Introduction

In 1962, former US president J.F. Kennedy formulated the Consumer Bill of Rights. The bill introduced the right to choose, defined as the right “to be assured, wherever possible, access to a variety of products and services at competitive prices”. The principle of consumer autonomy has been built on this right to choose and remains a foundational principle of liberal democracies today. In this context, Thaler and Sunstein’s (2008) “Nudge,” defined as “any aspect of the choice architecture that alters people’s behavior in a predictable way without forbidding any options or significantly changing their economic incentives,” emerged with the powerful promise to design interventions that respect consumer autonomy but manipulate the status quo of decision making in order to shift behavior.

Since the original definition, the concept of nudging has been further refined. Hansen (2016) extends the definition by emphasizing that nudges are intentional attempts to influence behavior in predictable ways by leveraging cognitive boundaries, biases, routines, and habits that often hinder rational decision-making. He argues that nudges exploit these inherent characteristics independently of forbidding or adding rationally relevant choices, changing incentives, or providing factual information and rational argumentation.

While nudging has become influential in recent years, it is not without controversy. Indeed, the extent to which nudges truly preserve consumer autonomy has become a key point of most ethical discussions in the scientific literature. A recent systematic review stated that 86% of ethical contributions to the nudging topic address autonomy (Kuyer and Gordijn 2023). The concept of autonomy in decision-making encompasses at least two integral aspects: freedom of choice and agency (Vugts et al. 2020). These facets, while distinct, are intricately connected, collectively contributing to the overall understanding of consumer autonomy.

Freedom of Choice entails the absence of restrictions on available options. In this context, it is important to consider whether, and to what extent nudges provide individuals with a genuine opportunity to resist the influence of a nudge (Kuyer and Gordijn 2023; Saghai 2013). This discussion sometimes pits effectiveness against autonomy, arguing that nudges can be either highly effective or easily resistible (Floridi 2016; Mills, 2018). However, this dichotomy oversimplifies the issue, as there are instances where nudges can be easily resisted and still prove effective, leading to the question of how to navigate this trade-off. This scoping review provides further empirical support of resistible, yet effective nudges.

Agency refers to an individual’s capacity to deliberate, critically reflect, and make choices (Vugts et al. 2020). Dold and Lewis (2023) further illuminate this distinction between these two aspects of autonomy by introducing the concepts of “opportunity freedom” (availability of choices) and “process freedom” (capacity to make reasoned decisions). While opportunities alone do not necessarily make one feel in control of their life, process freedom allows for control over the choice process and fosters the sense of being the “author of one’s life”. This is not necessarily the same as a rational decision outcome (Engelen 2019). Outcome rationality pertains to evaluating the most rational decision irrespective of the decision-making process, while process rationality seeks to understand the feasibility of a rational reflective process in the context of a given decision. The latter is what agency-respecting choice architects can strive to do. Nudges that undermine a decision-maker’s ability to reason threaten the agency dimension of autonomy (Vugts et al. 2020).

In summary, autonomy requires not only having options, but also the internal capacity to reflect on those options and freedom to act on them to achieve personal goals (Kuyer and Gordijn, 2023). Increasing agency through nudging can enhance desirable outcomes. For example, a study on charitable giving found that offering a list of donation options along with a default amount resulted in higher overall donations compared to providing just a single default option (Banerjee, John, et al. 2023). As Sunstein (2015) has long advocated, effective nudge design can preserve both agency and freedom of choice, ensuring that the success of the nudge, such as increased donation amounts, does not come at the expense of individual agency.

There remains a substantial gray area with regard to determining when a nudge does or does not, as well as to what extent, preserve autonomy. “Perceived intrusiveness” has emerged as a key construct to investigate the concept in survey research on nudge approval, with researchers regularly prompting consumers to indicate the extent to which they believe a nudge intrudes upon their capacity to choose - essentially their agency in the choice (Evers et al. 2018; Hagman et al. 2015, 2022; Yi et al. 2022). Up to 30% of the differences in approval of nudges is estimated to be explained by perceived intrusiveness alone, making it a very important concept to predict acceptance (Evers et al. 2018). However, “perceived intrusiveness” is not an ideal concept to judge the preservation of autonomy. For instance, when it comes to certain nudges, the level of controversy often stems from the fact that people tend to either strongly favor or strongly oppose their perceived intrusiveness, leading to polarized opinions among many individuals (Lemken et al. 2023). A judgment based on mean values of opinions would ignore a substantial group of citizens who voice concern or support with respect to autonomy. Moreover, it is worth noting that citizens may blend their views regarding the legitimacy of the nudge’s objective with their perception of its intrusiveness. This has been evidenced, for instance, in the case of Dutch public servants who generally support behavioral interventions but perceived simple reminders as paternalistic in an application targeting a behavior deemed unnecessary (Dewies et al. 2021). Additionally, the approach of having survey participants assess hypothetical scenarios rather than immersing them in real-life nudged decision-making situations is extremely sensitive to the specific wording employed. Consequently, there is a need for a more concrete and universally applicable conceptual framework in this regard.

The idea of this article, therefore, is to develop a typology of how nudges may hinder autonomy. The resulting typology will assist choice architects, policymakers, and other relevant stakeholders in critical thinking and systematic evaluation of nudges with respect to autonomy. Importantly, it provides the dimensions along the lines that choice architects have to think about when assessing or altering nudge design in pursuit of maintaining autonomy. Guided by the typology, choice architects may be able to identify nudge interventions that balance both respect for individual autonomy and effectiveness, or alternatively be poised to acknowledge the limits of nudging principles as to argue why more intrusive policy measures are demanded (Sunstein and Reisch 2014).

