Belief-sensitive norm nudging for household tap water use: an indirect-pathways argument

1 Introduction

Across many high-income countries, bottled water has become a routine substitute for tap water (e.g. Gleick and Cooley, 2009; Parag and Timmons Roberts, 2009). Household reliance on bottled rather than tap water, however, carries well-documented drawbacks. From an environmental standpoint, bottled water imposes burdens across extraction, processing, packaging, transport, and disposal; life-cycle assessments report markedly higher energy use and environmental impacts than municipal tap water (Gleick and Cooley, 2009). Economically, consumers often pay orders of magnitude more for bottled water despite the fact that much of it originates as municipal supply (Gleick and Cooley, 2009; Parag and Timmons Roberts, 2009). From a safety and governance standpoint, bottled water is not consistently superior in quality and is often subject to less visible public oversight than tap water, even though consumers frequently perceive it as safer (Doria, 2006; Parag and Timmons Roberts, 2009). Together, these features make the growth of bottled water consumption in high-income settings hard to make sense of on standard cost-benefit grounds.

In order to explain this puzzling phenomenon, sociologists have appealed to the notion of inverted quarantine: rather than pressing for collective provision of safe, high-quality public water, consumers purchase bottled water to insulate themselves from perceived environmental risks (Szasz, 2007; Willis and Schor, 2012). Why go through the hassle of campaigning for collective solutions to environmental risks, when you can—as sociologist Andrew Szasz aptly put it—shop your way to safety? This consumer strategy, however, may in the long run dull political pressure to maintain and upgrade public systems. Indeed, as bottled brands advertise health and purity, distrust of public water deepens—i.e. widespread exit into private protection reduces political pressure to fund and oversee tap systems, creating a feedback loop that further undermines confidence in public supply (Parag and Timmons Roberts, 2009). Most importantly, said exit can also reshape social norms in an unwelcome direction, as the visible prevalence of bottled water use shifts descriptive norms (what people do) while tacit approval shifts injunctive norms (what people approve). In effect, isolating oneself—rather than the perceived environmental threat—unwittingly signals that tap water is suspect: “if everyone buys bottled, tap must be unsafe.” This normative drift, as we shall see, is a major part of what norm-based interventions must confront.

This paper takes the above environmental and institutional backdrop as given and addresses a pressing behavioral question: how can we design effective interventions aimed at increasing household use of tap water and, by substitution, reducing bottled water use? This article develops a theoretical answer to that question by advancing an indirect-pathways account of norm-based intervention design.

We advocate green nudging—low-cost modifications of choice environments aimed at steering behavior without coercion—and argue that the most promising class of such nudges leverages social norms (what others do and approve of). As we shall argue, however, norm-based nudges cannot be treated as plug-and-play. They are context-sensitive policy tools, and their success or failure hinges crucially on the beliefs and motivations present in the target population. We frame this claim against a broader policy design insight: alongside natural, built, human, and social capital, societies can cultivate behavioral capital – the latent potential of behavior change to improve environmental quality (Beretti et al., 2013; Czap et al., 2012). Yet the very premise of behavioral capital implies a responsibility to diagnose which levers matter here and now. For norm-based interventions, that diagnostic task is empirical and, we argue, survey-based.

Our core thesis is twofold. First, social norms are powerful drivers of drinking-water choices, but their influence is often mediated rather than direct. People's personal norms (felt moral obligation), attitudes (evaluations of tap vs. bottled), and perceived behavioral control can transmit or block normative influence. Second, because those mediators—and the misperceptions that surround them—vary by locale and group, prior survey research is a precondition for designing norm-based nudges that actually work in situ. In short: to “nudge for tap” we must first measure beliefs.

This is a theoretical, hypothesis-and-theory paper. Its primary aim is not to report new behavioral data or to test a specific intervention experimentally, but to clarify what a belief-sensitive survey framework should measure if norm-based interventions for household tap-water use are to be designed responsibly. The proposed framework is concerned with perceptions, intentions, and behavior, but with an important distinction: perceptions and related beliefs are the main diagnostic target of the survey blueprint; intentions are treated as the main proximal outcome for survey-based modeling; and self-reported behavior serves primarily as a baseline/headroom indicator and, where available, as a downstream outcome to be interpreted alongside intentions. The paper's contribution is therefore to refine the conceptual and measurement logic linking perceived norms, mediating beliefs, intentions, and behavior in this specific domain.

At the level of mechanism, the present framework is individual-psychological. What affects behavior are not group-level regularities as such, but agents' perceptions of what relevant others do and approve, together with the personal norms, attitudes, and perceived behavioral control through which those perceptions are filtered. Yet these variables are not randomly distributed across persons. When measured across a target population, they reveal distinct local belief landscapes—patterned distributions of perceived norms, evaluations, and constraints through which normative influence is likely to operate differently from one setting to another. This is why local surveying is required before a norm-based intervention is designed. Figure 1 schematizes the indirect-pathways account defended in this paper. More specifically, it shows how perceived social norms may influence household tap-water choices not only directly, but also indirectly through personal norms, attitudes toward the behavior, and perceived behavioral control.

The paper proceeds as follows. Section 2 introduces green nudging, clarifies why the tap-over-bottled decision is a distinctive behavioral target, and situates it within the psychological and nudging literature. Section 3 focuses on how social norms are standardly taken to shape our behavior, exemplifies their context sensitivity through a concrete example, and draws the key lesson that nudge interventions must be grounded in prior survey-based research. It closes by posing the paper's central question: what kind of information about the targeted population should surveys focus on? Section 4 develops the central claim that social norms' influence is often indirect, working through personal norms, attitudes, and self-efficacy, and situates it within Fishbein and Ajzen's Reasoned Action Approach to behavior prediction and change, which assigns a central role to beliefs. From this, a methodological lesson follows: surveys informing the design of norm nudges should adopt belief-sensitive measures that explicitly elicit these mediating constructs on pain of underestimating the true influence of norms. Section 5 then applies this indirect pathways insight to the specific case of in-home tap water use, showing how a belief-sensitive survey blueprint can map the relevant mediators and, in turn, steer the design of norm-based nudges. Section 6 discusses the theoretical contribution of the argument, its implications for water policy, and the empirical work needed to test it, while Section 7 offers a brief conclusion.

2 Tap water as a belief-sensitive target for green nudging

Nudging has emerged as an effective tool for changing behavior (e.g., Mertens et al., 2021). By intervening on people's choice architecture—the environment in which decisions are made—policymakers can steer behavior in predictable directions while preserving their freedom of choice (Thaler and Sunstein, 2008, 2021; Congiu and Moscati, 2022). Given our psychological make-up, changing how options are presented can significantly improve people's choices (e.g., Marchiori et al., 2017; Schmidt, 2019; Schmidt and Engelen, 2020). Nudges complement rather than replace traditional instruments such as taxes, subsidies, and mandates, and they have been deployed across domains from politics to finance and health (Thaler and Sunstein, 2021; Halpern, 2019). By the mid-2010s, behavioral insights units and nudge-style programs had already been documented in well over half of UN member states (Whitehead et al., 2014).

