Public deliberation and policy design

Abstract What role can deliberative democracy mechanisms perform in relation to policy design? This article reports findings from an experiment which was conducted to answer that question. The experiment brought together a random sample of the public who debated public policy questions online and face-to-face. Three key findings emerged. First, deliberative democracy mechanisms are different from more typical participatory policy design tools because they offer something more dynamic than a ‘snapshot’ of public opinion. Second, public deliberation can contribute evidence to policy analysis processes that is more considered because it emerges from a process of citizen sensemaking. Finally, unique forms of policy relevant data can be produced by analyzing the justifications that citizens use to convince others during policy deliberations. These findings suggest that allowing citizens to “talk policy” through deliberative democracy mechanisms can produce unique forms of policy knowledge.


Introduction
For several decades, deliberative democracy mechanisms have been proposed as a means of including citizens in policy debates and subsequently reducing the "democratic deficits" that are said to undermine many polities (Dryzek 2000;Thompson 2008 ). This can be achieved, deliberative democrats argue, by allowing citizens to debate and decide upon the policy decisions that affect them (Elster 1998). These principles play out in the everyday world of policy through a series of unique institutions, often described as "mini-publics," which governments' use to share decision making with their citizens. The sheer number of mechanisms that have been used in this regard renders a full description impossible here but three well-used mechanisms are citizen juries, deliberative polling and consensus conferences (Smith and Wales 2000;Niemeyer 2014; Set€ al€ a and Smith 2018).
Across the mass of research that examines these mechanisms two fundamental findings hold constant. First, when deliberation is done right it regularly changes the opinions of participants (Himmelroos and Christensen 2014;Fishkin 2009). In other words, citizens who deliberate in mini-publics' often listen to the arguments of others and change their mind when it comes to public policy (Dryzek and Lo 2015). This is known as preference transformation. Second, that policymakers remain reluctant to share their authority with citizens (Goodin and Dryzek 2006;Bell and Hindmoor 2009;Davidson and Stark 2011). This means that deliberative mechanisms often produce decisions with are internally robust but regularly ignored by the external actors required to implement them. This contradiction led us toward a simple research question: if we take away the decision-making function from a deliberative mechanism, does it still have value for a policy designer? In other words, is there value, from a policy design perspective, in simply observing how everyday citizens debate policy?
A large mass of literature relating to participatory modes of policymaking (Fung 2006;Taylor 2011), alternative sources of non-technical policy evidence (Fischer 2017) and consultative policy tools more generally (Fraussen et al. 2020) would suggest that the answer to the question above ought to be an affirmative yes. The value of including citizens in policy design is presented in all these works but the reasons given for this varies. For some, the inclusion of citizen voices is a means of ensuring accountability (Fischer 2016), others emphasize legitimacy through consultation (Wang and Wart 2007) and another strand of participation studies, more pertinent to this research, suggest that participation can deliver policy formulation data (Hisschem€ oller and Cuppen 2015) and that "laymen" [sic] perspectives can complement the overreliance on technical expertise found in the policy sciences (Bekkers et al. 2018).
However, deliberative democracy mechanisms are fundamentally different from the more typical participatory mechanisms that tend to be examined in the policy design literature. They demand, for example, that citizens communicate through principles which respect the common good, that they listen to expertise and the rationales of others. To be considered deliberative in this regard debate also ought to reflect a series of principles: respect for the common good, respect for arguments and counterarguments and self-reflection about one's own position in relation to those arguments, for example (Steenbergen et al. 2003). It is these unique characteristics that have not been properly examined in relation to policy design work. Whether or not these unique features have a utility to policy design processes is an empirical question and such an evaluation, to our knowledge, has not been conducted before. We, therefore, created an experiment to explore this issue.
What might deliberation contribute to policy design?
The experiment we conducted was designed to replicate the processes that a typical mini-public deliberation might follow. However, our concern was to then analyze data generated about the process and the outcomes of the deliberation to determine whether they might be relevant to key policy design tasks and dilemmas. This of course begs the question of what deliberation might offer the policy designer. At its broadest policy design can be defined as "the application of knowledge about policy means gained from experience and reason to the development and adoption of courses of action expected to attain desired goals" (Howlett 2017, 129). We can think initially about what we might expect from public deliberation in relation to this broad goal by slightly reformulating this statement. We might say that deliberation offers a means of producing citizen knowledge about policy means through a discursive process that encourages reasoning about the adoption of a course of policy action. While the application of citizen knowledge to policy design is a feature of all participatory mechanisms, it is the dynamic process of reasoning, involving interaction between different arguments and changes in the opinion of participants, that makes deliberation unique in terms of process and outcome.
We can, therefore, hypothesize about some potential benefits that can be generated from this dynamism. A first potential benefit relates to instrument choice, especially around communication. For example, nudges that rely on behavior changing communications or modes of governance that utilize forms of persuasion and the delivery of information are growing in popularity (John 2013). As a deliberative process is designed to produce arguments which change opinions and develop consensus through communication, they offer the chance to identify narratives that citizens find compelling in these regards. A second potential benefit relates to the generation of evidence about policy complexity. In any deliberation, different citizens debate policy through their own lived experiences. This produces information which is characterized by a form of interactive complexity (in a discursive sense). Citizens compare and contrast their experiences through debate in ways which mean a program or instrument will be evaluated from many different positions and, cumulatively, via associations to multiple other policies and contexts. What this means is that deliberation might offer the opportunity to better understand the complex linkages between policies as they play out amongst the diversity of a target user audience. This seems pertinent to those interested in, inter alia, policy mixes and bundles (Flanagan et al. 2011). Finally, we can also theorize that deliberation is a means of prospectively exploring unintended consequences that might act as variables that affect policy once it has left the design table. Deliberation, for example, can alert us to the emotions, ideology or cognitive biases that target users may have which don't correspond with rationalistic assumptions. Deliberation may also provide signals to the ways in which changes in public opinion might affect policy implementation or give an indication of how the politics of policy more generally might play out in the public arena as ideas are tested through argument and competing discourses.