We note a caveat for our investigation: while we are focusing on autonomy, there are additional aspects influencing the ethicality of nudges beyond the scope of this study. The Nudge FORGOOD framework does address autonomy under “respect,” acknowledging the importance of autonomy and the freedom to choose (Lades and Delaney 2022). However, the framework also mentions other ethical concerns, such as the fairness of a policy’s redistributive effects, the availability of alternative policy options, citizens’ opinions on the goals addressed with a nudge, and the legitimacy of the choice architect to act out the role (Lades and Delaney 2022). Furthermore, some authors have counted in the concept of self-constitution to the fundamental principles of autonomous decision-making (Vugts et al. 2020). Presently, there remains uncertainty regarding whether nudges can actually impact higher-order preferences that make up self-constitution or if nudging someone to make a choice against their higher-order preferences merely serves to make the manipulation more apparent (Nys and Engelen 2017). The ongoing debate on this subject has yet to reach a definitive conclusion, and it presents challenges in terms of operationalization at this stage, which are not considered in the typology of this study.

The typology is developed and discussed in the context of nudges related to food choice. The rationale for focusing specifically on nudges related to food choice is three-fold. First, the food we eat can either support or threaten human health and environmental sustainability, thereby carrying a major impact on our well-being (Willett et al. 2019). As such, it is vital to understand how nudges impact our autonomy in this context in order to avoid undue infringement over basic human needs. Second, food is not just a matter of practical sustenance, but also an emotional, cultural, and moral aspect of our lives. Nudging food choices can therefore be particularly sensitive, and have the potential to be perceived as more intrusive than nudges applied in other behavioral domains (Sunstein et al. 2019). Finally, nudges on food choice address daily routine decisions that are made intuitively and instinctively (i.e., according to ‘System 1’ thinking) making them capable of greatly impacting individuals’ daily lives and habits (Wansink and Sobal 2007). Taken together, nudges aimed at influencing food choices have the potential for far-reaching implications.

The manuscript is structured as follows. In the method section, we provide an account of our approach to conducting a scoping review, where we delve into food nudging studies to identify the mechanisms that underlie autonomy and their representation in the empirical literature. In the results section, we demonstrate the ways in which nudge studies can impact autonomy and introduce a typology to better understand the nudge elements relevant to an autonomy assessment. The discussion section will then expand upon how these typologized dimensions have been addressed within the scientific literature and suggest potential applications for the typology in future research.

Methods

In the following, we describe the eligibility criteria, search procedure, title and abstract screening, data extraction and data synthesis of the scoping review:

Eligibility criteria

The current review adhered to the scoping review methodology recommended by the PRISMA-ScR reporting guidelines. An overview of the article selection process is illustrated (Fig. 1). We included articles published in peer-reviewed journals, excluding reviews, pertaining to empirical research on nudging individuals to choose healthy and/or sustainable foods, across any food consumption setting. Only articles that explicitly described their interventions as nudges were considered. Articles using the term “nudging” informally to denote behavior influence, without aligning their interventions with nudge terminology, were excluded. The publication date criterion considered studies published within the last ten years of the search period (01/2014 to 12/2023). Additionally, only articles written in the English language were included in our search.

Fig. 1
figure 1

PRISMA flow chart.

Full size image

Search procedure

After initial screening procedures, a systematic search was carried out on the Web of Science platform in January 2024. Various combinations of Boolean search terms were used in alignment with the research questions and defined research boundaries. Specifically, the search criteria included ((“Nudg*” OR “Choice Architecture”) AND (“Health*” OR “Sustain*”) AND (“Food” OR “Diet”)) within the topic field. Documents published in MDPI journals were omitted, resulting in the removal of 55 articles. The search was focused on the most pertinent Research Areas, which yielded over forty studies each. i.e., (“Behavioral Sciences” OR “Psychology” OR “Food Science Technology” OR “Public Environmental Occupational Health” OR “Business Economics” OR “Nutrition Dietetics” OR “Environmental Sciences Ecology”), thereby removing 87 articles. The systematic search yielded 385 articles.

Screening

The selected articles underwent a title and abstract screening process by the three investigators based on predefined inclusion/exclusion criteria. Article were excluded if they: i) were commentaries or theoretical contributions to the literature, ii) were not aimed at health or sustainability outcomes or were aimed at sustainability outcomes indirectly related to sustainability in food systems (i.e., reducing packaging or plastics, improving recycling), or iii) the study design did not allow to observe individual behavior. We considered various outcome measures, such as dietary outcomes (e.g., food choice), health metrics (e.g., BMI, weight, nutrient status), economic parameters (e.g., sales), and sustainability indicators (e.g., GHG emissions). In full-text screening, we also excluded studies i) combining nudges with non-nudge interventions, including a number of combinations with pricing strategies, ii) qualitative studies without empirical effectiveness examinations or without testing a nudge altogether, iii) studies lacking a nudge setup description or not yet implementing the nudge. A total of 146 articles remained for data extraction after the screening.

Data extraction, synthesis, and typology development

A single author (AE or SW) conducted data extraction utilizing a software for managing systematic reviews, with open-ended questions on each study (Supplementary file S1), outlining the i) study population ii) target behavior addressed with the nudge and choice setting, iii) status quo or control choice architecture, iv) nudge description, and v) nudge type — as campaigns, commitments, information mechanisms, transactional shortcuts, improved design strategies, warnings and reminders (Thaler and Sunstein 2008), vi) the classification of nudges according to the intrusiveness typology developed in this article, and lastly vii) whether the study observed significant, non-significant results with regard to the main outcome measure(s). The process of classifying nudges according to intrusiveness mechanisms involved assessing the nudge description to determine if it influences a mechanism relevant to autonomy. Subsequently, for those that do affect a mechanism, the assessment determines if this influence could potentially hinder autonomy. Any uncertainties identified by the authors were annotated and subsequently cross-verified by a second review author. For the typology of mechanisms that can hinder autonomy, we consolidated a preliminary set of identified studies. Each nudge was discussed amongst the research team to identify key mechanisms underpinning intrusiveness common across studies. The initial typology development was facilitated by an exercise amongst the authors, where a subset of the preliminary studies was considered for how the nudge design could be hypothetically modulated to reflect lower and higher degrees of intrusiveness. In addition, anonymous reviewers and colleagues with expertise in the field have commented on the initial typology, which has greatly helped to further develop it.

