Introduction
The democratic quality of decision-making is increasingly assessed not only by how citizens are involved in policymaking, but also by how well they consider diverse interests in doing so (Dryzek and Tanasoca, Reference Dryzek and Tanasoca2021). Dryzek et al. (Reference Dryzek2019) suggest that deliberation can help overcome polarization and echo chambers. They posit that deliberation generates social conditions under which individual reasoning improves (Dryzek et al., Reference Dryzek2019). Concretely, deliberation is said to expose participating citizens to a variety of interests, pooling different perspectives for them to consider. While the opinions of lay citizens taking part in participatory processes have become a key area of research and practice (Luskin et al., Reference Luskin2022), we have little empirical evidence that considered interests evolve in deliberation. As democratic innovations like minipublics become more prevalent (Elstub et al., Reference Elstub2019), understanding the extent to which participants integrate diverse interests into their deliberations is crucial for assessing how much they will reflect on more than their own egocentric concerns.
Deliberative minipublics bring together a diverse, near-randomly selected sample of the population to form considered judgments on public issues (Setälä and Smith, Reference Setälä, Smith and Bächtiger2018). Theorists have different perspectives on the types of interests that participants in minipublics take into account and the effects that incorporating these interests may have on the outcomes of the deliberative process (Barber, Reference Barber1984). Some contend that minipublic participants consider mostly their own interests and focus on their egocentric preferences in deliberations (Lafont, Reference Lafont2019). According to others, participants may integrate external interests into their considerations (Mansbridge, Reference Mansbridge and Mark1992). They are exposed to others’ views, feelings, and thoughts on an issue, weighing new arguments and justifying their opinions with reasons, which creates viewpoints beyond one’s self (Mansbridge, Reference Mansbridge, George and Russell1993). Participants may also consider overlapping interests, contingent upon the specific topic and deliberative process (Mansbridge et al., Reference Mansbridge2010).
We explore the role of other-regarding interests in deliberation, focusing on how participants balance their own interests with those of others as a consequence of participating in a minipublic. As a result, we offer a revised view of interest consideration within minipublics through a novel longitudinal analysis. We argue that the participants’ ability to consider self- interest versus other-regarding, that is, sociotropic interests, and to weigh multiple interests in a minipublic is essential for deliberation, as it enables them to arrive at well-informed and thoughtful judgments through discussion (Luskin et al., Reference Luskin2022).
Specifically, we leverage participant survey data and transcripts from five Swiss minipublics held between 2019 and 2023. These deliberative forums addressed the topics of affordable housing, climate change, and politicians’ pensions and were formally embedded within local and cantonal decision-making processes. Switzerland, with its strong tradition of consensus democracy and its direct-democratic instruments that allow citizens to participate directly in political decisions and referendums, is the most likely case for the formation of considered interests through deliberation. Both its inclusive nature and popular votes promote citizens’ knowledge of policy-related arguments in debates and lead them to consider a range of interests before making a policy decision (Colombo, Reference Colombo2018; Lacey, Reference Lacey2017). Albeit the available public discourses may influence participants’ own biases (Mercier and Landemore, Reference Mercier and Landemore2012), some characteristics of minipublics work against these potential biases. For example, the stratified random selection strikes a balance of views by drawing on demographic or opinion-related criteria. In terms of the generalizability of the findings, studying an easy case means that if we find no empirical evidence of the formation of considered interests, we can conclude that deliberation in more adversarial contexts is unlikely to lead to the expression of sociotropic interests. Moreover, elite cohesion against citizen-initiated popular votes is relatively high in Switzerland (Trechsel and Sciarini, Reference Trechsel and Sciarini1998), which might lead to a concentration of more like-minded perspectives within the minipublics connected to popular votes.
This article contributes to the literature by examining whether participants in minipublics are more likely to account for others’ interests during deliberation. The findings suggest that participants’ convictions are fluid rather than fitting into rigid categories of being fully aligned or partially aligned with certain interests. Individual traits are less decisive for an increased likelihood of sociotropic interest consideration than the type of process. Older participants and those with lower income are more likely to consider others’ interests. Crucially, coupling a minipublic to citizens’ initiatives enhances sociotropic considerations. Finally, we offer guidance on how to use our approach of a novel survey instrument to measure interest consideration for future applications. Overall, this study offers insights into improving the effectiveness of deliberative processes and fostering greater cooperation in democratic decision-making. Observing the potential transformation of participants’ interests in deliberative minipublics deserves scholarly attention because it can lead to the modification of self-interest and identities, especially those of disadvantaged groups.
Interests in deliberation
We follow Mansbridge et al. (Reference Mansbridge2010) in defining interests as deeply held beliefs of the truth discoverable through reason, action, and by enabling expression. We opt for their definition, because it aligns best with an empowered context, such as that of a minipublic. Interests differ from prereflective opinions in that they comprise ‘understandings of self and commitments that are more considered, emotionally and rationally, and more thoroughly tested in action’ (Mansbridge et al. 68). Mansbridge et al. (Reference Mansbridge2010) clarify that interests, while comprising monetary self-interest in general, run deeper than superficial policy arguments and positions when including ‘other-regarding and ideal-regarding commitment[s]’ (Mansbridge et al. 68). Distinct from empathy (Muradova, Reference Muradova2021), sociotropic commitments consider the deeper interests of others or ideals beyond one’s own egocentric concern. Therefore, sociotropic interests are also referred to as ‘considered’ or ‘other-regarding and ideal-regarding’ (Mansbridge et al. 68).
Mansbridge et al. (Reference Mansbridge2010) stress the importance of self-interest in deliberation by emphasizing its uses and legitimacy among deliberative virtues. Aggregated self-interest can act as information about what exactly qualifies as the common good. Yet, self-interest can also serve as a justification for supporting a specific policy. It becomes productive in deliberation when it is tempered by reason and guided by principles of respect and fairness throughout the process (Habermas, Reference Habermas and Bächtiger2018). In this vein, Habermas (Reference Habermas and Bächtiger2018) saw self-interest as the involved parties’ starting points whose transformation is possible only because a democratic society cannot function effectively if its citizens, in their roles as national citizens and co-legislators, focus solely on their own interests.