Method
To explore the value that deliberative democracy might have for policy design, we created an experiment in which a random sample of citizens were invited to discuss four policy proposals. The experiment was conducted online and via face-to-face deliberations in Brisbane, Australia.

The policy proposals
Four proposals were presented to the sample as questions. These were drafted with two key features in mind. First, they are proposals which can be discussed easily in the sense that they relate to policies that could affect everyday lives, they require little technical knowledge to be understood, and sources can be found about them easily via online searches. Second, each proposal was designed to have a contestable nature in terms of tradeoffs that would need to be made between a variety of competing policy objectives. This feature was necessary as a means of encouraging a context in which contestation and preference transformation was possible in the first instance. The proposals were: 1. Should an emissions trading scheme be introduced? Australia does not currently have an emissions trading scheme (ETS). This question reflected our research design interests as there are many examples of carbon emissions trading schemes around the world, which are easy to research online. This is also a policy that urges consideration of a range of issues relating to energy, renewables, tax and social policy and invites discussion of tradeoffs with other objectives, most notably, economic goals. 2. Should the full-time working week be reduced to 30 h or less? Clearly, this is not only a proposal that involves us all, but it is also one which requires labor market policy to be considered in relation to a variety of other policy objectives relating to, for example, taxation, superannuation and inward investment. In all these regards, it was deemed a question that would encourage deliberation. 3. Should the cost of vehicle registration be based on how many kilometers the vehicle travels? One method of achieving climate change and pollution goals is to use taxation to reduce the purchase of certain vehicles and their use generally. This question was therefore designed to facilitate a discussion of whether a preexisting tax that currently affects everyone who owns a car ought to be recalibrated for environmental purposes. 4. Should companies be taxed if they replace people with robots? Automation tax, colloquially known as "robot tax," is a policy measure that can be designed toward the production of several benefits. It can operate as a disincentive that can slow the shift to automation in a way that can make the transition manageable. Counterarguments in opposition to this form of tax typically draw on the problems with limiting innovation and technology.