The typology was integrated into the review process to evaluate a systematically selected set of studies to discern whether the typology appropriately captured the intrusiveness of the included nudge designs, or whether the definition needed to be expanded. While the overall typology was found to appropriately encompass intrusiveness mechanisms, the process proved useful for refining sub-dimensions within each mechanism. In the result section, we provide example studies that call into question the preservation of autonomy concerning the typologized mechanisms. Additionally, the complete set of studies, encompassing further examples and the authors’ classification for intrusiveness, is also accessible (see Supplementary file S1). Furthermore, we identified key criteria to operationalize a measurement of intrusiveness by sub-dimensions to facilitate critical thinking as to how nudges might be modulated to mitigate intrusiveness. Although we may not establish a definitive threshold for determining minimal or high intrusiveness, the criteria we outline can assist choice architects in making better evaluations.

Results

The total sample of food nudge studies (N = 146) encompassing 251 interventions was reviewed and evaluated by intrusiveness on individual autonomy. We delineate the overarching mechanisms of nudges that might alter an individual’s autonomy: (1) the effort to opt out; (2) the affective influence, and (3) non-transparency (see Fig. 2).

Fig. 2
figure 2

A typology of nudge intrusiveness.

Full size image

Of the 251 interventions reviewed, 74 (29.4%) altered the effort to opt out, 127 (50.6%) leveraged affective influence, and 164 (65.3%) exhibited non-transparency. Of those interventions that altered the effort needed to opt out of the nudged option, the majority (70.3%) acted upon this mechanism in such a way that did not pose threats to autonomy; however, just under a third (29.7%) of these studies employed obstacles to opting out that run the risk of hindering autonomous decision-making, either by the degree of physical (Ni = 11) or economic effort (Ni = 11) required to realize preferences against the nudged option. The majority (88.3%) of interventions that leveraged affective influence did not pose threats to autonomy; however, a handful of these studies posed threats to autonomous decision-making by either the extent of their exploitation of social norm influence (Ni = 5) or emotional appeal (Ni = 10). Finally, of those studies that posted risks to autonomy under the umbrella of non-transparency, the bulk were characterized as imperfect due to non-transparency of the intervention itself (Ni = 125), and a few were marked as threatening autonomy due to non-transparency of alternatives (Ni = 4). Additionally, we provide an indication of intrusiveness relative to the nudge type, namely campaigns, commitments, information mechanisms, transactional shortcuts, improved design strategies, warnings and reminders (see Supplementary file S2)

These intrusiveness mechanisms, and their respective sub-dimensions, are not necessarily independent, and rather can interact. In addition to the examples highlighted in our results, which pose minimal risk to autonomy, there are nudges that operate independently of our defined intrusiveness mechanisms. For example, information provision or self-nudging emerged as interventions that do not align with any intrusiveness mechanism. In one such study, participants were informed about nudges before autonomously selecting their own, such as a reminder to increase fruit consumption (van Rookhuijzen et al. 2023). As another example, a few studies added options to the choice set (Attwood et al. 2020; Gill et al. 2022), which does not intrude via any mechanism. However, it is important to note that, while these were self-proclaimed nudges, such interventions do not necessarily adhere to the definition of a nudge, as they alter the choice set. In the remainder of this paper, we draw upon examples from the literature that act upon a mechanism relevant to autonomy to better illustrate each concept (see Table 1 for a summary).

Table 1 Example food choice nudges identified in the scoping review organized by intrusiveness mechanism.
Full size table

Mechanism 1: effort to opt out

The effort to opt out refers to the resources demanded of individuals in order to realize a preference against the nudged option. This requisite effort can be modulated along two sub-dimensions that are relevant to autonomy – (a) economic and (b) physical. The former consists of both time and monetary resources, as both underpin economic thinking. The second sub-dimension includes various physical activities such as walking, reaching, and carrying. Substantial effort – either physical or economic – would constitute a restriction to individual agency.

Economic resources

By definition, nudging explicitly promises to keep economic incentives constant (Thaler and Sunstein 2008). This conceptualization of economic incentives must include time, as it is a vital tenet of economic thinking (DeSerpa 1971) and is closely connected to monetary resources. Despite this reality, altering the time it takes to opt out of a nudge is frequently employed in nudge designs. For example, one study in a corporate cafeteria limited access to all-inclusive payment terminals where all items could be purchased, but not to the payment terminal where only designated low-calorie and mostly meat-free items could be purchased. In effect, this could increase wait times for the non-nudged alternatives by considerable amounts in what the authors refer to as the “hassle factor” (Bauer et al. 2021).

Besides obvious examples where researchers directly manipulate time resources, there can also be more hidden time costs in opting out of nudged options. In digital environments, for instance, opting out of a pre-selected option is just a quick click away. While this alone is not intrusive, the cumulative effect of facing numerous preselected items, each demanding individual action for removal, can become significant. In two studies, online grocery shopping carts were pre-filled with nutritious groceries such that shoppers could delete, add, exchange, or keep items in their cart. With upwards of twenty items preloaded into the carts, the time investment required to opt out of each individual selection could become considerable, unless efficient design features enable the selection of multiple options at once (Coffino et al. 2020, 2021).