Within this wider toolkit, green nudging—i.e. behavioral interventions aimed at promoting environmentally responsible choices—has proved particularly effective and cost-effective (Benartzi et al., 2017). Many governments and institutions are increasingly treating green nudges as key complements to conventional climate and resource policies (Beckenbach and Kahlenborn, 2016; Lehner et al., 2016; Schubert, 2017; Carlsson et al., 2019). Unlike pro-self nudges (e.g., retirement savings defaults), green nudges are considered pro-social, as their primary aim is to reduce negative externalities—such as air pollution or food and water waste—by facilitating sustainable everyday practices. This section situates the choice of tap water over bottled water within the psychological and nudging literature.

At home, choosing what to pour at the table is routine, low-salience, but symbolically loaded. Unlike many pro-environmental actions, this decision occurs multiple times a day within a ritualized household context, such as meals, where hospitality and purity cues matter. The literature on bottled-vs.-tap water consumption consistently points to four broad determinants of household water choice: perceived health and safety of tap water, taste and organoleptic expectations, trust in providers and public oversight, and presentation cues that shape expectations before consumption (Doria, 2006, 2010; Parag and Timmons Roberts, 2009; Wilk, 2006; Fedi et al., 2021). Perceived safety and health concerns are central in reviews of public drinking-water preferences (Doria, 2006, 2010). Trust in public providers likewise matters, especially where bottled-water use reflects distrust in municipal supply and its governance (Parag and Timmons Roberts, 2009). Taste expectations also play a major role, but are often shaped by symbolic framing and serving conditions rather than by objective quality alone; studies on bottled-water symbolism and controlled tasting contexts suggest that presentation can significantly affect evaluation (Wilk, 2006; Fedi et al., 2021). Doria (2006)'s review further cautions that risk and health beliefs alone are not sufficient predictors of bottled-water use, as practical considerations such as convenience also shape behavior. In the specific case of in-home consumption, however, convenience should not be understood as mere physical access, since tap water is already available at the point of use. Rather, what matters is perceived meal-time convenience: bottled water may be experienced as the more immediately ready, predictable, and socially low-risk option, especially when households associate it with stable taste, easy table presentation, and less need to reassure family members or guests about quality. By contrast, serving tap water may be perceived as requiring additional practical or symbolic steps, such as chilling it, pouring it into a carafe, filtering it, or making it look and feel appropriate for the dinner table (Doria, 2006; Wilk, 2006; Fedi et al., 2021). In this sense, contextual access in the household is not about literal distance from water, but about which option feels more frictionless and more readily acceptable at the moment of serving.

Trust links the public sphere to the kitchen table. In this regard, Parag and Timmons Roberts (2009) argue that bottled water growth in high-income contexts reflects a trust problem in public water. Indeed, as we noted in section 1, purity branding and private exit erode political pressure to invest in tap systems, which in turn undermines confidence in tap water—the result is a vicious, self-reinforcing feedback loop in which growing reliance on bottled water undermines confidence in tap, and progressively influences households' daily drinking decisions. To counter this cycle, the authors recommend transparent monitoring, credible third-party checks, and visible role-modeling (e.g. officials serving tap water) to rebuild trust. Wilk (2006) argues that bottled water is sold through symbolic associations (health, class, nature) that can eclipse tiny or nonexistent sensory differences; blind tastings often fail to support claimed superiority of bottled water. Empirically, recent Italian work with students finds exactly that pattern: stronger tap risk and bottled goodness beliefs among subjects who prefer bottled water, yet blind tastings often favor tap water when presentation is controlled (e.g., chilled) (Fedi et al., 2021).

The above features all make in-home tap adoption a belief-sensitive target behavior, well-suited to green nudging: small changes to presentation (carafe, chilling), salience (glasses set out), and social information (what similar others do or approve) can shift the intention–behavior link without coercion.

For present purposes, the crucial point is that, psychologically, the tap-over-bottled choice seems driven less by objective quality and more by expectations and normative signals. Reviews and cross-national evidence show that skepticism about tap water, beliefs about bottled water's purity and health, and taste expectations explain much of the variance in household preference; these factors can indeed be surprisingly decoupled from measured quality (e.g., Doria, 2006). The implication here is not that information is irrelevant, but rather that information, by itself, often lacks traction unless accompanied by norm cues and easy routines that make the desired behavior feel both approved and feasible at home. Campus and community studies—used here only as mechanism probes—show that pairing norm activation with belief work significantly outperforms message-only approaches (e.g., Fedi et al., 2021).

In light of the above, the most promising way to make tap water feel both approved and doable at home is not to oppose norm cues and environmental tweaks, but to combine them. In the present argument, physical or presentational modifications such as a chilled carafe, glasses set out, or brief quality framing are not treated as alternatives to norm nudges. Rather, they are understood as supporting devices that can either reinforce the credibility and social intelligibility of a norm cue or remove practical and symbolic frictions that would otherwise prevent normative influence from translating into routine behavior. With that clarification in place, a household strategy may combine norm cues with simple environmental supports such as:

• Defaults and salience: make tap the visible default (a filled fridge carafe; glasses on the table), so that the norm-congruent option is also the most immediately available one.
• Framing and brief transparency: elevate perceived quality (who tests, when) and shape taste expectations (serve chilled; carafe aesthetics), thereby making a pro-tap norm cue more credible and easier to accept.
• Hassle-factor reduction (self-efficacy): micro-scripts such as “fill the carafe nightly; add lemon if desired” help convert norm-consistent intention into routine action.
• Social norms: descriptive (“most households like yours serve tap at dinner”) and injunctive (“families here approve of avoiding plastic”) messages align what is normal and appropriate.

Social norms have proved effective in steering people toward environmentally responsible behaviors. As Farrow et al. (2017) demonstrate in their meta-analysis of 87 studies on pro-environmental behavior, social norms have had robust and consistent effects across various domains, including energy use, recycling, mobility, and water conservation. Notably, norm-based interventions have been shown to outperform purely informational campaigns and rival economic incentives in their capacity to shift behavior.

Although much norm research has targeted energy or waste, there is now field evidence bearing directly on drinking-water choices. A recent Italian restaurant study, for instance, made a simple, local descriptive norm salient: “Two in three people from this area drink tap water” (Dorigoni and Bonini, 2023). Bottled orders fell by 12 percentage points during the poster month (from 96% to 84% of water orders), with temperature controlled. What matters for our purposes is not the restaurant setting, but rather the strategy: making a localized, credible norm visible at the moment of choice – precisely the kind of cue a household can mirror (e.g., condo newsletters, utility mailers, or front-of-fridge prompts stating a verified local norm). At the same time, the psychological translation from restaurant to household should not be overstated. Ordering water in a restaurant is a relatively public act, performed in front of other diners and staff, whereas choosing what to drink at home is often more private and less exposed to immediate social observation. For this reason, a norm cue transplanted into the household is unlikely to work merely by reproducing public compliance pressure. Its success is more likely to depend on whether it reshapes internal expectations and evaluations—for example, what households think families like theirs do, approve of, and can comfortably serve at the table. This is precisely why the argument developed in the following sections focuses on indirect, belief-mediated pathways of normative influence in private settings. Moreover, integrative reviews emphasize that social norms are among the most effective levers across many pro-environmental domains (including water), provided messages are salient, credible, and context-matched—i.e., aligned with what similar others actually do or approve (e.g., Farrow et al., 2017). This in turn aligns with meta-analytic evidence that social-norm interventions can reliably shift conservation behavior, and with practical guidance that norm messages work best when they cite a nearby, meaningful reference group, and are shown at the moment of choice (e.g., Abrahamse and Steg, 2013; Bergquist et al., 2019). Taken together, the above considerations suggest that norm-based nudges (norm nudges) are especially well-suited to increasing in-home tap use. In the next section, we will therefore take a closer look at the psychology of social norms.

