attributions of contribution may reshape expectations, responsibilities, and relationships within and between groups. A team recognized as increasingly capable may receive greater autonomy and organizational support, but also more demanding assignments; recognition concentrated on particular contributors may change the standing of others. Such responses can become conditions of subsequent collective functioning. The Job Demands–Resources perspective provides one means of examining how these developments affect resources and demands associated with engagement and strain (Bakker and Demerouti, 2017). Collective intelligence and employee wellbeing consequently remain distinct outcomes whose relationships may vary across members, organizational arrangements, and time.

8.2 Making OHAI actionable: from research questions to organizational tools

A central opportunity for OHAI is to translate this relational perspective into practical tools and organizational recommendations. Research can establish how choices about task allocation, decision authority, and participation shape relationships among human members, and how these relationships evolve alongside collective capability and employee wellbeing. The practical aim is to identify which organizational arrangements sustain supportive working relationships, for whom, and under what conditions.

Several forms of practical support could emerge from this program. Role-and-expertise maps could help teams examine how formal allocations compare with actual patterns of consultation, contribution, and verification. Participative reviews could make differences in members' experiences of support, recognition, voice, and responsibility visible, including tensions between team and organizational expectations. Longitudinal assessment profiles could examine how changes in consultation, mutual support, and responsibility accompany changes in collective capability and employee wellbeing, including whether new patterns of dependence concentrate workload and strain in particular roles. These possibilities identify instruments to develop and evaluate, with measures appropriate to each phenomenon and analytical level.

Their organizational value would depend on connecting assessment to decisions. For example, evidence that improved collective capability accompanies increasing verification burdens could inform how checking responsibilities and available resources are allocated. Evidence that persistent expertise representations narrow whose knowledge is consulted could inform when role assignments and information-routing practices should be reviewed. Comparisons across configurations, participative field studies, and evaluations of organizational interventions could establish which responses improve working conditions, for whom, and under what circumstances. This would give OHAI a practical trajectory: from identifying the organization of persistent hybrid collectives to supporting informed decisions about how those collectives can function effectively and sustain employee wellbeing.

9 Conclusions

OHAI makes visible the persistent organizational human-AI team as a distinct theoretical object. It is an organized collective in which distinguishable human and artificial contributors participate repeatedly in interdependent organizational activity and prior collective history can affect later participation. Roles, dependencies, authority relations, information flows, and successive activity together define this focal unit.

The framework distinguishes system configuration, collective functioning, CI, and downstream consequences. Its central inferential rule is that component properties do not by themselves establish team-level phenomena; relations among these objects require evidence appropriate to the claim and level of analysis.

Persistence makes earlier activity part of the conditions of later activity, bringing recursive updating, path dependence, reconfiguration, and possible variation in CI over time into organizational inquiry. It creates history-sensitive collective organization whose trajectory remains empirical: accumulated history may support continuity, learning, or recovery, or preserve error, obsolete representations, or ineffective routines. Similar collective functions may be realized through different combinations of human capabilities, artificial features, task structures, organizational arrangements, and shared infrastructures. Functional integration therefore accommodates multiple AI architectures and psychological asymmetry between human and artificial contributors.

OHAI also opens a research agenda on how persistent artificial participation and changes in collective capability reshape human working relationships and the conditions of employee wellbeing, providing a basis for developing and evaluating practical tools and organizational recommendations.

10 Limitations

Direct evidence from persistent organizational human-AI teams, especially longitudinal evidence, remains limited. Available literature largely examines isolated artificial capabilities, bounded human-AI interactions, human-team reference phenomena, or systems outside the focal domain. These sources are useful for establishing computational possibilities, identifying candidate processes, and supporting construct development, while their relevance to collective functioning or CI in persistent hybrid teams remains indirect. OHAI therefore remains theoretical and programmatic, with claims proportionate to an informative but uneven evidence base.

OHAI's present generalizability is bounded by its focal domain of persistent organizational human-AI teams. Its applicability to transient or high-turnover collectives, loosely bounded or platform-based work, episodic human-AI interaction, highly asymmetric arrangements, and other out-of-domain configurations remains to be established. Some OHAI distinctions may prove useful in these settings, but extension requires evidence that the framework's unit, constructs, and relations retain appropriate meaning in the new context.

Construct validity and measurement remain substantial challenges. Component properties cannot substitute for team-level evidence; cross-substrate comparisons require validated functional or representational linking rules when indicators differ; and longitudinal CI claims require comparable estimates across periods that distinguish changes in collective capacity from episodic performance and material reconfiguration. Table 3 summarizes the evidence required for these claims and the uncertainty that remains.