Considering the over 200 food choices we make on a daily basis (Wansink and Sobal 2007), even a five minute demand to opt out can be a significant ask. In contrast, durations of less than one minute — such as the time required to request an alternative from a restaurant server (e.g., (Ferrante et al. 2022; Gravert and Kurz 2021; Radnitz et al. 2023) — should be considered negligible and can hardly be avoided in the implementation of any decision.

Monetary incentives are seldom associated with nudging due to their general exclusion from the nudge framework, with the exception of near negligible (dis)incentives that are easy and cheap to avoid (e.g., 5 cent plastic bags in supermarkets) (Hansen 2016). Therefore, pricing strategies are only considered as complements to be used with nudging tools (Kraak et al. 2017). Nevertheless, some nudges can indirectly influence the monetary cost of opting out. For example, the bundling of products, changes in portion sizes, and use of non-monetary rewards can lead to increased relative purchasing costs of non-nudged options. Imagine a scenario where a burger is either bundled with a side of fries or a side salad. In cases where it is not possible to switch the bundled side free of charge, or if the information regarding this option is not readily evident (Diaz-Beltran et al. 2023), customers may end up paying extra to add on the additional desired side. In this self-proclaimed nudge instance, the current state of affairs determines how a basic marketing strategy — the bundling of options — alters financial incentives on the decision-making process.

Regarding portion sizes, a restaurant intervention increased the default portion size of vegetables and decreased the portion size of meat in their dishes (Qi et al. 2022). In this scenario, the original size of components needs to be offered at the same price to avoid any monetary incentives. This particular study did not grant this alternative to the default option, effectively restricting choice.

Rewards or gifts employed as nudges, though potentially negligible in monetary value, may limit personal autonomy. Consider an intervention in which glow-in-the-dark bracelets of relatively low material value were affixed to white (but not chocolate) milk cartons in a school cafeteria (Lai et al. 2020). In this scenario, choosing chocolate milk incurs a financial disadvantage, particularly for young children who might place excessive value on such items. This issue is echoed by the long-standing debate in the US concerning toy incentives in kids’ meals at fast-food establishments, with some advocating for their prohibition due to concerns about limited autonomy and the negative influence on children.

Conversely, this review included multiple studies that successfully employed economic incentives without casting doubt on the preservation of individual autonomy. These approaches included strategies such as monetary framing (Carroll et al. 2018; Policastro et al. 2017; Yi et al. 2022), e.g., healthy food bundles offered without a discount, framed as “5 items for $5” (Carroll et al. 2018), or scarcity cues (Fennis et al. 2020; van Rookhuijzen et al. 2021) (e.g., “available while supplies last”).

Physical resources

Physical effort has been identified as a key component of the desire for “convenience” that drives food choices (Wales 2009). Campbell-Arvai and colleagues (2014) leverage this desire for physical convenience by serving a default vegetarian menu at the tables, while informing verbally and in writing of a second menu with meat options posted on the wall approximately 3.5 meters away from their table, observing significant increases in the percentage of patrons selecting vegetarian meals relative to the control condition. Baskin and colleagues (2016) also observe significant declines in snacks selected by employees in a large company when the snack station was placed an additional 2 meters away from the beverage station in the company break room. Other changes that require seemingly small shifts in physical effort navigating the space available, such as placing healthy (unhealthy) items closer (further) within reach (Knowles et al. 2019; Quinn et al. 2018; Seward et al. 2016; van Rookhuijzen and de Vet, 2021) can result in significant changes in behavior. Less obvious applications within this category are changes to encourage smaller portion sizes, such as providing smaller spoons or plates for self-service. The degree of physical effort required to opt out of the nudged option distinguishes between a minor rearrangement of choices and a more intrusive one that may compromise individual agency to choose freely against the nudged option.

Mechanism 2: affective influence

Engaging emotionally-laden eating goals through “healthy eating calls” and “hedonic enhancements” have been classified as affectively-oriented nudging (Cadario and Chandon 2020). We build upon this consideration of nudges that aim to influence decision-making through affective means – i.e. emotions – to also include social norm nudges. Social norms possess the capacity to encroach upon personal autonomy in decision-making situations where privacy or discretion is lacking, thereby making individuals feel unable to opt out of a nudge due to peer pressure and fear of social judgment.

Social norms

Social norms serve as codes of conduct to guide socially appropriate action and have been found to strongly influence food choice, including quantity, healthiness, and hedonic evaluation of food consumed (Higgs and Thomas 2016). By conforming to social norms, individuals experience positive emotions associated with social acceptance and belonging. In the case of sustainable consumption, for example, individuals have been found to experience feelings of satisfaction when they choose ethical or environmentally-friendly foods, in what is called “warm glow” (Iweala et al. 2022). Conversely, fear of social judgments around non-conformity can also have powerful steering effects on individuals (Higgs 2015). The effect is moderated by group identity (or lack thereof) (Liu et al. 2019).

Social norms can be cued by setting defaults (e.g., Coffino et al. 2020, 2021; Dalrymple et al. 2020; Diaz-Beltran et al. 2023; Gravert and Kurz 2021; Hansen et al. 2021), which individuals tend to view as an implicate recommendation or normative option (Everett et al. 2015). Alternatively, re-configuring menu designs to position or otherwise frame certain choices as the normal option (e.g., Bacon and Krpan, 2018; Bergeron et al. 2019; Boronowsky et al. 2022; Campbell-Arvai et al. 2014; van Kleef et al. 2018) and altering default portion or plate sizes (Davidson et al. 2021; Libotte et al. 2014; Qi et al. 2022; Zhang et al. 2024). Social norms can also come in the form of explicit messaging that conveys a descriptive norm (Gottselig et al. 2023; Jesse et al. 2021; McGrath 2023; Otto et al. 2020; Reinholdsson et al. 2023; Suleman et al. 2022) such as signage that reads “most people choose fruit and vegetables” (Bauer et al. 2022) or a recommendation or injunctive norm such as “improve your score” when grocery shopping (De Bauw et al. 2022; Kroese et al. 2016; Panzone et al. 2021).