3 The standard model of norm nudging and its limits

Social norms are shared expectations about how people typically behave and how they ought to behave within a group or context. They are not formal rules but informal social constraints that regulate our behavior by shaping our perceptions of what is normal, acceptable, or valued. Following Miller and Prentice (2016), we treat norms as psychological constructs—i.e., not directly observable features of the social world, but rather beliefs about what others do and approve of (see also Schultz et al., 2007). It is customary to distinguish between descriptive norms, which refer to what most people actually do (e.g., “most people drink tap water”), and injunctive norms, which concern what most people approve or disapprove of (e.g., “most people think we should drink tap water”) (e.g., Cialdini et al., 1990, 1991; Farrow et al., 2017). These two types of norm are distinct, but can often interact in complex ways. Evidence about what people do, for instance, is frequently read as evidence about what they value, and vice versa.

Three additional clarifications will be relevant in what follows. First, we must separate perceived norms from the underlying social patterns to which they refer. Perceived norms are agents' beliefs about what relevant others do and approve. By contrast, what we call the “actual norm” is shorthand for the actual distribution, within the relevant population, of the behavior in question (descriptive side) and of approval, disapproval, or normative expectations concerning that behavior (injunctive side). This distinction is especially important for behavior change, because those perceptions are often biased (e.g., pluralistic ignorance), and the gap between perceived and actual local prevalence or approval is precisely where interventions can operate. A methodological complication should be noted, however. For relatively private behaviors such as household tap-water use, the actual descriptive norm is not easily established by survey evidence alone, since self-reports of behavior may themselves be affected by recall problems or social desirability bias. For this reason, estimates of the gap between perceived and actual norms are stronger when survey responses can be paired with more independent behavioral benchmarks, such as administrative indicators, observational proxies, or other non-self-report sources. In practice, then, the relevant contrast is often not between perception and a perfectly transparent behavioral baseline, but between perception and the best empirically grounded estimate available in a given context. Second, compliance with a norm is conditional on expectations: people are more likely to follow a norm when they believe that many others comply (empirical expectation) and that others expect them to comply (normative expectation). This conditionality explains the stability and fragility of norms: they persist when expectations align and can shift quickly when expectations move (Bicchieri, 2006). Finally, norms are usually distinguished from mere habits or conventions. Unlike habits, norms are socially constructed and sustained; and unlike conventions (e.g., driving on the right), they involve a moral or evaluative component. Social norms are not just behavioral patterns, but socially shared rules for behavior, rooted in collective expectations and sustained through mutual monitoring and reinforcement.

How can social norms be relied on in order to nudge people's behavior in a desired direction? According to the currently standard way of thinking about the issue, norm-based behavior change would take place via norm debiasing. Two broad assumptions about our psychological makeup are widely taken to underpin the effectiveness of this strategy. On the one hand, people often misperceive peer behavior or approval (e.g., “everyone drinks bottled water at dinner”). On the other, correcting those misperceptions can realign what people think others do or approve, which not only reframes the behavior but increases the likelihood they adopt it. This logic underwrites two widely used approaches, namely social norm marketing and personalized normative feedback. The former consists in broadcasting a clear, credible factual message about the actual norm to a whole target population via, e.g., posters or mailers. The latter provides instead household-level or person-level feedback which compares one's behavior to a relevant peer distribution (e.g., “you use more water than 60% of similar neighbors”).

A final qualification is needed, however. The standard norm-debiasing model works most straightforwardly when the corrected norm supports the target behavior. When the underlying descriptive norm is itself unfavorable—for example, when bottled-water reliance is genuinely widespread—broadcasting that norm may normalize the undesired behavior rather than reduce it. In such cases, intervention design cannot rely on a bare factual majority message alone. It must instead look for a more behaviorally promising referent, or combine descriptive information with injunctive or dynamic cues—for example, highlighting local approval of tap water, positive trends, or subgroup-specific pockets of adoption—so as to avoid reinforcing the very pattern the intervention seeks to change.

Two design points matter across both approaches. The first one concerns reference-group relevance. Messages work better when they point to, say, “people like me” in “places like mine” (e.g., neighbors, similar households, the same building), rather than to broad or distant audiences. The second concerns credibility and salience. The message must be both trusted (e.g., clear source, plausible numbers) and, crucially, encountered at the moment of choice (e.g., where the decision is actually made), or else even accurate norm information will easily be ignored. When norm information is well designed and credible, it can change what people do—not just what they think—by realigning their sense of what's normal and acceptable. In addition, in public-goods domains like energy and water, there's often a ripple effect: once people believe that conservation is common and approved, they are more likely to encourage it in others, praise compliance (or quietly disapprove of waste), and further normalize the pro-environmental choice. Those social reactions amplify the initial effect of the norm message.

Bicchieri's framework remains relevant in this domain, but the observability condition should not be interpreted too narrowly. Household tap-water use is less directly public than many conservation behaviors, yet it is not socially invisible in any strong sense. Drinking practices can become partially observable through ordinary domestic and neighborhood interactions: what households serve at the table when hosting others, what visitors or relatives encounter in the kitchen, and even visible proxy behaviors such as the regular purchase, delivery, storage, or disposal of bottled water. For this reason, norm-based influence in this domain need not depend on constant public monitoring or overt sanction. It may instead operate through a combination of socially legible cues, empirical and normative expectations, and the perceived appropriateness of serving tap water in semi-public household settings such as shared meals. In this respect, the dinner-table framing is not a theoretical necessity, but it is an especially relevant site because it increases the social legibility of a normally more private routine.

Norm-based interventions, however, are highly context-sensitive policy tools. A very instructive case in point comes from the use of norm nudges aimed at increasing indirect water conservation by encouraging hotel guests to reuse their towels, and hence reduce unnecessary laundering. Let us then consider two seminal studies in turn. Goldstein et al. (2008) tested bathroom cards that either appealed to environmental concern or highlighted a descriptive norm (“the majority of guests reuse their towels”). Across two field experiments in a US hotel, descriptive-norm signs outperformed the standard environmental appeal; moreover, a provincial norm referring to “guests in this room” performed best among descriptive variants. Bohner and Schlüter (2014) ran an extended replication of Goldstein et al. (2008) in two German hotels. Overall reuse was already high (Study 1 ≈ 82%), both environmental and descriptive messages beat a no-message baseline, but descriptive messages were not more effective than the standard environmental appeal—and the provincial (same room) variant was sometimes less effective than a general hotel-guest norm. Their discussion, and related European replications, underscore mixed evidence for a robust descriptive-norm advantage in this domain.