Deliberative minipublics present a particular form of democratic innovation wherein deliberation has demonstrated the capacity to cultivate an environment conducive to transforming interests (Grönlund et al., Reference Grönlund2015; Niemeyer et al., Reference Niemeyer2024). Mansbridge (Reference Mansbridge, George and Russell1993) details potential pathways to interest transformation. For example, participants transform their identities by taking different perspectives after being exposed to the alternative viewpoints they consider while deliberating. They might also modify or override their (self-)interest due to developing new or modified perceptions of justice (Mansbridge et al., Reference Mansbridge2010). Participants may change their identities in response to giving more weight to the interests of communities that have been less considered before the deliberation. Collectively, they undergo a metamorphosis of their (self-) interests, and, as a result, their sense of identity evolves through the adoption of a perspective that considers the interests of others (Mansbridge et al., Reference Mansbridge2010).
Conditions of interest transformation
Practices of including broader other-regarding views in deliberation like intersubjective reasoning, storytelling, and empathy have been subject to extensive empirical analysis (Black, Reference Black and Nabatchi2012; Luskin et al., Reference Luskin2002; Niemeyer et al., Reference Niemeyer2024). However, existing studies have rarely focused on the degree to which and why minipublic participants consider self-interest or other-regarding interests. Myers (Reference Myers2017; Reference Myers2018) investigated interests in deliberation specifically but focused on participants’ beliefs about conflicting interests in non-facilitated discussions. Considering others’ interests made it more likely that a person would transform their own interests to include those of others.
This transformation of interests allows participants in minipublics to develop considered judgments – choices they would make under ideal conditions that reveal their true preferences (Luskin et al., Reference Luskin2022). These considered judgments – participants’ opinions after deliberation – were first studied systematically within the Deliberative Poll model of a minipublic (Fishkin, Reference Fishkin2009; Yankelovich, Reference Yankelovich1991). Deliberation has been shown to play a crucial role in making judgments more thoughtful, aligning them closer to enlightened preferences as the minipublic progresses (Dryzek et al., Reference Dryzek2019).
Here, we present two key arguments. First, we argue along Mansbridge et al. (Reference Mansbridge2010) that, as people talk things over and listen to each other, they begin to think more deeply about their needs and those of others. This transformation can make participants’ self- and other-regarding interests not only more refined, but also allows them to make representative claims (Montanaro, Reference Montanaro2017; Saward, Reference Saward2010). Second, we argue that considering other-regarding interests can help participants form representative claims, which are defined as individual and institutional non-electoral aspirations to represent another actor or entity politically (Saward, Reference Saward2010). Guasti and Geissel (Reference Guasti and Geissel2021) have demonstrated that elected and self-selected citizen committees and assemblies in Germany make representative claims based on claimed constituencies and their interests. This arises from the fact that contemplating the interests of others pertains to a particular constituency (Guasti and Geissel, Reference Guasti and Geissel2019). These constituencies’ interests may form part of a claim of self-appointed, non-electoral representation voiced by participants (Montanaro, Reference Montanaro2017). When they claim to represent someone else, participants refer precisely to these constituencies (Guasti and Geissel, Reference Guasti and Geissel2019). Thus, representative claims can be the product or consequence of a considered interest. At the same time, interests can only be precursors or elements of fully-fledged representative claims (Guasti and Geissel, Reference Guasti and Geissel2021). Other-regarding interests might form the basis of the claimed constituency as an object of representation of a potential representative claim (Guasti and Geissel, Reference Guasti and Geissel2019). In this way, considered interests can help generate citizen representatives (Warren and Pearse, Reference Warren and Pearse2008).
Interest transformation was found to strengthen the individual deliberative capacity of minipublic participants (Dryzek, Reference Dryzek2009). Mansbridge et al. (Reference Mansbridge2010) argue that participants in a deliberation arrive at first primed with their self-interest. Exposure to alternative viewpoints ideally enables them to transform both their initial judgments as well as their interests by including and processing a range of alternative viewpoints. This process results in multiplier and self-realization effects, occurring by successfully transforming participants’ identities. Although these effects occur at the individual level, their collective impact manifests as a systemic influence. This is particularly the case when deliberating peers bridge the gap between the broader citizenry and experts, who in turn can refine their deliberative skills through interaction with minipublics (Dryzek and Niemeyer, Reference Dryzek and Niemeyer2019). These normative functions of considered interests help delineate the democratizing potential of interest consideration within minipublics.
However, interest transformation does not automatically happen in deliberation, but depends on the characteristics and purpose of a given minipublic. Notably, we know little about what participatory design (e.g., size, selection, issue, process) contributes to this transformation. For example, Chong et al. (Reference Chong2001) found that individuals are more likely to rely on self-interest when an issue has a tangible impact on their personal life (e.g., through redistribution effects) and when that impact has been made clear by previous communication. Especially, forums handling issues like climate change or social housing can highlight this connection to citizens’ daily lives (i.e., gas prices and rents), eliciting a wider range of other-regarding interests among participants.
A direct link with citizens can also be created by clearly defining and explaining the purpose of an institutional connection (El-Wakil and McKay, Reference El-Wakil and McKay2020). For example, this happens when a minipublic is tied to a public vote. Parkinson (Reference Parkinson2020) outlines further conditions under which popular votes give leeway for deliberative practices. The latter must be simultaneously consequential regarding a popular vote’s outcome, plugged into a narrative and detect foul arguments in public discourse. This way, they can amplify argumentative quality by linking campaign claims with reasons present in the discourse.
Ideally, a process would meet these criteria: Coupling a citizens’ initiative and voters’ deliberation by circulating the minipublic’s output to eligible voters (Már and Gastil, Reference Már and Gastil2023). This output could contain a summary of key facts and arguments on the upcoming vote. This procedure would hand voters independently vetted arguments instead of unverified claims. The statement could influence citizens’ voting behavior and increase their initiative-relevant knowledge (Gastil et al., Reference Gastil2018). Two mechanisms explain why (Parkinson, Reference Parkinson2003): (1) the presence of a protected space enabling a respectful discourse, and (2) the institutional accountability created by a publicly available summary statement sent out before the vote. This accountability would contribute to participants’ interest inclusiveness by equipping them with a sense of public responsibility.