We argue that the intrusiveness of this class of interventions is contingent upon the presence of social pressure. On this note, privacy is a key factor which facilitates discretion in decision-making, enabling individuals to make choices without immediate social repercussions. The connection between privacy and decision-making has been previously explored (Huh et al. 2014). In settings that offer anonymity or a degree of discreteness, individuals may have more agency in their choices. In such environments, the use of social norms to subtly guide decisions, while still allowing for individual discretion, can be seen as an autonomy-preserving intervention. However, in embedded choice settings characterized by a highly social environment, the presence of strong social norms should be considered as a potential threat to autonomy, as individuals may feel compelled to conform to the prevailing social expectations in public settings. Relevant examples identified in this review include prompts at checkout counters encouraging patrons to choose fountain water over soda to support a local soup kitchen (Policastro et al. 2017), instructions given to parents in community settings to make healthy choices for their children (Loeb et al. 2017), and requests for patrons to downsize to smaller meals to combat food waste (Qi et al. 2022).

The impact of social messages varies between a communal cafeteria setting with frequent social interactions and more detached decision contexts (e.g., online grocery stores). Since the social dynamics of decision environments are often not described in the nudging literature, we cannot conclusively address the utilization of social norms in highly social settings. However, we highlight the potential risk to individual agency in such scenarios where social pressure is empirically confirmed.

Emotional appeals

In the realm of food nudges, choice architects can aim to elicit a salient emotional response to make nudged options more appealing and/or make non-nudged options less appealing. Concerns to autonomy arise in the event that the effectiveness of a nudge hinges on the motivation to elicit negative emotions, such as fear, anger, sadness, shame, guilt, envy, disgust, or contempt (Plutchnik 2001). A recent review of studies examining the relationship between affective influence and agency observed that negative stimuli were associated with a lower sense of agency, as indicated through both self-reporting and implicit measures (Kaiser et al. 2021). Loss aversion, and the resulting negativity bias, in which humans tend to pay heightened attention to, learn from, and consider negative information relative to positive information during decision-making (Rozin and Royzman, 2001; Vaish et al. 2008), may also be a pertinent consideration to agency. As such, careful consideration must be taken in the use of information that could be construed as negative to ensure that such interventions promote autonomy rather than hinder it.

This discussion is particularly relevant in the context of interventions meant to “warn” consumers against adverse health and/or environmental impacts of certain food choices. In the case of front-of-package nutrition labels, several studies have demonstrated that these tools enhance consumer understanding of the nutrition composition of packaged foods and beverages (Temple 2020). Several studies focused specifically on warning labels, which label foods “high in” or in “excess of” sugar, salt, and/or saturated fats, have found that consumers indicate high acceptance of these labels and consider them useful to inform purchases (Bopape et al. 2021; Sato et al. 2019; Vargas-Meza et al. 2019). However, choice architects should be wary of other “warning”-type interventions that may cross the line into emotional manipulation. For instance, Aldrovandi and colleagues (2015) examine the effect of presenting rank information (e.g., “you are in the most unhealthy 10% of eaters”) on students’ willingness to pay for healthy foods, an intervention which runs a higher risk of effectiveness based on triggering shame. This intervention also overlaps with social norm messaging. Similarly, Caso and colleagues (2023) test the influence of fear-based messages that communicate the irreversible consequences of a high intake of red and processed meat in terms of disease and death on self-reported future meat consumption.

On the other hand, several studies were identified in this review which leveraged emotional appeals without linking to negative emotions or posing a risk to autonomy, such as those that sought to highlight healthy and/or sustainable options through the use of hedonic descriptions or sensory appeals, or adding smiley faces (Mecheva et al. 2021) or cartoon characters to healthy options (Ozturk et al. 2020).

Mechanism 3: non-transparency

A prominent criticism of nudges is that they shift behaviors through the manipulation of biases. In this context, transparency has emerged as a key concept to preserve consumer autonomy (Hansen et al. 2021; Michaelsen 2024; Wachner et al. 2020). The concept has generally been defined as making both the existence of the nudge and its intended objective known (Michaelsen 2024). We incorporate and broaden the concept of transparency to also evaluate whether, and to what degree, the nudge clarifies alternatives to the nudged choice — a topic scarcely addressed in the literature on the ethics of nudging.

Non-transparency of intervention

To date, the empirical literature on nudging offers limited insights into whether individuals can actually recognize a nudge and its intended purpose. On the one hand, attempts by choice architects to openly disclose nudges often go unnoticed, indicating that people frequently fail to understand the information meant to enhance transparency. Various studies have reported accuracy rates around or below 50% in tests with simple multiple-choice questions designed to check for manipulation awareness (Michaelsen 2024). On the other hand, there is evidence that individuals can often identify nudges even without explicit notification, implying individuals might recognize nudges even when they are not overtly disclosed (Michaelsen 2024). In this review, our attention centers on the actions of choice architects that affect the autonomy of individuals, encompassing deliberate efforts to inform about the presence or aim of the nudge. Several studies have aimed to explore the effects of enhancing the transparency of nudges, with the goal of ensuring that their effectiveness is not solely due to exploiting unconscious biases.

For example, consider a nudge intervention to shift the default options at the cash register from unhealthy to healthy snacks in a store. By placing a sign stating “we help you make health(ier) choices” (Cheung et al. 2019; Kroese et al. 2016), the sign informs on the purpose of the nudge in the shop. In some cases, these communications directly highlight the implementation of an intervention. For example, in an aforementioned study in an online supermarket, participants encountered a shopping cart preloaded with selections intended to mirror a “nutritionally balanced grocery shopping cart tailored to their gender and age”, effectively making them aware of the intervention’s purpose and existence (Coffino et al. 2021).