The sample
The questions were given to a random sample of the public to deliberate upon. The sample was chosen using the Vox Pop Labs (Vote Compass) Database. A random sample of 6000 people from the Greater Brisbane area in Australia were sent an email invitation by Vox Pop Labs inviting them to register to be part of the project, asking them whether they preferred to participate in online or face-to-face discussions, and whether they would support each of the four policy proposals (the choice of answers was "yes," "no," and "not sure"). A stratified random sample matching the basic demographics of the Greater Brisbane area for age, gender, housing tenure, disability, and cultural and linguistic diversity was then selected and offered places in the discussion of their choice. Eighty participants were originally registered but as is always the case in recruitment of this nature, the sample reduced (to 48 participants) and representativeness was consequently compromised. However, this was anticipated, and the intention of the research design was not to produce a form of generalizability through representativeness. Instead, the intention was to put as much variety into the deliberations as possible and, ultimately, to produce findings that are exploratory and hypothesis-generating rather than generalizable.

The deliberative process
The citizens were put into a two-step process.
Step one involved the whole sample participating in a series of webinars in which experts, who were co-opted from academia and think-tanks, discussed the advantages and disadvantages of the policy proposals. The purpose of the webinars was to give time-poor citizens easy access to research and arguments in a simplified form. The webinars were conducted via Zoom technology, which allowed the sample to ask questions of the speakers and discuss the issues with each other (albeit in a limited electronic "chat" medium). Once the webinar sessions were complete, the sample was split between those who would discuss online and those who would discuss face-to-face. Twenty-seven participated in the online discussions and sixteen in the face-to-face discussions. The online deliberation took place via Synthetron software, an online platform which places participants into overlapping virtual small groups to allow the anonymous sharing of ideas in writing. This process involves a real-time moderated discussion, which in this case lasted across two evening sessions. In this discussion, participants make statements, and these are ranked by others in terms of degree of agreement. The result is a clear picture of where participants agree and disagree, a series of statistics about that agreement and a transcript of the discussions which can be analyzed in more depth. A similar moderation script was used for each topic to allow comparison between the discussions. Participants were asked to reflect on the webinars: what arguments made sense to them and which did not; what arguments or policy options and issues that were missing from the webinars; what views from the deliberation added value to their thinking about the proposals.
The face-to-face deliberation was loosely based on a citizens-jury model (see Smith and Wales 2000). However, the webinars had to take the place of the more typical citizen-witness dynamic, purely because project funds could not cover the costs of experts in situ. In the sessions, the participants worked through each proposal, first in small groups and then as a plenary, with the support of three facilitators who each have knowledge of deliberative democracy principles and experience of working directly with the public in mini-public settings. At the end of each session the outcomes from the plenary were recorded in terms of support or rejection for a proposal and/or mediating proposals which presented alternatives to the proposed policy.

Analysis
The small group discussions and the plenary were recorded and transcribed verbatim. The transcripts were then subjected to a "light" content analysis using the Discourse Quality Index (DQI) categories (see Steenbergen et al. 2003), which allows for the analysis of a range of deliberative criteria within a speech act. The DQI criteria covers seven dimensions that reflect mainstream idealizations of deliberative democracy (Thompson 2008). These are open participation, the justification of assertions, consideration of the common good, respect for social groups, respect for participants, respect for counterarguments and preference transformation. However, the DQI was not used here to produce a content analysis, and associated quantitative measurements, but rather as a means of broadly identifying moments in the discussion that had a deliberative quality to them. Once identified in this way, these instances were analyzed in a qualitative sense and the most frequently used arguments were coded.
Both the online and the face-to-face participants completed pre and post surveys of the same nature. In both they were asked about their views in relation to the four proposals to ascertain whether preference transformation had taken place. In the second survey, they were also asked a series of questions about the process of deliberation that they experienced to explore the deliberative process itself as a series of outcomes.
Once our data was generated, we finally analyzed it relation to the challenges of policy design. Two findings stood out clearly in this regard.