Hypothesis 1: Enhancing the purpose of the process (i.e., coupling a citizens’ jury to a consequential popular vote) increases the likelihood of interest inclusiveness.
Furthermore, the quality of the deliberative process likely influences interest inclusiveness. Scholars investigated the impact of discursive quality on participants’ opinion change (Fishkin, Reference Fishkin2009), political attitudes (Grönlund et al., Reference Grönlund2010), empathy (Muradova, Reference Muradova2021), and emotional transformation (Johnson et al., Reference Johnson2017). They show that opinion change is, on the one hand, an outcome of successful deliberation, understood as the weighing of reasons (Luskin et al., Reference Luskin2022). On the other hand, the presence of additional discursive elements and facilitation also increases the likelihood of such a transformation happening (Afsahi, Reference Afsahi2021). Such conducive factors include, but are not limited to, being treated with respect by the other participants and facilitators and having fewer problems following and comprehending the discussion. Interest consideration should hence also become more likely in the presence of increased discursive quality, not solely by weighing interests (Mansbridge et al., Reference Mansbridge2010). We thus expect an improvement in participants’ subjectively experienced discursive quality to exert an exogenous influence on interest consideration. As improved ratings of discursive quality have been found to promote participants’ opinion change, we similarly expect it to lead to a higher likelihood of interest inclusiveness.
Hypothesis 2: A higher perceived discursive quality of the minipublic’s process increases the likelihood of interest inclusiveness.
Finally, we examine the characteristics of participants and how these impact the inclusiveness of interests. Past research revealed that disadvantaged groups engage less in institutional formats of political participation, including voting, standing as a candidate, and participating in campaigns (Bailer et al., Reference Bailer2022; Paxton et al., Reference Paxton2007). For example, disadvantaged groups (i.e., women, younger individuals, and individuals with lower levels of income and education) are likelier excluded from discursive participation (Beauvais, Reference Beauvais2021; Mansbridge, Reference Mansbridge1983) and likelier exposed to inequalities in discussion time during political discussions (Beauvais, Reference Beauvais2020; Mendelberg et al., Reference Mendelberg2014). Simultaneously, political representatives with a disadvantaged background are more likely to convey interests in policy areas affecting their rights and autonomy (Lovenduski and Norris, Reference Lovenduski and Norris2003; Mansbridge, Reference Mansbridge1999). With the group composition of minipublics representing significant parts of the disadvantaged population, these groups were found to possess more frequent opportunities and time to voice their perspectives (Flynn and Sood, Reference Flynn, Sood and Grönlund2014), including those who participate less in elections. Deliberative and facilitated discussions increase this likelihood even more, furthering women’s discursive participation (Afsahi, Reference Afsahi2022; Field, Reference Field2022). This expectation is evidenced by studies of rhetorical sensitivity that also posit that, especially the disadvantaged, will develop more fine-grained social skills in discussions (Hart and Burks, Reference Hart and Burks1972), allowing them to take the perspectives of others. Note that the inclusion of the interests of disadvantaged groups in democracy as a marker for societal progress has been a central concern for theorists of deliberative democracy for decades (Bächtiger, Reference Bächtiger and Bächtiger2018; Chambers, Reference Chambers2009; Young, Reference Young1990).
In the measure introduced later, we do not present the precise list of Mansbridge’s (Reference Mansbridge1999) definition of disadvantaged groups (e.g., women and people of color) for two reasons. First, equipping participants with this exact set of options possibly creates contradictions between more or less objective and subjective views of which group counts as disadvantaged. Second, priming any of these groups could activate inherent stereotypes or prejudices participants carry. By contrast, the selected options presented later possess a clear scope for most jurisdictions and thus leave the interpretation of disadvantaged groups to what participants perceive society or a vote’s proponents or opponents suggest.
Hypothesis 3: Participants belonging to disadvantaged groups are more likely to convey considered interests than their more advantaged counterparts, using the following markers: (a) gender, (b) age, (c) income, (d) education, (e) previous political participation.
Empirical strategy
We collected participant data of five instances of Swiss in-person deliberative minipublics between 2019 and April 2023, following a single-country study approach (Pepinsky, Reference Pepinsky2019). We analyzed the pooled data totaling 343 responses nested within 110 participants in three local and two cantonal minipublics (see Online Appendices A and B for items and summary statistics). Crucially, we include almost the entire Swiss population of these minipublic participants within that period, regardless of the procedure used in each case, thereby enhancing representativeness and addressing concerns about statistical power when generalizing the results.
Studying Switzerland is advantageous for several reasons. First, the Swiss context serves as a most-likely case for the formation of considered interests due to its, on average, conducive institutional architecture. As a consensus democracy, its political system prioritizes powersharing and compromise over majoritarian conflict. This culture is reinforced by direct-democratic instruments, such as popular initiatives and referendums, which necessitate frequent, high-stakes public reasoning (Sciarini, Reference Sciarini, Morel and Qvortrup2018). These institutions do not merely exist in parallel to deliberation; they actively shape the cognitive environment of the citizenry. The constant exposure to policy-related arguments leading up to popular votes promotes a high baseline of political literacy and encourages individuals to weigh a range of competing interests before casting a ballot (Colombo, Reference Colombo2018). While broader public discourses may introduce heuristic biases, the shared design of Swiss minipublics involving stratified random selection aims to mitigate these by ensuring a demographic and ideological cross-section.
Focusing on this ‘easy case’ provides a rigorous test: if empirical evidence of other- regarding interest is absent here, which results will later show is not the case, it is unlikely to emerge in more adversarial or majoritarian contexts. However, the high degree of Swiss elite cohesion often found in opposition to citizen-led initiatives introduces a counterweight: The federal executive and legislative chambers frequently position themselves against citizens’ initiatives to defend the status quo (Kriesi, Reference Kriesi2006). This external influence might concentrate like-minded perspectives within minipublics, potentially narrowing the multitude of available arguments even within a consensus-oriented system.