Messages conveying transparency can either directly highlight the purpose or presence of a nudge, or they might necessitate more advanced inferential reasoning through indirect cues, such as health-related posters in the decision-making environment that lack a clear spatial or thematic connection to the specific nudges implemented (Antunes et al. 2024). The latter method signifies a compromise on autonomy protection, despite its potential to help consumers recognize the intentional design of the choice architecture. Another aspect worth noting in detecting interventions is the frequency of exposure (singularity) to both the choice and the choice architecture. Interventions aimed at frequent patrons are more likely to be noticed as a change, particularly by customers dissatisfied with the nudged choice, who will promptly opt out. The frequency of exposure serves as a safeguard against misleading nudges (Lemken 2021). In contrast, irregular visitors may have difficulty discerning the nudge.

Enhancing the transparency of interventions is one approach to enable a deliberative process. However, it’s worth noting that a lack of transparency in nudges does not necessarily obstruct the deliberative process. In addition to instances where individuals frequently identify the nudge, and choice architects are recommended to disclose this, there are also scenarios where such disclosure is not necessary. This is clear for purely descriptive nudges which automatically point to the presence of the intervention, for example, simple labels indicating “organic” or “local” meal quality (Migliavada et al. 2022). Another example is the use of floor arrows to direct customers towards healthier food choices in retail and/or food serving settings (Allan and Powell 2020; Bauer et al. 2021; Chapman et al. 2019; Luomala et al. 2023). The awareness of the nudge requires some level of processing the intervention; otherwise, the intervention cannot be effective or suspected of working in the dark. In such cases, additional transparency messages seem unnecessary. This becomes even more evident for self-nudges (van Rookhuijzen et al. 2023) or the provision of commitment tools (Jia et al. 2022; Panzone et al. 2024; Samek 2019), where cognitive reflection on the choice is inevitable, and consumers actively modify the choice architecture according to their preferences. Further transparency is not deemed necessary.

Non-transparency of alternatives

In decision-making processes, ensuring transparency regarding alternatives is paramount. A significant concern arises when alternatives become invisible. This poses a threat to consumer autonomy by reducing the choice set that is actually considered and limiting the ability to make informed choices. A nudge designed to change the visibility of alternatives acts upon transparency of options, though without necessarily making options invisible. In most cases, the nudge intends to increase the visibility of nudged options but accidentally influences the prominence of alternatives. The extent of this influence varies widely, ranging from subtle interventions like positioning meat alternatives alongside meat products in supermarkets, to harmonize the chance of finding such products (Vandenbroele et al. 2021), to more intrusive ones where consumer awareness of alternatives is severely limited, suggesting that freedom to choose only exists in the abstract. A deliberative decision-making process necessitates, at the very least, a reasonable opportunity to notice the presence of alternatives. This requirement becomes particularly concerning when choice architects actively conceal alternatives to impede deliberation, such as hiding sugary beverages at the bottom of coolers behind a frosted film on the glass front (Mikkelsen et al. 2021). In this case, the use of frosted film presents an intentional barrier to the deliberation process. A modified version of this study, which merely repositions sugary drinks to the bottom of coolers, might be viewed more favorably because it merely re-organizes products based on available space. The latter constitutes a forced choice architectural decision that must prioritize products.

Achieving complete parity in product presentation is often impractical or impossible. Numerous studies (e.g., Meeusen et al. 2023; Young et al. 2020) explore repositioning nudges that simply change the positioning of nudged and non-nudged options to alter visibility, without making options invisible. It is crucial to understand that the status quo should not serve as the benchmark for evaluating visibility in a particular context; rather, the focus should be on how difficult it becomes to notice an option. Additionally, there may also exist methods to purposefully decrease the visibility of alternatives without unduly limiting consumers’ ability to consider them. For instance, implementing a nudge on an online ordering platform could involve adding a partially opaque white layer over the images and product information of unhealthy products (Michels et al. 2023).

Several researchers have noticed the autonomy issue that arises when alternatives become challenging to consider due to their lack of visibility. To address this, researchers have devised a workaround by still reducing the visibility of alternatives while actively referencing them to increase the likelihood that consumers are aware of the possibility to opt out (Campbell-Arvai et al. 2014; Erhard et al. 2023; Gravert and Kurz 2021). For instance, this approach might involve presenting a default plant-based meal with an option to opt out to a meat meal with a simple click (Erhard et al. 2023). This setup aims to enable a reflective choice process, allowing consumers to evaluate the nudged option first while being explicitly informed of alternative choices, typically with minimal effort required in switching. Therefore, such prompts can serve as a choice architectural tool to enhance autonomy and possibly preserve effectiveness. It’s worth noting a nuance in this approach. Choice architects can choose to explicitly name alternatives or simply prompt their existence. For example, Gravert and Kurz (2021) redesigned an “a la carte” menu to offer a choice between a vegetarian and fish dish versus a meat and fish dish, informing patrons that they could request meat without providing a further description of the dish. While providing more information is generally beneficial from an autonomy perspective, the cognitive deliberation process may have its limits in real-world settings.

Another interesting nudge approach that may maintain autonomy yet initially hides alternatives requires that options be made unavailable or not visible during the initial phase of making a future or delayed selection. All choices are then revealed upon a second evaluation at the time of the final decision (Schlegel et al. 2021). While this commitment nudge aims to engage consumers in a more thorough decision-making process, empirical evidence may find most consumers do not reassess their options, leading them to perceive a more restricted choice set mistakenly.