The general lesson is that the impact of a “Most people do X” message varies systematically with audience and setting. In particular, the contrast between the two towel-reuse studies—and the broader literature they sit within—highlights conditions that have a significant impact on whether a descriptive-norm message will work. The four factors can be summarized as follows:

• Baseline level and headroom. If the target behavior is already high, a descriptive norm may have little room to lift behavior further and could even risk boomerang toward the mean among very compliant guests. In Goldstein's setting, for instance, there was enough headroom for a majority message to move behavior; in Bohner and Schlüter's hotels, on the other hand, high baselines likely capped additional gains.
• Locality of the referent (provincial norms). People often treat very local norms (e.g. same room) as more diagnostic than broader referents. Goldstein found a “same-room” advantage; Bohner and Schlüter, on the contrary, observed this inconsistently, suggesting that hyper-local cues help only when perceived as credible and meaningful in that context.
• Message match and audience involvement. In more environmentally involved audiences, a well-crafted environmental rationale can rival or beat a bare descriptive cue—especially if the descriptive statistic seems unremarkable or thinly sourced. This arguably helps explain why a standard environmental appeal could keep pace with (or beat) a descriptive card in Bohner and Schlüter's German hotels.
• Perceived diagnosticity and credibility of the number. The level (e.g., 75%) and provenance (sample size, recency, source) of a norm claim tend to shape its influence. Whereas a number that is above guests' priors and clearly sourced can persuade, a number that seems poorly matched to the local context or methodologically weak can be discounted—thereby blunting the provincial advantage. Bohner and Schlüter, for instance, explicitly probed credibility in pilots and still found muted effects in situ.

Taken together, the above four factors suggest that the design of a norm-based intervention should be informed by questions such as: What is the actual baseline and headroom in the target population? Which referent feels genuinely local and credible? How involved is the audience? Will our percentage and source pass a plausibility check? Without these answers, designers risk broadcasting a norm that is inert, ambiguous, or counterproductive. The general lesson is that norm nudges must be firmly grounded in prior survey-based research. Surveys provide the knowledge base required to design credible, targeted messages: they reveal perceived descriptive and injunctive norms (and their gaps with actual norms), identify the salient reference groups, gauge baseline trust and credibility, and locate channel constraints that could block translation from changed perceptions to action. Put differently, the effectiveness of a norm message is conditional on what people already believe, who they take as “us,” and how the setting affords action. Absent such a map, designers risk choosing the wrong referent, signaling the wrong prevalence, or cuing a boomerang. Survey diagnostics – ideally paired with credible administrative data (e.g., metering, audits)—are thus not a luxury but a precondition for effective norm nudging.

If survey-first is the design imperative, however, what should those surveys measure? Behavior counts and intentions matter, but they are not enough. The central claim we elaborate on in section 4 is that norm influence is often of an indirect sort, operating through mediating belief-like constructs. It follows that focusing only on the direct path that links norms to behavior runs the risk of seriously underestimating the role of norms and obscuring exactly where a local design should push. Our key question then becomes: Which beliefs, in the target population, are the levers through which norms will move tap-over-bottled choices at home? The next section will set out this indirect pathways framework and show how a belief-sensitive survey instrument can surface the relevant mediators—so that, later, norm cues can be aligned with them rather than broadcast in the dark.

4 Indirect pathways of normative influence

As we saw in section 2 above, as a behavioral policy tool green nudging has shown great potential in fostering pro-environmental decisions. Most importantly, it holds promise for gently steering households toward choosing tap over bottled water. At the same time, however, nudge-based approaches to behavioral change face a serious replicability challenge, which reflects negatively on nudging's status as an evidence-based policy.

One important limit of many nudging studies is currently represented by an almost exclusive focus on the effects of various nudging interventions, and a concurrent lack or insufficient amount of effort devoted to the investigation of the specific cognitive mechanisms responsible for bringing about the observed effects. While the enthusiastic proponents of nudge-based approaches currently present their recommended policies as “evidence-based,” they typically do not do much, or at least not enough, in the way of investigating and explaining through which mechanisms such policies operate. As a consequence, current theoretical research on nudge-based policies has recently highlighted a need to engage more closely with the psychology of nudging or, as some have suggested, to “put the psychology back in nudging” (Marchiori et al., 2017; see also Berg and Gigerenzer, 2010; Grüne-Yanoff, 2016; Hansen, 2019; Marchionni and Reijula, 2019). As we saw in section 3, what one often observes is that a specific nudge—e.g., a norm message aimed at increasing indirect water conservation—which proved remarkably effective in one particular context, will prove largely ineffective (or even backfire) in other contexts (e.g., Sunstein, 2017; Osman et al., 2020). What this means in practical terms is that, as it has been put, when it comes to predicting the future effectiveness of individual nudges—and of behavioral policies in general—the journey from “it worked there” to “it will work here” is far from straightforward (Cartwright and Hardie, 2012).

The purpose of the present section is therefore not to reopen a general discussion of nudging, but to draw from that broader replicability problem a more specific lesson for norm-based intervention design. Our claim is methodological rather than empirical: if norm nudges are to be designed responsibly, they must be informed by evidence about the mediators and moderators through which normative influence operates in the target setting. The argument advanced here is theoretical and literature-based. It does not test those pathways directly, but clarifies why a belief-sensitive survey framework should be structured so as to detect them. The focus of the present argument is not to develop a general framework for behavioral public policy, but to clarify a more specific behavior-change problem: how norm-based interventions for household tap-water use should be designed in light of mediated normative influence.

A much-felt need in current research on nudging, then, is that of reaching a deeper understanding of the psychological processes that underlie a specific nudge, and can hence be credited for having made a difference to the behavioral variable of interest—i.e., having brought about the desired behavior change. The point here is not to reintroduce social norms in general, but to shift from the standard descriptive account discussed in Section 3 to a more specific question: through which psychological pathways does normative influence operate, and what follows from this for the design of belief-sensitive surveys? Now, that same need seems to be reflected in the literature on social norms. On the one hand, it has long been known that social norms have a strong impact on human behavior—i.e., people's decisions and choices are heavily influenced by social information about what relevant others do or approve of, as well as about what the common behaviors in a given situation are. In their seminal contribution to the study of social influence, for instance, Deutsch and Gerard (1955) refer to this phenomenon as normative social influence. On the other hand, however, it is far from unusual to find psychologists lamenting that the process of normative influence is still not well understood (e.g., Göckeritz et al., 2010; Thøgersen, 2014).

The task that currently awaits nudge theorists could hence be summarized in a slogan: we urgently need to open nudges' black box and take as good a peek into it as we possibly can. What does this mean in practice? What kind of evidence should we be looking for? According to (Marchionni and Reijula 2019), behavioral public policy is premised on the idea that behavioral interventions should be grounded on a psychologically realistic picture of human behavior and its causes. As a consequence, here is the answer they provide to our question:

“Extrapolating behavioral policies requires evidence about the invariance of the causal relationship between the policy lever and the outcome. Knowledge of whether and when a given causal relationship remains stable typically … relies on evidence about the variables that mediate and those that modulate it, that is, on mechanistic evidence.” Marchionni and Reijula, (2019: 62)

By their lights, then, the kind of evidence we need to open nudge's black box should be looked for in empirical research aimed at unearthing the mediators and moderators of nudges' effects. How do matters stand on this front with respect to norm nudges?

In this regard, two distinct but complementary points from the literature should be kept apart. First, Nolan et al. (2008) show that normative social influence is often underdetected: people can be influenced by descriptive norms while failing to recognize that influence in their own causal self-reports. This finding points to a limit of introspection-based evidence and cautions against treating respondents' explicit accounts of what motivates them as transparent evidence of mechanism.