The five cases comprised all advisory citizens’ juries, with participant numbers ranging from twenty (Citizens’ Initiative Review (CIR) Sion, CIR Geneva, Citizen panel Uster, CIR Bellinzona) to thirty (Citizens’ Forum Geneva). Each minipublic followed a consistent three-stage process: (i) random selection, (ii) deliberation, and (iii) a concluding statement presented to either the political executive or the wider public, with a roughly equal split between learning and deliberation phases. All events lasted two weekends except the Citizens’ Forum, which took four weekends to complete. This standardized structure facilitates a representative comparison across Switzerland’s three main linguistic regions – German-, French-, and Italian-speaking communities. By reducing variation in key areas such as sortition, size, and core elements of the deliberative process, the setup ensured meaningful comparisons. These cases also provided diversity in topics (climate change, housing, and pension payments), engagement duration (at least four full days), advisory purpose (targeting policymakers, the public, or voters in a referendum), and comparable levels of the political system (cantonal or local).
The CIR Sion (November 2019) involved twenty participants evaluating a federal initiative to extend affordable housing policies, with a summary statement sent to all voters before the vote. Similarly, the CIR Geneva (September 2021) had twenty-one participants assessing a cantonal initiative to abolish pensions for politicians, with a summary statement sent to eligible voters. Lastly, the CIR Bellinzona (Spring 2023) involved nineteen participants discussing a referendum on the CO2 law aimed at reducing emissions; their summary statement was also distributed to voters. This case adds the missing combination between CIR processes and climate change, disentangling issues and purpose. If climate change is a transversal issue, interest inclusiveness should be consistent across all cases discussing it. However, the three CIRs revealed significant variations in interest inclusiveness across climate change cases. More detailed information on each case is available in Appendix D. The Citizens’ Forum Geneva (September 2020–June 2021) consisted of thirty participants who, over four weekends, developed recommendations for improving communal life and addressing climate change in Geneva; these were presented in a report to local policymakers without broader public dissemination. The Citizen Panel Uster (August–September 2021) engaged twenty-one participants in creating recommendations on climate change and waste reduction, which were shared with both the city council and all local households.
The first hypothesis will be tested by comparing the interest inclusiveness of participants in the three cases involving citizens’ initiatives with those in the two cases that did not, using fixed effects. This comparison helps assess whether connecting deliberative practices and direct democracy affects interest inclusiveness, based on the CIR (Gastil et al., Reference Gastil2023). The CIR follows the criteria outlined by Parkinson (Reference Parkinson2020), which specify that popular votes can create space for deliberative practices if these processes are consequential for voting outcomes, embedded within a broader narrative, and capable of identifying flawed arguments in public discourse. By linking campaign claims with reasoned arguments, the CIR might enhance discursive quality at a systemic level (Curato and Böker, Reference Curato and Böker2016). It integrates direct and deliberative democracy by providing eligible voters with a summary statement containing independently vetted facts and arguments about an upcoming citizen-initiated vote (Már and Gastil, Reference Már and Gastil2023). This approach ensures voters receive validated information instead of unverified claims, thereby influencing voting behavior and improving initiative-relevant knowledge (Gastil et al., Reference Gastil2018).
Methods
We employ a mixed-methods approach that triangulates a survey measure of interest inclusiveness with automated transcripts from the deliberation of the CIR Sion case to study the scope of other-regarding interests and the context in which participants express them (see Online Appendix E). For each of the cases, participants completed self-evaluation questionnaires either at the end of each day (CIR Sion, Geneva) or weekend (Citizen panel Uster, Citizens’ Forum Geneva, CIR Bellinzona) of deliberation. Participants who were unable to complete the paper questionnaire at the end of the weekend deliberations received a link to the same online survey by e-mail, which they completed until two days later.
This study measures interests evoked by deliberation, which inherently requires deliberation to assess these interests. A key limitation is the inability to account for preexisting other-regarding preferences on the topics prior to deliberation. To address this, we use repeated measures, enabling us to examine how individuals’ considered interests evolve from a shared baseline established on day one to the end of the process. This approach also allows us to assess the measure’s test-retest reliability.
We operationalize interest inclusiveness using a self-report measure that captures interests directly after each meaningful period of deliberation. For the initial case, the question read: ‘When forming your judgment on the popular initiative today, have you taken the interests of one or more of the following groups into consideration? My judgment was based on… (1) My own interests; (2) The interests of the inhabitants of the City of Sion; (3) The interests of the inhabitants of the Canton of Valais; (4) The interests of the inhabitants of Switzerland; (5) The interests of disadvantaged groups in society according to the proponents; (6) The interests of disadvantaged groups in society according to the opponents; (7) Other(s)’.
The question design was based on the literature on non-electoral representation, assuming that participants conveying interests mirror to some degree a non-electoral representative (Rehfeld, Reference Rehfeld2018). This role hands them three elements to consider (Rehfeld, Reference Rehfeld2009): (1) the legislative objective, (2) the nature of judgment, and (3) the degree to which they are sensitive to consequences. Among these, the legislative objective is most relevant for how participants consider other-regarding interests. This criterion distinguishes between promoting the greater good, often aligned with national interests, and advocating for the interests of a specific constituency. For example, participants might refer to interests of groups of a particular jurisdiction (e.g., locals or the nation’s inhabitants). We adopt the distinction between the whole (national interests) and the part (specific group interests, such as the local or regional constituency) in measuring these considerations (Rehfeld, Reference Rehfeld2011). Moreover, using a rationale to offer options allows us to compare a homogeneous set of interests that is likely important to the deliberation.