Findings 1: preference transformation and policy design
Each of the debates that we staged resulted in preference transformation among the sample. In other words, citizens changed their opinion about policies after being exposed to expertise and different perspectives from other citizens. In some discussions, the degree of preference transformation was significant enough to shift opinion around completely, meaning that proposals that had been supported by a majority of participants pre-debate subsequently lacked approval after it (and vice versa). In other deliberations, the support for a policy remained constant but changes in preference were strong enough to deliver a clear message about aspects of the policy proposal.
Consider, as our most emphatic example of this finding, our car registration deliberation. A typical consultation on that question would have yielded results like those found in our pre-discussion surveys, which indicated that this was a policy that would be likely to generate support amongst these target users. In our pre-discussion survey of those who deliberated face-to-face, 75% of the sample said yes to a question that asked if car registration costs should be calculated on kilometers traveled with 25% saying no (and no one in the unsure category). However, after the deliberations 100% of the participants said no to the proposal. In the surveys completed by those who debated online, a similar shift was also notable but in a less dramatic way because of the existence of a large percentage of "not sure" participants. Nevertheless, the largest pre-discussion category on this question were yes's (with 52%), but this subsequently became the smallest category post-deliberation (23%). Indeed, after the online discussion, the majority category was no (46%) with a significant number unsure (31%).
What we have, therefore, is a discussion which led to the dramatic withdrawal of support from a policy proposal.
This finding is reinforced in the evidence from other deliberations. In relation to the working week deliberation, for example, a smaller amount of preference change made a big difference in terms of outcome by transforming the majority view from negative to positive. In the face-to-face pre-discussion survey 44% rejected the proposal and 38% said yes with the remainder unsure. However, after the deliberation these positions reversed with 50% of participants opting for yes and 43% rejecting. In the online deliberation, the same trend was also substantiated, and the same outcome materialized. However, the margin was slightly larger in the online deliberations as 50% again said yes but only 36% rejected. This was a reversal from the pre-discussion positions which had the no's as the largest category (with 48%) and the yes's sitting at 37%.
The emissions trading scheme (ETS) discussion, however, represents a departure from this pattern. In this deliberation, a majority of participants supported an ETS before and after the deliberation. However, examination of the preference transformation that took place revealed a dramatic decline in support for an ETS as participants moved themselves out of the yes category and into no and unsure boxes. After the face-to-face deliberation on this question 50% of those who deliberated changed from their pre-deliberation position and 63% of those who changed went from a yes or unsure to a no. In the online deliberation, a similar but slightly less dramatic trend was evident. After these deliberations, only 25% of participants changed their preferences. However, 71% of those switched from a pre-discussion yes to a post-discussion no or unsure. Thus, while the final consensus was still in support of emissions trading, the deliberations encouraged a negative shift in the degree of support behind it. The message being sent here is that this is a policy proposal that might gain approval but not without a significant political discussion in the public sphere.
The debate on an automation tax represents a deliberation in which a large percentage of unsure participants shifted into yes and no categories. In the face-to-face surveys, for example, the pre-discussion positions contained 38% who were unsure yet post-discussion this reduced to 6% and the no category grew to a significant majority of 75%. However, this was the only deliberation in which online and face-to-face opinion differed in terms of preference transformation. Prior to the online discussion 26% of the sample were unsure with the yes and no categories split precisely at 37% each. However, post-discussion the unsure category dropped to 7% and as consequence 52% of yes's became the majority.
In many ways, this is an ideal deliberation to finish this section on as it allows us to underscore that we are less concerned with outcomes and more interested in the fact that citizen attitudes changed in some way because of them. As we discussed above, this dynamic is well known among deliberative democracy researchers, but we believe it bears repeating to a practitioner audience because it casts doubt on the more typical consultative tools used in policy formulation. If those tools only take a snapshot of opinion in one moment in time they will only capture what we did in our pre-deliberation surveys. However, as we show above this is far from ideal as opinions change when citizens make sense of policy and experience deliberations about it. This would seem to be analogous to the real world. When policy leaves the design table and enters party political arenas, and when it gets implemented and begins to have effects, it can be exposed to the kind of "everyday" questions and debate that we observed in our deliberative processes. Consequently, preferences are likely to change as they did in our experiment. A public deliberation therefore offers an opportunity to observe and reflect on those kinds of changes prospectively.