Second, John Thøgersen's work highlights a different, though related, problem in much survey-based research on pro-environmental behavior. When perceived social norms are modeled as one antecedent among others, and attention is restricted primarily to their direct effects after controlling for other proximal predictors, their total influence is likely to be underestimated (Thøgersen, 2014; see also Ajzen, 1991; Armitage and Conner, 2001; Bamberg and Möser, 2007). The reason is not simply that respondents fail to detect social influence, but also that normative influence may operate indirectly through mediating constructs such as personal norms, attitudes toward the behavior, and perceived behavioral control.

This distinction is important for the present argument. Our claim is not that a more elaborate survey instrument can, by itself, identify the true causal mechanism of normative influence, nor that self-reports can fully overcome the introspection illusion documented by Nolan et al. Rather, our more limited claim is that intervention design is likely to be poorer when these indirect pathways are not measured at all. Belief-sensitive survey research should therefore be understood primarily as a diagnostic tool for mapping the local configuration of plausible mediators, moderators, and misperceptions through which normative influence may operate.

For this reason, the survey blueprint proposed in this paper is not intended as a substitute for experimental or quasi-experimental work. Where possible, survey-based diagnostics should be complemented by mediation analysis and more strongly causal designs. Still, if designers rely only on self-reported behavior or on models that estimate only the direct contribution of perceived norms, they risk overlooking important channels through which norms shape intentions and behavior. The practical implication is modest but important: surveys cannot settle causality on their own, but they can still improve intervention design by revealing which belief patterns and psychological pathways are most plausible in a given local context.

This indirect-pathways account should not be read as exhaustive for the specific case of household tap-water use. In that domain, other highly relevant cognitions also matter, including perceived safety, taste expectations, and perceived cost or convenience. WASH-oriented frameworks such as RANAS (Risks, Attitudes, Norms, Abilities, and Self-regulation) explicitly foreground these factors (Mosler, 2012). In the present argument, however, they are not treated as competing alternatives to the indirect-pathways account, but rather as behavior-specific belief contents that can shape attitudes toward tap water and, in some cases, perceived behavioral control. For this reason, the survey blueprint developed in the next section is designed to elicit these locally salient beliefs rather than assuming that one mediator alone captures the relevant psychological structure.

At the same time, the present argument should not be read as denying the importance of moderators. As Marchionni and Reijula (2019) emphasize, evidence relevant to behavioral policy extrapolation concerns not only mediators but also modulators of a causal relationship. In the specific case of household tap-water use, contextual factors such as local water quality, plumbing conditions, trust in public supply, and entrenched taste expectations may function as moderators or boundary conditions of normative influence (e.g., Doria, 2006, 2010; Parag and Timmons Roberts, 2009; Fielding et al., 2012). Even a well-designed norm message may therefore fail if these background conditions are unfavorable. For this reason, the survey-first logic defended here is meant not only to identify plausible mediating pathways, but also to detect the local conditions under which those pathways are more or less likely to operate.

This brings us to the next question: if norm effects are often indirect, and if surveys are to be used diagnostically rather than as stand-alone causal tests, which beliefs should they target in the specific case of household tap-water use?

5 A belief-sensitive survey blueprint for household tap-water use

This section will translate the indirect pathways insight into a belief-sensitive survey rationale for the in-home tap-over-bottled decision. The guiding premise is that social norms frequently affect behavior indirectly—that is, by affecting personal norms (felt moral obligation), attitudes toward the behavior, and self-efficacy—and that, without measuring these mediators, designers risk misdiagnosing why a norm message succeeds or fails in a given locale. On this view, prior survey research is not an optional add-on but the first step of responsible norm-nudge design for household drinking water. A norm message that is psychologically elegant in the abstract may still be inert—or worse, counterproductive—if it is launched into a belief landscape it does not match.

The case for surveying first rests on the specific psychology of this behavior. What households pour at the table is belief-sensitive: choices correlate less with objective quality and more with expectations (taste, purity, hospitality), normative signals (“what families like ours do/approve”), and simple feasibility scripts that make tap routine (a chilled carafe, glasses set out). If normative information is to carry weight in this domain, we must first map the belief environment in which it will operate: what people believe, how those beliefs are distributed across the target population, and which of them mediate intentions to serve tap at home. Absent that information, norm cues are broadcast in the dark, amplifying the context-sensitivity problem.

Concretely, the measurement target is a triad of belief categories that corresponds to the belief architecture of Fishbein and Ajzen's Reasoned Action Approach (RAA): behavioral beliefs (about outcomes and their desirability), normative beliefs (about what salient others do and approve), and control beliefs (about barriers, supports, and one's capability to act). Although several applications to pro-environmental behavior are presented by their authors under the label “Theory of Planned Behavior,” we use “Reasoned Action Approach” throughout, given this paper's focus on the belief structure underlying attitudes, perceived norms, and perceived behavioral control. This terminological choice draws on Fishbein and Ajzen (2010) and Ajzen and Kruglanski (2019), while also acknowledging the earlier Theory of Planned Behavior (Ajzen, 1991, 2020) and its applications to pro-environmental behavior (e.g., Klöckner, 2013; de Leeuw et al., 2015; Fedi et al., 2021; Benincà et al., 2022; Canova et al., 2023, 2024; Benincà et al., 2025). We do not deny the relevance of WASH-specific frameworks, including RANAS, which are often especially useful for identifying a broad range of determinants in water-related behavior. However, the present paper adopts RAA as its primary organizing framework because its central concern is not to propose a general model of water-related behavior, but to specify how a belief-sensitive survey should be structured when the aim is to understand how normative influence may operate indirectly through attitudes, perceived norms, and perceived behavioral control. On this view, factors emphasized by WASH-oriented frameworks—such as risk perception, taste expectations, and perceived convenience—are not excluded, but are treated as behavior-specific belief contents to be elicited within a belief-based measurement logic. The point is not to rehearse a general theory, but to use RAA's measurement architecture to pin down behavior-specific beliefs for this case—serving tap at home—because design leverage lives at the belief level.

Against that behavioral anchor, the first category to elicit is behavioral beliefs for attitudes. What matters here is (a) respondents' perceived likelihood that certain consequences will follow from serving tap at dinner, and (b) how desirable or undesirable those consequences are. Expectancy-value composites (likelihood × evaluation) then identify which beliefs actually drive attitudes. Content must be domain-specific: taste/organoleptics as served (e.g., chilled in a carafe vs. lukewarm from the tap), purity/safety backed by quality information (who tests the water and how often), convenience/hassle (ease of setting the table; no need to buy, carry, and store bottles), cost savings, and perceived social signaling at the table (hospitality, modernity, thriftiness). The month-long horizon ties beliefs to an intention window that invites realistic appraisal. What often emerges at this stage is that small presentation cues—carafe aesthetics, fridge placement, the routine of chilling—are already represented in people's taste expectations and hospitality concerns; the survey's task is to quantify their weight.