The options were presented in randomized order, with the residual option placed at the bottom. Semi-open-ended question formats are said to yield valid results in the Swiss context (Zollinger, Reference Zollinger2024). Online Appendix C lists the wording for the remaining cases, in which we adapted the options to match the topic of the minipublic. If certain groups were essential to its core question, we added further options, that is, the interests of the economy or neighboring regions. Crucially, we adjust for the content and number of response options among the juries. Space limits constrained the question’s ability to exhaust all possible interests that participants could consider. Thus, a residual option allowed respondents to specify additional groups or concepts they perceived to have considered. These supplementary statements indicated that participants understood the question and engaged with it. Additional considerations were matched to appropriate categories, if possible, otherwise counted as separate interests. In the Uster case, two more residual options probed the measure’s validity, labeled ‘Considered no interests at all’ and ‘Tried to consider all interests equally’. These remained unused throughout, underscoring participants’ comprehension.
The question format supports capturing multiple interest objects explicitly and independently from one another. When self-interest and the interest of a disadvantaged group overlap for some respondents, they may check having considered both self-interest and the interests of that particular group. Validation results presented next support that participants differentiate between their self-interest and the interests of disadvantaged groups. We do not try to objectively determine the interest consideration of a given participant.
Moreover, the question contains important details that enhance its comprehension and the measurement of interest inclusiveness. The possessive pronoun in the phrase ‘my own interests’ clarifies to respondents that personal material self-interest as opposed to sociotropic interest is at stake (Schaffer and Spilker, Reference Schaffer and Spilker2019). The measure thus clearly distinguishes between egocentric interests and sociotropic interests irrespective of overlaps between participants’ self-interest and their objective or subjective belonging to one of the other options listed (Mansbridge et al., Reference Mansbridge2010). We also avoid naming any specific partisan groups, except for disadvantaged groups in society or as framed by initiative proponents or opposers. The term ‘disadvantaged’). We also avoid naming any specific partisan groups, except for disadvantaged groups in society or as framed by initiative proponents or opposers. The term ‘disadvantaged’ can thus capture a wide range of individuals or subgroups. The options offer mainly entities fitting the political-administrative jurisdictions of the whole and the part. Hence, they accommodate two dynamics during the deliberation: the content of the presentations (and subsequent Q&As with stakeholders and experts) and the participants’ interpretations and narratives (Parkinson, Reference Parkinson2003).
Our measure of considered interests differs in three ways from how the established index based on coding actual speech acts, the Discourse Quality Index (DQI) conceptualizes interests (Steiner et al., Reference Steiner2004): First, our measure is self-assessed and isolates interests as a distinct analytical category rather than assessing them against the common good. The DQI, on the other hand, scores the ‘content of justification’ indicator higher when participants shift from narrower group interests toward justifications of the common good. However, as Steiner et al. note (58), these orientations can and in fact do often coexist. By focusing specifically on isolating interests, we avoid conflating a participant’s potentially strategic rhetoric with their underlying interest formation. This distinction matters because self-interest and broader group interests are not always identical.
Second, our approach references a participant’s own interests, a dimension not separately captured by the DQI’s notion of referencing narrower (group) interests. As results will show later, including a deliberator’s own interest enables observers to assess whether individual considerations are maintained or modified. Third, the item references other-regarding interests likely to emerge during the discussion rather than locating a speaker along levels of common good versus interest. This way, the instrument records context-dependent concerns.
Measuring interest inclusiveness
The question on considered interests introduced in the previous section allowed for the coding of two measures, representing the number and proportion of inclusive interests, referred to as the continuous outcome I and binary outcome II. With these measures, we tap into both the quantity and extent of participants’ consideration of others’ interests.
-
Continuous outcome (I):The number of considered interests checked, other than self- interest. For subsequent analyses, this count was rescaled to a continuous variable to provide a uniform seven-point scale across all cases.
-
Binary outcome (II): The proportion of considered interests, coded as (1) if participants checked at least one other-regarding interest and left self-interest non-selected, and as (0) if they selected self-interest only.
Outcome I represents the total number of other-regarding interests considered by participants. Outcome II indicates the proportion (and later, through logistic regression, likelihood) of interest inclusiveness by coding whether participants exclusively took into account interests beyond their own. This distinction is crucial as it delineates the consideration of the number and proportion of other-regarding interests besides self-interest.
Outcome II serves as the primary dependent variable in the subsequent analyses. Drawing on our main conceptual source, Mansbridge et al. (Reference Mansbridge2010), which sets a high standard for interest transformation, we chose to focus on the measure requiring participants to have transformed their interests only if they ‘forego’ their self-interest in favor of other-regarding interests. This strict condition is reflected in the binary coding. For all observations, the proportion of considered interests with foregone self-interest is 56%. Figure 1 and Table A2 of Online Appendix A visualize the distribution of each outcome. Interest consideration was measured two or four times on each available day or weekend. In the two events with four measurement days, there was a longer gap between the second and third day.
Average proportion of interest inclusiveness relative to all instances of considered interests (right-hand side panels) and number of other-regarding interests considered other than self-interest (left-hand side panels) at the end of each day/weekend of each case.
Figure 1. Long description
Panel A: The line graph shows the number of interests considered over four days or weekends for three different cases: CIR Sion 2019, Citizens' Forum Geneva 2021, and CIR Geneva 2021. The x-axis represents the days or weekends, and the y-axis represents the number of interests. The lines are color-coded and use different symbols to represent each case. Panel B: The line graph shows the proportion of interests considered over four days or weekends for the same three cases. The x-axis represents the days or weekends, and the y-axis represents the proportion of interests. The lines are color-coded and use different symbols to represent each case. Panel C: The line graph shows the number of interests considered over two days or weekends for two different cases: Citizen Panel Uster 2021 and CIR Bellinzona 2023. The x-axis represents the days or weekends, and the y-axis represents the number of interests. The lines are color-coded and use different symbols to represent each case. Panel D: The line graph shows the proportion of interests considered over two days or weekends for the same two cases. The x-axis represents the days or weekends, and the y-axis represents the proportion of interests. The lines are color-coded and use different symbols to represent each case.