Discussion

In this paper, we delve into a crucial topic: autonomy preservation in nudging strategies. Namely, drawing upon insights from existing literature, we have constructed a typology for evaluating and categorizing the diverse mechanisms that underlie the intrusiveness of nudges in the context of food choices. In devising these three mechanisms – effort to opt out, affective influence, and non-transparency – and relative sub-dimensions, we lay the foundation for a more sophisticated comprehension of how nudges can affect an individual’s ability to make independent and deliberate choices. Here, we discuss how each mechanism has previously been touched upon by other researchers and how to move forward with the typology, including summarizing possible criteria from the results that can be used by choice architects to evaluate nudge intrusiveness along the identified mechanisms (see Table 2).

Table 2 Proposed measurement criteria for each intrusiveness mechanisms.
Full size table

Effort to opt out

The first criterion that choice architects and other relevant stakeholders should consider in evaluating the intrusiveness of a nudge is whether, and to what extent, a degree of effort is required to opt out. In terms of economic resources, this could refer to elements of time, such as search time needed to identify an alternative option, or transaction time needed to execute the decision against a nudged option, such as by filling out a form or making a call. Depending on the size, the use of monetary aspects, such as foregone material gifts or rewards for those who opt out, and the extent to which these might be valued by those to be nudged, should also be considered as a potentially autonomy-threatening dimension. In terms of physical resources required to opt out, this involves a demand on fitness of some sort, such as standing up, walking, or reaching.

In contrast to the studies reviewed, where opting out required significant effort, it’s often observed in practice that the collective effort to opt out of individual nudges — termed “nudge stacking” — is more prevalent. Nudge stacking has been identified as objectionable to the paternalism libertarian framework in that multiple nudges can sum up to a “shove” (Coons and Weber 2013). In addition to the time and effort to opt out, nudge stacking also relates to non-transparency insofar as layered nudges make it more difficult for even watchful decision-makers to identify the mechanism behind nudges and therefore easier for choice architects to hide nudges (Ivanković and Engelen 2019). The prevalence of nudge stacking in the marketplace is one reason why market nudges have been identified as particularly autonomy-threatening (Ivanković and Engelen 2024). This has particular relevance in online environments. Consider frequently employed “dark pattern” nudges in which multiple buttons that should be clicked in order to proceed as desired are in bigger font, centered, and/or boldly colored to draw attention, while alternatives are tucked away in small corners of the screen (Reisch 2020).

Affective influence

Another crucial aspect for evaluating the intrusiveness of nudges is their affective influence, encompassing social norms and emotional appeals. Social norms may leverage social pressures and normative expectations. Negative social norm messages meant to discourage behaviors can have a somewhat stronger impact on affect compared to positive messages designed to encourage behavior. This is primarily attributed to the well-documented “negativity bias,” where humans tend to pay heightened attention to, learn from, and consider negative information during decision-making (Rozin and Royzman 2001; Vaish et al. 2008). Nonetheless, such negative normative cues need not necessarily pose a threat to autonomy. Substantial social pressure limiting deliberation can primarily be anticipated in settings where decisions are made publicly and are subject to controversy.

Emotional appeals tap into negative emotional cues such as fear, anger, sadness, shame, guilt, envy, disgust, and contempt, driving decision-making processes. Relatedly, emotional responses to negative stimuli tend to be stronger than those to positive stimuli (Vaish et al. 2008). This heightened emotional reactivity, particularly under stress, can potentially impede an individual’s ability to process information rationally. It’s important to emphasize that, in response to text-based warning messages, which can sometimes be found on ultra-processed foods, any potential impact on agency remains relatively manageable, as most individuals can still engage in a deliberation process when reflecting on a written message. However, the emotional processing of graphics (consider cigarette packaging in many countries) can be involuntary, so that agency is reduced for better or worse. Not yet considered are stimuli that trigger positive emotions. Positive emotional stimuli have been found to be associated with an increased sense of agency (Kaiser et al. 2021) and improved decision-making processes (Tran et al. 2012). Skillful use of these stimuli presents a promising opportunity to implement effective nudging interventions that preserve autonomy.

Non-transparency

A significant criticism of nudges involves their potential to manipulate biases and influence behavior without individuals’ awareness. Transparency has thus become crucial for preserving consumer autonomy. For non-transparency of intervention, we consider the lack of direct or indirect disclosure regarding the presence and purpose of the nudge. This includes instances where nudges are subtly hinted at, such as through posters on the topic of intervention. The singularity of decision refers to the frequency of exposure to both the choice and the choice architecture. Interventions targeting frequent patrons are more likely to be noticed as changes, providing a safeguard against misleading nudges, whereas irregular visitors may struggle to discern the nudge (Lemken 2021). Finally, non-transparency of alternatives entails assessing the visibility of alternative options or prompts, as well as the absence of prompts directing individuals to consider alternative choices.

Providing a transparency statement transforms an intervention into a “double nudge,” potentially amplifying its impact on behavior and individual agency. This is crucial for nudges that sidestep traditional decision-making processes (Michaelsen 2024; Wachner et al. 2020). Yet, many nudges in this review, such as messaging nudges and self-nudging strategies, clearly do not bypass decision-making processes. For other nudges, adopting a precautionary approach, the inclusion of a disclosure statement seems helpful, with initial studies showing that it does not compromise effectiveness while boosting agency (Bruns et al. 2018; Cheung et al. 2019; Dranseika and Piasecki 2020), identifying a potential sweet spot for autonomy enhancement. However, the applicability of this approach across different settings and how well such nudges target specific audiences still warrants investigation.