The second category is normative beliefs for perceived norms. The aim is to localize the referent and to separate descriptive from injunctive content. Rather than asking whether “tap is popular,” the survey should ask what relevant groups actually do and approve. Examples include: “In the past 2 weeks, how many of your close friends/neighbors in this building or on your block you think served tap at dinner at home?” (descriptive), and “How much do these groups approve of serving tap at dinner at home?” (injunctive). A supplementary “motivation-to-comply” item (e.g., “How much do you care what your friends/neighbors think about this issue?”) may still be useful, but it should not be treated as a standalone test of whether normative influence matters, since respondents may underdetect or misreport the role of social influence in their own decision-making. For that reason, evidence about the relevance of norms should be inferred from a broader pattern of indicators, including perceived descriptive and injunctive norms, misperception gaps, and the salience of specific local referents. Including several proximal referents allows the data to reveal which “local” truly feels local to respondents—an issue that Section 3 showed is pivotal for avoiding backfire. Where feasible, these perceived-norm items should be paired with benchmarks derived from administrative indicators or small observational samples to estimate misperception gaps (pluralistic ignorance or false consensus). Those gaps, when present, are the leverage points for norm-debiasing messaging.

The third category concerns control beliefs and perceived behavioral control, including self-efficacy more narrowly understood as confidence in one's ability to perform the target behavior across relevant situations. Many households express pro-tap intentions yet fail to turn intention into routine. For that reason, a belief-sensitive survey should elicit both (a) direct efficacy beliefs and (b) perceived barriers and supports. More direct self-efficacy items might include formulations such as: “I am confident that I can serve tap water at dinner on most nights over the next four weeks,” “I could do so even if other household members initially preferred bottled water,” or “I could do so even when hosting guests.” These can then be complemented by control-belief items targeting specific obstacles and facilitators, such as: “We have space for a carafe,” “I can chill a carafe each evening,” “Our tap tastes fine when chilled,” “I know who tests our tap and when,” or “I can set glasses and carafe out before dinner without extra hassle.” Because perceived control is often socially shaped, it also helps to include items probing vicarious experience, such as: “People like us manage this easily” or “I have seen friends serve tap without issues.” Taken together, these items distinguish confidence from contextual constraints and help reveal where simple scripts or environmental supports would be most useful.

The item wordings offered in this section should be understood as illustrative examples rather than as already validated psychometric scales. Their purpose is to show the type of belief-sensitive content that a survey should elicit in the specific case of household tap-water use. In line with the logic of the Reasoned Action Approach, the exact wording of such items should be refined through pilot elicitation in the target population and, before full-scale deployment, subjected to context-specific pretesting and psychometric validation. The point of the present blueprint is therefore not to provide a finalized measurement instrument, but to specify the categories of beliefs that such an instrument should be designed to capture. In this respect, the proposed measurement strategy is bottom-up in spirit: rather than imposing a fixed list of determinants in advance, it begins by eliciting locally salient beliefs from the target population and only then structures the survey instrument within the broader RAA architecture.

To complete the picture, and beyond the elicitation of perceptions and related beliefs, the survey should also measure intentions and baseline behavior over a short horizon, using time-bounded, behavior-specific items (“Over the next four weeks, how likely are you to serve tap at dinner on most nights?;” “In the last two weeks, on how many dinners was tap served?”). These measures serve as proximal outcomes for path or mediation modeling and as indicators of headroom: when baseline is already high, a descriptive norm may have limited scope and greater boomerang risk. Finally, a set of segmentation variables helps identify moderators of belief patterns: presence and age of children; housing type (apartment vs house; shared kitchen constraints); district within the municipality; prior unpleasant taste episodes; plumbing age; and general environmental involvement. These variables should not be understood as a closed or universally validated set of predictors for household tap-water use. Rather, they are proposed as context-sensitive segmentation candidates, motivated by prior research on household water behavior and conservation, public perceptions of drinking-water quality, and the role of trust, prior experience, and infrastructure in shaping water-related practices (e.g., Doria, 2006, 2010; Parag and Timmons Roberts, 2009; Fielding et al., 2012). Their exact inclusion should therefore be refined through pilot elicitation and local pretesting. In the present blueprint, intention to serve tap water at dinner is treated as the main proximal outcome for survey-based modeling, whereas self-reported behavior is used primarily to estimate baseline/headroom and, where available, as a downstream outcome. Likewise, “segments” should not be understood as purely demographic categories fixed in advance, but as empirically identifiable subgroups defined by recurring combinations of belief profiles, baseline patterns, and relevant moderators within the target population. The segment examples discussed later in this section are therefore stylized illustrations of the kinds of profiles that a belief-sensitive survey could help distinguish.

Why insist on this belief-centered elicitation? Because it mirrors the structure that Section 4 argued is essential for interpreting norm effects: attitudes formed from behavioral beliefs, perceived norms from normative beliefs, and self-efficacy from control beliefs. The deepest, most actionable explanations sit at the belief level: it is there that one discovers not only whether taste expectations matter, but precisely which expectations (e.g., “warm tap tastes flat,” “chilled tap tastes clean,” “a carafe looks ungenerous”) and how strongly they shape intentions relative to normative approval or ease of routine. Our aim is indeed to move from the general claim—norms often work indirectly—to a local diagnosis of which pathways are carrying normative influence in this population now.

A helpful methodological template is provided by recent Italian work on the intention to use refillable bottles outside the home with a two-stage design (Benincà et al., 2025). The researchers first ran a pilot to elicit salient beliefs in the local context, then fielded a large survey structured on TPB constructs and extended with descriptive and moral norms. They measured behavioral, normative, and control beliefs, linked them to attitudes, subjective norm, and perceived control, and examined which factors best predicted intention. For our purposes, the value of that approach lies not in the specific target, but in the logic: pilot elicitation to surface local content, belief batteries to quantify levers, and modeling to identify what carries intention. That same logic can be adapted to the in-home dinner setting. For ease of reference, Table 1 summarizes the main components of the proposed belief-sensitive survey blueprint, linking the three belief categories to illustrative survey content, the corresponding psychological mediators, and the design implications discussed in this section.

Table 1. Summary of the proposed belief-sensitive survey blueprint for household tap-water use.

Analytically, the proposed survey blueprint calls for techniques that are appropriate to mediated relationships rather than simple descriptive ranking alone. Because the present argument concerns indirect pathways, the relevant analyses would ideally include path analysis or Structural Equation Modeling (SEM), so as to estimate how behavioral, normative, and control beliefs shape attitudes, perceived norms, and perceived behavioral control, and how these in turn relate to intentions and, where available, behavior. In applications that do not support fuller path analysis or Structural Equation Modeling (SEM), researchers may instead rely on simpler regression-based mediation analyses as a preliminary strategy for examining indirect relationships. In line with the methodological caution already noted in Section 4, however, such analyses should be interpreted as evidence about theorized indirect associations rather than as conclusive proof of causality, unless they are complemented by longitudinal, experimental, or quasi-experimental designs.

Beyond mediation modeling, empirical implementation would also require some form of segmentation analysis. In smaller applications, this may take the form of theory-guided subgroup comparisons; in larger datasets, it could be supported by cluster-analytic or profile-based procedures that identify recurring combinations of beliefs, baseline patterns, and moderators. On this basis, labels such as high, weak, or adequate should be interpreted comparatively, relative to the local distribution of the relevant measures, rather than as fixed thresholds assumed in advance. The stylized scenarios discussed later in this section are therefore meant as illustrative summaries of the kinds of empirically grounded profiles that such analyses could yield.