The approach to measurement stems from ambiguity in the literature about what constitutes evidence of genuine inclusion and consideration of interests. Cappella et al. (Reference Cappella2002) require individuals to recite divergent arguments, but interests are more consistent than issue-specific opinions that might change. Furthermore, reciting arguments within groups is subject to hindsight bias (Stahlberg et al., Reference Stahlberg1995): When group members recall or discuss arguments, they may overestimate the predictability or inevitability of an argument, skewing their judgment and reinforcing collective viewpoints. Measuring the perceived deliberative quality of the discussions in parallel adjusts for the potential post hoc rationalization and social desirability bias. If interest consideration was triggered by a good discussion rather than by genuine interest consideration, we would observe extremely strong positive associations to the point of multicollinearity between the two measures. Therefore, the study’s focus is on the binary measure indicating that interest inclusiveness, that is, advancing interests other than one’s self-interest, has occurred in the eyes of the participant.
We follow the best practices of construct validation (Adcock, Reference Adcock2001; Flake et al., Reference Flake2017), supporting the minimal measurement approach toward considered interests in a self-administered questionnaire. Our three-part validation strategy demonstrates the measure’s reliability and validity through: (a) substantive validity, including linguistic validity across three languages and content relevance assessed with word-embedding techniques (Online Appendix E1); (b) structural validity, using test-retest reliability (alpha = 0.7–0.8 for the continuous outcome, mean rho = 0.3 for the binary outcome, monotonicity, and Mokken scale analysis in Online Appendix E2); and (c) external (predictive) validity, evaluated through hierarchical and panel regressions (Online Appendix E3). Together, these tests indicate that the longitudinal self-report measure used to operationalize interest inclusiveness effectively captures participants’ interests immediately following each meaningful period of deliberation.
A concern for the measure’s reliability is that participants may prefer appearing as motivated or think more other-regarding interests are more socially desirable (see Elster, Reference Elster and Elster1998). Therefore, they might check more interests than they have considered, exacerbating measurement error. However, the additional response option offered in the Uster Citizen Panel, that is, ‘Tried to consider all interests equally’, remained completely unused. This finding indicated no substantial social desirability bias. Participants possessed an understanding of the depth of the question, enhancing content validity.
A related limitation of the multiple-choice format is that it cannot capture the compelling arguments participants might have articulated regarding their interests. Therefore, the textual analysis of transcripts detailed in Online Appendix E1 helps contextualize how participants use the response options provided by the measure when making claims. In addition, we ran a pretest several months prior to the first case among a stratified random sample of twenty undergraduate political science students from a Swiss politics lecture. They participated in a pilot citizens’ jury on a Swiss federal referendum addressing the adoption of the EU’s changes in private arms regulations.
Covariates and estimation
The regression models adjusted for differences in the outcome between each case of a minipublic. This adjustment permits estimating the baseline differences between each case for H1. For the remaining hypotheses, we added the perceived discursive quality for H2 as an unweighted additive index of four randomized 11-point items adapted from Gastil et al. (Reference Gastil2015) (reliability Rk = 0.8, see Shrout and Lane (Reference Shrout, Lane, Matthias and Tamlin2014)), labeled from ‘Never’ to ‘Almost Always’: ‘During today’s [this weekend’s] discussions, when other participants expressed opinions different from yours, did you listen carefully to what they had to say? (1), did you feel that other participants treated you with respect? (2), did you have difficulty understanding or following the discussion? (3 – reversed), did you feel pressured to agree with something you weren’t sure about? (4 – reversed)’ As an equally reliable alternative, the perceived DQI (pDQI) is often used in empirical studies involving deliberative forums, presenting a similar set of items for self-assessed discursive quality (Caluwaerts and Deschouwer, Reference Caluwaerts and Deschouwer2014). However, since three out of five cases in this study followed the CIR’s process, this study used the items developed for it by Gastil et al. (Reference Gastil2015).
For the disadvantaged groups in H3, we included participants’ gender (binary), education (5-point), age (continuous), monthly net household income (5-point), left-right self- placement as a proxy for ideology (11-point), and vote frequency (5-point). These data were available from the initial random selection and surveys conducted before the deliberation. In addition, we added quadratic effects for age and left-right placement to account for the possibility of age effects and less inclusive interests among the ideological extremes. We also adjusted for the number of checked interests separately, because considering other-regarding interests might be more likely when the number of considered interests checked in the question increases (coefficient ‘No. of interests’ in Table 1).
The data set comprises variation within and between participants, along with the clustering of citizens’ juries. To account for the nested data structure, we employed linear and logistic hierarchical and panel models (Snijders and Bosker, Reference Snijders and Bosker2012). We avoid cross-lagged panel models because the cumulative effects are underidentified in two out of five of the cases and, therefore, cannot be consistently estimated in two-wave panels (Lucas, Reference Lucas2023).
The results, presented next, were based on imputing missing observations of income, ideology, vote frequency, and discursive quality, using single imputation with predicted values from linear regressions from the samples of positive replies to the initial invitations to the juries. In addition, we replicated the models based on conditional multiple imputation with identical conclusions, see Online Appendix F.
Results
We investigated whether members of a minipublic progressively consider a broader range of interests over the course of their participation, employing both descriptive and explanatory approaches. In line with the literature on interest transformation, minipublic participants considered interests different from their own for both outcomes, though the extent varied considerably between cases. Figure 1 shows both the proportion of interest inclusiveness and the number of interests considered other than one’s self-interest. A heterogeneous pattern indicates two to three interests, on average, that participants took into account other than self-interest. However, tracking the proportion of interest consideration over time suggests that it was only slightly more likely to occur as each event progressed. Not all participants became uniformly more likely to consider others’ interests, either. What drove the consideration of other interests? To explore this, we modeled process and individual differences using regression analysis.
First, the hierarchical logit model (1) shown in Table 1 regressed the binary measure of interest inclusiveness on the fixed effects relative to the first measurement of each series. In the second step of the regression analysis (models 2–4), we tested the hypotheses H1–H3 regarding the factors that influence changes in interest consideration. We used the same binary outcome and estimation. Model (3) presents the main results without adjusting for the number of interests that respondents checked, while model (4) adjusts for the full set of available controls. We obtained identical results to the ones presented here using random-effects panel models in Online Appendix F, after an insignificant Hausman test χ2(5) = 0.78 (p = 0.98) indicated that random effects were appropriate. Further robustness checks presented there using the continuous outcome supported the findings.