Estimating CI may not be feasible or substantively justified in every OHAI setting. Highly specialized work may not provide sufficient task diversity or repeated observations, and performance may be better represented by task-family-specific dimensions than by a common capacity. In such cases, studies should limit their claims to qualified capability profiles, collective functioning, or task-specific performance rather than impose a unitary CI interpretation. This limitation constrains the available evidence for CI without removing the setting from the OHAI domain.

Rapid change in artificial configurations intensifies these limitations. Model or version, architecture, memory, retrieval, information access, stochastic behavior, orchestration, prompting, or deployment context may change while the same general product or capability label is retained. Such changes complicate replication and evidence accumulation, reduce longitudinal comparability, and can confound change in collective functioning with change in the artificial component.

OHAI's functional, configuration-based approach therefore directs attention to the features material to a claim, with interpretability depending on relevant deployment details being available and consistently reported, and on material changes being assessed for their implications for measurement comparability and attribution.

Finally, OHAI's non-omnibus architecture has limited predictive specificity. Substantive causal explanation depends on local theories with explicit boundary conditions, and evidence for particular relations supports those relations rather than validating OHAI as a complete causal system.

11 Future directions

A first priority is to explain how configuration features become realized in collective activity. Research should identify when available artificial capabilities are accessed, interpreted, and used by the team and how organizational arrangements enable, constrain, or redirect their use. Relevant features may include human expertise, artificial functionality, information distribution, task interdependence, roles, authority, communication arrangements, and shared infrastructure. These elements operate in combination: similar artificial systems may function differently across teams, tasks, or periods, while different combinations of human practices, computational features, procedures, and infrastructures may realize the same collective function. Explaining multiple realization can reveal which features are necessary, substitutable, or implementation-specific for a local phenomenon across alternative AI architectures.

A second priority is to explain which forms of collective functioning are associated with collective intelligence. Coordination, partner modeling, TMS-related coordination, representational alignment, informational synergy, and other processes may be relevant, but their relations to CI remain empirical. Research should determine when a function contributes to general collective capacity across tasks, when coordination generalizes across task classes, and how different functions complement, substitute for, or interfere with one another.

History-sensitive teaming creates a third priority: explaining the tension between continuity and reconfiguration. Persistent state may support learning, coordination, recovery, or useful context, while also carrying forward errors, obsolete representations, inefficient routines, biased routing, or inappropriate reliance. Research should examine which forms of collective history are functional to retain, revise, discount, or forget, under what conditions, and how memory and continuity generate benefits or liabilities. It should also determine when changes in models, information access, human membership, roles, or procedures constitute adaptation within a continuing team and when they materially reconfigure the collective. This distinction may depend on which relations and historical dependencies remain consequential rather than on whether the system retains the same name.

A fourth priority is to investigate the relational pathways connecting persistent hybrid teamwork to employee wellbeing, as developed in Section 8. Research should examine how changes in collective functioning and intelligence reshape relationships within and between teams, and how these relationships subsequently influence collective work. Longitudinal and intervention studies could identify organizational arrangements that sustain mutual support, recognition, and meaningful participation, while assessing how benefits and demands are distributed across members and roles. Such evidence would support the development and evaluation of the organizational tools outlined in Section 8.2.

Finally, future research should investigate cross-substrate divergence as a substantive source of theory development. Persistent human-AI teams may realize familiar functions through different organizations, develop interaction patterns without close human analogs, or display substrate-specific coordination structures and failure processes. Such cases can show where inherited human-team constructs remain useful, where they require revision, and where genuinely hybrid concepts are needed. Rigorous evidence of divergence requires adequate measurement and matched conditions and can reveal constraints and possibilities that a program centered primarily on correspondence would miss.

Generative AI statement

The author(s) declared that Generative AI was used in the creation of this manuscript. The author(s) used OpenAI's ChatGPT (GPT-5.6 Thinking, OpenAI, San Francisco, CA, United States) to support the translation of the manuscript from Italian into English, to improve grammar and readability, and to provide graphical support for Figure 1. All AI-assisted text and graphical content were critically reviewed, revised, and approved by the author(s), who take full responsibility for the final content.

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.

Footnotes

Edited by: Piers Steel, University of Calgary, Canada

Reviewed by: Kyana Van Eijndhoven, Tilburg University, Netherlands

Isabel Barbosa, Portucalense University, Portugal