At the same time, any one-wave survey implementation of this blueprint remains vulnerable to reverse causality. Current intentions or established habits may shape reported attitudes, perceived norms, and perceived behavioral control, rather than the other way around; households that already serve tap water, for instance, may retrospectively report more favorable beliefs or perceive pro-tap norms as more salient. For this reason, the present blueprint should be understood primarily as a tool for identifying plausible pathways and design-relevant associations, not for settling temporal direction on the basis of cross-sectional data alone. Stronger empirical implementations should therefore, where feasible, incorporate longitudinal or multi-wave designs, so that baseline beliefs can be related to later intentions and behavior, or linked to changes following a norm-based intervention.

In practical terms, a belief-sensitive survey yields three results for nudge design. First, it shows which beliefs matter most for intention and how they are distributed across segments. Analyses identify the behavioral, normative, and control beliefs with the greatest predictive weight and show their prevalence (for instance, taste pessimism in older buildings or low self-efficacy in families reporting mealtime pushback from teenagers). This helps identify which beliefs to target first: widespread, high-impact beliefs become priorities, while rare or low-impact ones move down the list. Second, by pairing perceived-norm items with benchmarks, the survey locates misperception gaps: many households may quietly serve tap while assuming almost no one does, or may underestimate neighbors' approval of tap. These gaps show where norm-debiasing messages are warranted and which referent (building, block, school network) is local enough to carry the claim. Third, the instrument distinguishes logistical from psychological barriers: if respondents rate “carafe chilling” as easy but “family acceptance” as hard, the design brief is not more information about mineral content, but legitimation and a hospitality-affirming script. Conversely, if taste is the problem only when tap is served warm, the brief is to normalize chilled tap and presentation rituals, not to litigate safety.

With those diagnostics in hand, the path from survey to design becomes tractable. Four stylized scenarios, each reflecting a typical mediator profile, will illustrate how the diagnosis steers the intervention:

If personal norms are comparatively high (a felt duty to avoid plastic), attitudes favorable, but self-efficacy low, households need help in converting intention into routine. Here a credible local descriptive norm (“most households on your block served tap at dinner last week”) can be paired with micro-scripts that reduce friction—e.g., “fill the carafe nightly,” “keep it on the middle fridge shelf,” “add lemon if desired”—alongside simple commitment prompts (e.g. “tap on weeknights”). A light injunctive overlay (e.g., “families here approve avoiding plastic at dinner”) legitimates the practice without moralizing.

If attitudes and self-efficacy are comparatively supportive but personal norms comparatively weak, the behavior is seen as optional rather than a “should.” The design shifts: combine a descriptive message with a gentle injunctive cue tied to shared local values (e.g., waste reduction; children's future), and ensure the referent is genuinely local to avoid the perception of generic moralizing. Visible role-modeling (e.g., parent-teacher newsletters showcasing “tap nights”) can transform latent acceptance into a felt obligation.

If self-efficacy is comparatively strong but attitudes are comparatively weak—taste skepticism or diffuse safety concerns—the priority is attitude repair. Brief transparency (who tests, how often) and presentation framing (“serve chilled;” “use an attractive carafe;” “place glasses on the table”) set taste expectations before the first sip; a localized descriptive norm then normalizes the practice. Where skepticism is entrenched, it is often useful to lead with injunctive cues from trusted referents (neighbors, building association, pediatricians) before layering descriptive claims.

Finally, if the survey reveals a misperception-driven barrier—households already serve tap privately but believe that “no one else does” or that others disapprove—a pure norm-debiasing message is indicated. Communicate the actual local majority, with source and scope—e.g., “In your building, seven in ten households served tap at dinner last week (source: concierge waste-audit; sample of N = …).” Because the baseline is already high, the intervention should be modest and placed where it will be encountered at the decision point (front-of-fridge; condo newsletter before dinner hours), then scaled back or periodically refreshed once alignment is achieved, so as to balance maintenance against message fatigue.

These scenarios are not mutually exclusive. The same neighborhood may exhibit segments in different quadrants: a cluster with high duty and low efficacy; another with strong efficacy but entrenched taste hostility; a third with quiet adopters who underestimate their peers. That is precisely why the survey's segmentation is not a cosmetic add-on but the means by which interventions are matched to belief profiles, minimizing the risk of a one-size-fits-all broadcast that some audiences discount and others read as tone-deaf.

A final clarification concerns persistence over time. The present argument does not assume that belief-sensitive norm nudges automatically generate permanent, self-sustaining change once the original cues are removed. In some cases, repeated pairing of norm messages with supportive routines may help stabilize new household habits and reduce reliance on external prompts. But where bottled-water use is reinforced by entrenched taste expectations, distrust, or longstanding domestic routines, some degree of maintenance is likely to remain necessary. The practical implication is that utilities should think in terms of adaptive reinforcement rather than one-off messaging: cues may sometimes be reduced in intensity, rotated, or periodically refreshed, but without assuming that effects will persist unchanged once prompts disappear.

Two closing considerations reinforce the ‘surveys first' stance—one practical, the other conceptual. Practically, norm interventions are fragile to context. Section 3′s towel-reuse contrast showed that identical descriptive cues can lift behavior in one site and stall it in another because baselines, referents, and credibility differ. A belief-sensitive survey surfaces those parameters before rollout, reducing the odds of inert or counterproductive messages. Conceptually, belief-sensitive surveys honor the rationale for building on behavioral capital in policy design: behavior change should be cultivated where it is cheapest and most effective, which requires knowing which beliefs carry behavior here and now. Because norm influence is often mediated, a design that measures only direct norm → behavior links will systematically underestimate normative leverage. What looks like a small norm effect can mask a large indirect effect operating through personal norms, attitudes, or self-efficacy; without measuring those pathways, effort may be misallocated, for example by investing in more information when a simple efficacy script or injunctive cue would do more work.

For decision-makers, the most useful survey deliverable is therefore compact and action-oriented. It should list, in order of predictive importance, the behavioral, normative, and control beliefs that shape intention to serve tap at dinner; report the prevalence of those beliefs by segment; quantify misperception gaps (perceived vs actual local norms); and offer a one-page design memo mapping belief patterns to message type (descriptive vs injunctive blend), referent (building/block/municipality), and supporting scripts (what to put in the fridge; where to place the cue). Because our focus is the in-home decision, we have treated workplace/campus/restaurant studies as mechanism probes rather than templates; even so, their common lesson is fully portable to the household: norm cues work when they are localized and visible at the decision point, and belief work—repairing attitudes, boosting efficacy—often amplifies them beyond information alone. A field intervention in restaurants, as we have seen, reduced bottled orders by making a local descriptive norm salient with a credible source; the household translation is not “turn the kitchen into a restaurant,” but rather to mirror the mechanism (a locally credible norm encountered at the moment of choice) through building newsletters, utility communications, or front-of-fridge prompts.