Hierarchical regression of outcome II (binary interest inclusiveness measure)
Table 1. Long description
The table presents hierarchical regression results for a binary interest inclusiveness measure. It has four columns and multiple rows, including an intercept and various factors. Column headers are (1), (2), (3), and (4). Row labels include Intercept, 2nd session, 3rd session, 4th session, Gender: Female, Age, Education, LR Ideology, Vote frequency, Income, LR Ideology^2, Age^2, Discursive quality, Minipublic (Non-CIR GE), Minipublic (Non-CIR UT), Minipublic (CIR GE), Minipublic (CIR BZ), No. of interests, SD (Intercept id), Num. Obs., R2 Marg., R2 Cond., AIC, ICC, and RMSE. Each row contains values corresponding to the columns. Notable trends include variations in coefficients and standard errors across different models and factors.
a Ref. = 1st session.
b Ref. = Male.
c Ref. = Minipublic (CIR Sion).
Note: p + = 0.1, *= 0.05, **= 0.01, ***= 0.001. Standardized logit coefficients for non-binary coefficients. GE – Geneva, UT – Uster, BZ – Bellinzona. Cluster-corrected SE in parentheses. N(responses) = 343, N(participants) = 110. Dep.var: Binary measure of interest inclusiveness: 1 = Considered at least one other-regarding interest, leaving self-interest unchecked.
Model (1) indicated that no statistically significant transformation of interests happened in terms of exclusively considering others’ interests as the minipublics progressed. The models’ intra-class correlations and conditional R² for random effects showed the sizeable variation between the respondents, backed by the low conditional R² for the fixed effects in model (1). Within the sample only, the results indicate an overall sizable (23%) increase in the likelihood of including other interests after an initial dip and recovery throughout the process. Following the initial reduction in considering others’ interests, they became more considered toward the last session compared to the first, and participants were more likely to take stances different from their own, focusing less on self-interest. A trend model where fixed effects were replaced with a continuous duration variable yielded an insignificant 6.3% increase per session of the likelihood to consider other-regarding interests.
In support of H1, the two coefficients indicating the relative difference in the likelihood of interest inclusiveness occurring between the Sion CIR reference category and the two non-CIR minipublics attain significance in model (4) of Table 1, from which all following results stem. The CIR reference case heavily boosted interest inclusiveness beyond the sample of participant responses, making interest consideration eight times (Citizens’ Forum Geneva) and four times (Citizen Panel Uster) less likely for the other types of minipublics not following the CIR model. This pattern held for the Bellinzona CIR, where the likelihood to consider others’ interests was, on average, very much higher than in Sion. This result was likely due to combining a consequential popular vote using the CIR model with climate change as an issue that elicits various interests. Since all cases shared a roughly equal split between learning and deliberation phases, the CIR’s format and connection to a popular vote might be responsible for its positive association with interest consideration, showing an increased probability compared to the other two cases.
Hypothesis 2 about perceived discursive quality was not supported, only its direction within the sample was consistent with expectations, that is, each unit along the eleven-point scale of perceived discursive quality was associated with a roughly 1.3-times increased likelihood of considering others’ interests, all else equal. However, perceived discursive quality was significantly associated with consideredness only using the continuous outcome reflecting the quantity of considered interests (see Online Appendix F, Table F3).
The results for disadvantaged group status (H3) did not indicate that any of the markers of disadvantaged status except a lower income attained significance, almost entirely disproving this hypothesis. Only within the sample, women were indeed more than 1.6 times as likely as men to consider others’ interests, as stated in H3(a). We observed a similar and significant effect size for scoring a lower income category, reflecting support for H3(c). A higher monthly income reduced interest inclusiveness considerably, all else equal. Age H3(b), education H3(d), and previous participation H3(e) did not align with the expected directions of the hypotheses. On the contrary, the main coefficient for a SD-unit increase in age (OR = 15.5, p = 0.031) indicated that older participants were significantly more likely to consider others’ interests as they have encountered various preferences throughout their lives. Note that this finding only holds for the likelihood of consideredness but not the quantity of interests. Furthermore, these sociodemographic indicators frequently function as broad proxies that can mask more complex, underlying mechanisms driving interest formation. Consequently, these variables should be interpreted with caution, as they may obscure the nuanced attitudinal or structural dynamics that more precisely shape individual preferences and should be subject to future investigation.
Considering a higher number of interests was negatively associated with interest inclusiveness, that is, checking an additional other-regarding interest reduced the likelihood of considering these exclusively without also factoring in self-interest. The more considerations weighed, the likelier individuals compared these against personal interests because of the deliberative process. Although neither education, nor vote frequency or ideology were significant, results indicate that center-right ideology as a combination of the linear and quadratic terms boosted considered interests within the sample.
Besides the evidence for the measure’s validity gathered in Online Appendix E, we also conducted further robustness checks using the continuous outcome (number of other-regarding interests considered) in Online Appendix F, which support the substantial conclusions.
Discussion and future application
The primary research question investigated whether participation in deliberative minipublics (i.e., citizens’ juries) increases participants’ sociotropic orientations and explored the association between this shift and factors such as the procedural purpose of the minipublic, perceived deliberation quality, and participant demographics (i.e., socioeconomic status). First, we observed a substantial albeit non-significant interest transformation in all the cases studied, which aligns with theorists who argue that such transformation only occurs when a deliberative process is perceived as effective (Mansbridge et al., Reference Mansbridge2010). The coupling of a citizens’ jury (H1) to a popular vote and better discursive quality (H2) strengthened interest consideration. Crucially, it was not only the coupling to the larger public that promoted more considered interests. Instead, using the CIR’s output for guiding voters in a consequential popular vote motivated participants to look beyond what their individual characteristics predicted them to do. The results do not support the expectation of an unconditional increase in interest consideration among disadvantaged groups as stated in H3 (i.e., women, young people, and individuals with lower levels of education and participation). However, some demographic patterns were associated with interest inclusiveness: It was more likely to occur among women (a), and significantly more likely among older participants (b) and low-income earners (c). Other factors neither enabled nor prevented interest consideration: The findings regarding previous political participation and education purveyed that interest inclusiveness does not vary much as a function of these traits. Interest inclusiveness, though expected to be transformed throughout the process of the minipublic, appeared less influenced by demographics.