Decision-making is often characterized by bounded rationality, suggesting that decisions, especially in the food domain (Wansink and Sobal 2007), are not always based on rational thinking, even without the influence of nudges. Nudges can encourage more thoughtful consideration of options without requiring explicit transparency. Nonetheless, incorporating a transparency message for decisions that occur less frequently is recommended to safeguard autonomy. In certain scenarios, where more intricate reflection is achievable, ethical nudging becomes particularly pertinent, aligning with behavioral public policy’s goal of enhancing citizen autonomy (Banerjee, Grüne-Yanoff, et al. 2023). For example, the “nudge + ” initiative aims to bolster citizen empowerment by promoting critical analysis and transparent assessment of nudges in advance (Banerjee, Galizzi, et al. 2023; Banerjee, Grüne-Yanoff et al. 2023). This approach facilitates individuals in maintaining decision-making autonomy, marking a progression towards more ethical nudging practices where feasible.

The risk of overlooking alternatives was acknowledged before. Lades and Delaney (2022) explain how default settings, which dictate the outcome if individuals take no action, could cause busy and rationally limited individuals to perceive that they lack choice. Consequently, the freedom of choice for these individuals is diminished when they are unaware of the available options (Lades and Delaney 2022). In general, this design feature was widely overlooked in ethical nudge assessments despite the substantial autonomy risks of the invisibility of alternatives, while a number of empirical studies in the review have reported on efforts to make individuals aware of alternatives to the nudged option. For autonomy-enhancing nudges, we recommend to not purposefully lower the visibility of alternatives or apply explicit prompts to alternatives in case the intervention may have reduced the visibility of non-nudged options (see Table 2).

Limitations

This review focuses exclusively on food choice nudges, which may limit the generalizability of the findings to other domains such as health, finance, or environmental behaviors. Additionally, nudges and their impact on behavior can change over time as individuals become more aware of them. This study does not account for the dynamic nature of nudges and how repeated exposure might alter their effectiveness and intrusiveness, nor does it consider other dimensions that are important to an ethical evaluation of nudging such as fairness, consent, and the potential for manipulation. The emphasis on autonomy may overlook other critical factors that influence the acceptance and effectiveness of nudges, such as cultural values, social norms, and individual differences in decision-making processes. A more holistic ethical analysis is necessary to fully understand the implications of nudge strategies. Finally, this study is limited in its ability to assess autonomy threats of nudging in the case of nudges that are used as part of broader policy mixes and integrated with other policy tools (Holz et al. 2023; Merkelbach et al. 2021).

Policy implications

The developed typology of nudge intrusiveness provides a framework for choice architects and policymakers to design and evaluate nudges that respect individual autonomy. This framework can guide the creation of interventions that are less intrusive while still promoting desired behaviors, potentially increasing public acceptance and the ethical validity of nudging practices. By highlighting the mechanisms through which nudges can undermine autonomy — such as effort to opt out, affective influence, and non-transparency — this study informs policymakers about key ethical considerations necessary when implementing nudge strategies. It underscores the importance of maintaining transparency and providing easy opt-out options to uphold consumer autonomy, ensuring that nudge strategies can be both effective and ethically sound.

Future directions

There remain open ethical questions for nudging. For instance, the deliberative nature of increasing transparency (of an intervention or alternatives) often translates to increased cognitive effort. In principle, furnishing consumers with more information neither constrains freedom of choice nor diminishes personal agency. For instance, the inclusion of nutritional labels, nudge disclosures, or details on alternatives serve an informative purpose. As such, we argue that cognitive effort aimed at prompting deliberation enhances autonomy by enabling informed decision-making. However, there is a level of information which risks overwhelming the deliberation process, but additional cognitive effort does not automatically translate into an autonomy risk. This boundary condition and the cognitive effort required to resist nudging attempts might be considered a threat under other ethical frameworks. In line with this, we question the overuse of disclosures to alert decision-makers to the presence and purpose of nudges to be burdensome or even autonomy-threatening (think nudge stacking). This is an area for future research on the kind of context that demands and allows for disclosures. Relatedly, there is a link between individuals’ awareness of their own limited cognitive capabilities and willingness to outsource regulatory mechanisms to governments (Grelle and Hofmann 2024; Kukowski et al. 2023). This is an interesting area to explore with regard to cognitive effort and nudge acceptance.

Future studies could also explore complementing the use of this typology by assessing decision-maker’s opinions on nudge elements. Rather than solely inquiring about their sense of freedom to choose, as in the traditional perceived intrusiveness approach, more focused questions based on the typology presented here can be devised. For instance, gauging perceived social pressure from peers may shed light on the autonomy in decision-making regarding affective influence. While this approach remains subjective and potentially contentious, employing more targeted questions minimizes the likelihood of conflating ethical concerns unrelated to autonomy, such as opinions on the nudge’s objectives and similar matters. Additionally, this approach ensures that patrons are confronted with a cognitive concept commonly understood, while the freedom to choose and consumer autonomy remains a topic not widely comprehended even among researchers.

In conclusion

Offering a more nuanced understanding of the factors influencing nudge intrusiveness, our paper adds a valuable perspective to the ongoing discourse surrounding the legitimacy and feasibility of employing nudge strategies. As behavioral interventions continue to exert a significant influence on public behavior, our typology serves as a valuable resource for encouraging critical thinking and responsible decision-making among choice architects. Ultimately, the insights presented herein can serve as a compass for a more ethical use of nudges, ensuring that these interventions align with societal values and uphold individual autonomy. Autonomy-preserving nudges will find it easier to gather widespread support in public policy and with private actors, although they should not be misunderstood as a necessarily sufficient solution to an underlying problem. Depending on the success of lowly intrusive policy measures, a restriction of autonomy can be demanded to improve the functioning of markets or mitigate environmental issues. While nudges may alter decision-making environments, they should be assessed against alternatives like subsidies or taxes, which also influence free choice (Lades and Delaney 2022; Mukerji and Mannino 2023). Nevertheless, redesigning nudges to lowly intrusive policy instruments could be the smallest common denominator to initiate behavioral change.