One final qualification concerns cost. Green nudges are often described as low-cost interventions, but the survey-first strategy defended here is not costless. Pilot elicitation, segmented survey work, and administrative benchmarking all require organizational capacity and resources, especially when implemented at scale. The relevant comparison, however, is not between a zero-cost diagnostic phase and a fully specified intervention, but between more and less informed intervention design. In many cases, an upfront diagnostic investment may be justified if it helps avoid poorly targeted campaigns and directs norm messages toward the segments where headroom and leverage are greatest. At the same time, the blueprint need not always be implemented in its most resource-intensive form: in lower-capacity settings, a lighter version may rely on shorter elicitation stages, smaller stratified samples, or the reuse of existing customer-survey and administrative data. The practical claim of the present paper should therefore be understood in qualified terms: the nudges themselves may be low-cost, but responsible design still requires a diagnostic investment whose scale should be matched to the policy context.

A related practical point concerns implementation at scale. The present blueprint does not imply that water authorities must design fully individualized nudges for each household. A more realistic operational strategy is to work with a limited number of empirically identified segments and to align message type, referent, and supporting scripts with those segments through channels that utilities already use, such as billing inserts, customer newsletters, neighborhood pilots, school-based communication, condominium networks, or app and website prompts. In this way, belief-sensitive design can remain operationally tractable: the aim is not one-to-one personalization, but better matching between local belief profiles and a manageable set of intervention packages. A water authority might, for example, deploy one package for households with strong pro-environmental norms but low perceived behavioral control, another for segments marked by taste skepticism and low trust, and a third for areas where the main barrier is misperceived local prevalence. This kind of segmented deployment is more demanding than a one-size-fits-all campaign, but it remains compatible with routine utility communication infrastructure.

In sum, if the paper's central claim is that norm-nudging tap water use must be grounded in belief-sensitive survey work, the present section has specified what to measure, why those measures are the right ones, and how to convert survey outputs into matched designs. The broader implication is straightforward: because norm effects in this domain are often indirect, effective policy follows a two-step sequence—diagnose beliefs first, then deploy norm nudges that those beliefs can carry.

6 Discussion

6.1 Theoretical contribution

This article has advanced a theoretical claim about the design of norm-based interventions aimed at increasing household tap-water use. Its central argument is not merely that such interventions should be grounded in prior survey research, a point that is already widely accepted in the nudging literature. Rather, the more specific claim is that the knowledge base produced by such surveys should be belief-sensitive. The reason is that social norms often influence environmentally relevant behavior indirectly, through mediating constructs such as personal norms, attitudes toward the behavior, and self-efficacy. If those mediators are not measured, the role of norms is likely to be underestimated and the design of interventions correspondingly misdirected.

The contribution of the present paper should therefore be understood in delimited terms. It does not propose a new general theory of pro-environmental behavior, nor does it deny the explanatory value of existing approaches such as the Reasoned Action Approach, extended TPB models, expectation-based accounts of norms, or WASH-oriented frameworks such as RANAS. Rather, its contribution is integrative and application-specific. It brings together research on mediated norm effects in pro-environmental behavior (e.g., Thøgersen, 2014; Helferich et al., 2023), the belief architecture of the Reasoned Action Approach (Fishbein and Ajzen, 2010; Ajzen and Kruglanski, 2019), and the literature on bottled- vs-tap water choice and tap-water interventions (e.g., Doria, 2006, 2010; Parag and Timmons Roberts, 2009; van der Linden, 2015; Kasdan et al., 2020; Dorigoni and Bonini, 2023). What is new, in this limited sense, is the claim that these strands jointly support a belief-sensitive survey blueprint for designing norm-based interventions for household tap-water use around indirect pathways rather than direct norm-behavior links alone. The stylized scenarios developed in Section 5 should accordingly be read not as empirically estimated findings from the present paper, but as literature-guided illustrations of the kinds of locally grounded intervention profiles that such a blueprint could help generate.

6.2 Implications for water policy and intervention design

This indirect-pathways argument has immediate implications for the design of household water interventions. In the case of tap-water use, the relevant behavioral target is highly context-sensitive and strongly shaped by expectations about taste, purity, hospitality, convenience, and social approval. It follows that interventions cannot be designed effectively on the basis of self-reported behavior alone. Instead, policy design should begin with the identification of the specific behavioral, normative, and control beliefs that organize the local belief landscape. Only then can descriptive and injunctive norm messages be tailored so that they are credible, psychologically local, and capable of operating through the mediators that actually matter in a given population.

Such context-sensitive tailoring consists in matching the content and delivery of a norm-based intervention to the local belief landscape identified through prior diagnostic work. This matching is important because the same norm message may be ineffective or even counterproductive when its reference group, prevalence claim, communication channel, or supporting cues do not fit local baselines, misperceptions, trust levels, taste expectations, or infrastructural constraints. In practice, the process is sequential: first diagnose these local conditions; then select the most appropriate descriptive or injunctive message, referent, source, communication channel, and supporting scripts; and finally field-test the resulting intervention package before wider deployment.

6.3 Limitations and future directions

The argument developed here is theoretical rather than empirical. It does not report original survey data on household tap-water use, nor does it test the proposed framework experimentally. Its aim is instead to offer a new interpretation of how norm nudges should be understood in this domain and to derive from that interpretation a practical blueprint for survey design. For this reason, the next step is empirical. The empirical blueprint sketched in this paper also has important methodological and practical limitations that should be acknowledged explicitly. Even when implemented carefully, a predominantly survey-based design can identify plausible pathways and design-relevant associations, but it cannot by itself settle causal direction or fully overcome the limits of introspection-based self-reports. In particular, one-wave implementations remain vulnerable to reverse causality, and estimates of misperception gaps are stronger when survey data can be paired with independent behavioral benchmarks. In addition, the full blueprint may be demanding to implement at scale: pilot elicitation, segmentation, benchmarking, and tailored communication strategies require organizational capacity that not all utilities will possess to the same degree. For this reason, future empirical research should assess not only whether belief-sensitive designs outperform more behavior-centered or information-only approaches, but also which lighter and more scalable versions of the blueprint retain sufficient diagnostic and practical value in real policy settings. Future research should implement belief-sensitive survey instruments in specific local contexts, estimate which beliefs have the greatest predictive weight, identify misperception gaps between perceived and actual norms, and then test whether interventions designed on that basis outperform more behavior-centered or information-only approaches.

The scope of this argument should also be kept explicit. The paper is primarily concerned with contexts—especially high-income settings—in which tap water is objectively safe and broadly available, but bottled-water reliance remains common for social, perceptual, or routine-based reasons. In settings where access and safety are genuine structural constraints, the explanatory balance between beliefs, norms, and material conditions may differ substantially, and the present blueprint would require adaptation. Future empirical research should therefore test the framework across different infrastructural contexts while keeping the distinction between intention and observed behavior explicit.

7 Conclusion

The broader implication is straightforward. In the domain of household tap-water use, effective norm nudging should proceed in two steps: first diagnose the belief landscape, then deploy norm-based interventions that those beliefs can carry. On this view, what matters most is not simply whether social norms influence behavior, but through which belief-mediated pathways they do so. Clarifying those pathways is both a theoretical task for the psychology of normative influence and a practical requirement for the design of more effective water policy.

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.

Footnotes

Edited by: Saket Pande, Delft University of Technology, Netherlands

Reviewed by: Mithun Raj, Amrita Vishwa Vidyapeetham University, India

Diana Carolina Callejas Moncaleano, Ministry of Environment and Sustainability of Colombia, Colombia