The results prompt an updated scholarly understanding of participants’ beliefs of whose interest they consider within minipublics in three key areas: First, minipublic participants perceive themselves to consider other group interests during the deliberation. The Swiss case instructs us to reconsider the conventional wisdom that participants have only their own thoughts and interests on their minds when engaging in these events. Only a minority of participants viewed self-interest as the exclusive principle guiding their judgment. Contrary to what social desirability bias or masked self-interest would predict, participants reported only a marginal inclination toward being entirely impartial, having no distinct interests, or focusing exclusively on general interests.
Second, we find these convictions are also associated with the three citizens’ juries following the CIR’s process design. Importantly, interest consideration is related to a well-structured deliberative process coupled to a popular vote and thus combines deliberative and direct democracy in a hybrid mode. Hence, under the extensive process of vetting a popular vote proposal and routing the final citizens’ summary statement to voters, the CIR’s process excels at providing incentives to participants to consider others’ interests.
Third, measurement matters when it comes to investigating whose interests participants take into account: Considered interests can be defined flexibly within the measure, but the options offered have to be fixed over time to allow for substantial conclusions. We proposed two synthetic measures based on a novel survey question: interest inclusiveness as the likelihood of considering other interests and the number of these interests. The latter indicates solely interests other than self-interest; the former references the presence of considering other group interests excluding self-interest.
This study cannot cover all aspects of considered interests. First, gaps remain between the conceptual foundation and the subsequently introduced measures: Scholars conceive of considered interests differently. The present focus is on other-regarding interests, but future work also needs to investigate ideal-regarding interests. Second, we present a single-country study with a related set of issues and the same type of democratic innovation to suppress confounders when analyzing interest transformation. This study, though covering most of the population of participants from 2019 to early 2023, operates on a small data set with 110 individuals from five events on a limited number of issues. Moreover, interest transformation as an outcome of deliberation is not bound to a specific type of democratic innovation. Future research should look into other countries, processes, and issues. Third, due to the real-life character of the events, it was impractical to introduce treatments that would have investigated more complex individual- or process-level drivers of interest consideration. As interest formation is a cognitive process (Mansbridge et al., Reference Mansbridge2010), survey experiments can be a useful tool to isolate other potential traits leading to considering other-regarding interests across a more diverse set of issues.
Researchers may easily adapt the measure to their specific cases and context by modifying the response options of relevant interests. To ensure comparability when examining multiple cases, these interests should be phrased in similar terms, matching each case’s context. The measure’s ease of application implies its importance for future endeavors to investigate further self- and other-regarding interests in deliberation. This way, research can examine various issues and types of settings, for example, deliberation between citizens and politicians. The measure is easy to field in principle and yields reliable longitudinal data. Another way of extending the results would be to also include the strength or relevance of the purported interests in the outcome.
Conclusion
The article investigated whether participants in deliberative minipublics become more sociotropic through deliberation and how their views on considering other-regarding interests evolve over time. Deliberative theory suggests that participants may either consider other- regarding interests or arrive mainly with their own interests at the table of a deliberative minipublic. We focused on participants in five Swiss minipublics from 2019 to early 2023. Citizens’ juries, as deliberative minipublics, are characterized by the inclusion of citizen representatives, which has repercussions for the concerns of groups. Participants consider group interests beyond self-interest in deliberation on housing, pensions, and climate change, particularly with a process model coupled with a popular vote. We examined whose interests participants considered based on their self-evaluations and textual analysis of minipublic transcripts. We fielded a novel survey item eliciting whether participants in these minipublics had based their judgments on a variety of interests, including the population at the local, regional, and national levels, or on the interests of disadvantaged groups. We contribute a mixed-methods investigation, including textual analysis of considered interests and their inclusiveness both within and between different instances of minipublics.
This article presents an updated understanding of interest consideration within minipublics. Treating participants’ perceptions as a fixed concept or assuming them to be constant, risks failing to lose track of their interests’ actual extent and inclusiveness. Rather, the results support imagining participants’ convictions as shifting (Saward, Reference Saward2014). These attitudes expand the established categories of actor-centered representation (Rehfeld, Reference Rehfeld2018). The normative desirability of participants going beyond the constraints of their self-interest to act as citizen representatives remains contestable. Our main findings point to the existence of individual and process traits conducive to a constrained and deliberative interest that acts as a stabilizing force in the give and take of reasons in the cases under investigation. These insights direct attention toward a more concise understanding of the relationship between letting participants find reliable arguments and the image of whose interests they consider in minipublics.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1755773926100605.
Data availability statement
The raw data and replication materials supporting the conclusions of this article are available in the OpenScienceFramework (OSF) repository: https://osf.io/dxf7m/.
Acknowledgments
Previous versions of this article were presented at the Centre for Deliberative Democracy and Global Governance (2020), the NextGDC Conference (2022), and the APSA Annual Meeting (2022). We would like to thank the participants for their valuable feedback and constructive insights. We would also like to thank Katherine Knobloch and John Gastil, as well as Healthy Democracy for providing us with the CIR’s evaluation questionnaires and more original materials. We are grateful to Camille Bedock, Andrew M. Engelhardt, Matthew Ryan and Pierre-Etienne Vandamme for their valuable comments on previous versions of the manuscript, and Victor Sanchez-Mazas and Nenad Stojanović for their suggestions concerning the survey item used to capture interests. Finally, we are deeply grateful to Tina Freyburg for her excellent comments and insightful suggestions.
Funding
This study was supported by a project grant from the Swiss National Science Foundation (no. 176760). The municipal councils of the cities of Sion and Uster, Switzerland, and the cantonal chancellery of Geneva provided financial assistance.
Competing interests
The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Ethics statement
The University of Geneva’s Ethics committee gave IRB approval. The cantonal chancellery and the department of urban development of Geneva, and the City Councils of Sion and Uster approved the project.
Open access