Findings 2: citizen justifications and policy design
We now turn to a more detailed analysis of some of the justifications that influenced the preference transformation that we recorded. Our central claim in this section emerges from a series of findings that show that the justifications that change opinion in a deliberation offer significant insights for the policy designer. More simply put, there is value in examining the arguments that change opinions in a citizen-led deliberation as they provide a unique data source.
A few further methodological points are warranted here. In the face-to-face discussion, we assume that the most frequently aired justifications affected the changes in opinion we recorded above. Two characteristics in the evidence support this assumption. First, the content of those justifications aligns with the direction of change we see in terms of opinion. Second, we can point to specific evidence in the transcripts in which our participants openly acknowledge that their thinking has changed as a consequence of an argument that they have heard. These acknowledgements litter the transcripts indicating justifications that were influential. Similarly, in the online discussion, we have a very clear indication of what statements attracted support. When a participant makes a statement, the software we employed allows other participants to manually rank the degree to which they support it. When a statement generates support from a significant number within the sample the software designates it a "synthetron" and produces statistics about the degree of support it has enjoyed. We relay two specific statistics below, which are "reach" (the percentage of participants who actively scored a statement, which indicates the extent to which it was recognized by the sample) and "net agreement" (the percentage of participants who agreed with the statement).
Rather than work through the nuance of each discussion's detail, we have chosen here to simply record the strongest theme that we saw across every debate, which highlights our argument about the value of analyzing public "policy talk" vis-a-vis policy design.
This theme centers around what policy design scholars call the "policy mix" (Howlett and Rayner 2007). This is a term under which researchers have sought to understand the challenges involved in bundling multiple policy instruments into single policy programs (see Howlett 2017 for an overview). Many of the justifications that were influential in these deliberations spoke directly to these challenges via arguments that: (1) defined the policy proposals as blunt instruments that would not be effective if implemented in isolation from other policies; (2) identified a range of associated policies that would have to be rethought if the proposed policy was to be implemented; (3) suggested alternative policies that could be more effectively coordinated into preexisting policy landscapes. Ultimately, analysis of these justifications delivered evidence about how the target users of policy perceive the preexisting policy mixes that affect them and the potential effectiveness of any new instrument or program in relation to that mix.
A very good example of this is the discussion on vehicle taxation in which opinions dramatically turned against the policy proposal. The transcripts of the face-to-face deliberation clearly show how support withdrew for this policy via concerns about the lack of a supporting framework of associated instruments that would be required to make it work. Negative justifications in this regard were communicated in relation to the (in)capacity of the public transport system to assist reductions in car usage, problems in urban planning that meant Australian governments were unable to assist a breakaway from car reliance and fuel excise measures that already institutionalize a user pays policy for taxing kilometers traveled. Similarly, the need to think about transport policy represented one of the online discussion's strongest synthetrons here, which emerged from a statement that the proposal "would put pressure on governments to improve public transport" (92% reach and 30% agreement) and a swathe of lesser supported statements that called for improvements and subsidy in public transport.
These arguments about associated policies formed into a larger theme which was recorded in our coding of transcripts as an "equity" argument. This captured many claims that lower socio-economic groups, car owners living outside cities and those with car reliant occupations would all be unfairly disadvantaged by this policy. This meant that there was a widespread view that "the incentive becomes a punishment for people because of the nature of Australia or the nature of certain jobs" (Participant H) and that "it is not effective to say I am going to tax this bad behavior in the hope you will move to something better if they have nothing to move to. Then we are just taxing" (Participant D). In the online discussion, the equity code was also echoed through two statements that made it into the top five of the most agreed upon synthetrons. These were that "people who live in outer suburbs that are poorly serviced by public transport would be penalized. These people are often on lower incomes" (a reach of 88% and supported by 36%) and that any proposal "would need to ensure there were exemptions for disability. But that would not cover people who bought further away from their work because the housing was cheaper" (72% reach and 22% support). This was a statement that elaborated on another synthetron, which was indicative of the exchanges in this discussion, that simply stated that the policy "would punish those who have no choice over their car usage." Negative arguments of this nature culminated in a consensus developing in the faceto-face plenary session that articulated that car registration was a blunt policy instrument. Thus, it was rejected by one group because it required "a system of strategies and adjustments to deal with the equity issue and fit in with the current tax system" (Participant D) or because "it would need congestion pricings on top of it and a registration fee for heavier vehicles" (Participant E) and because it would need "lots of tweaks, discounts and adjustments to avoid damage, physical and practical, for people who live in the country, old people, disabled people, or whoever" (Participant H). This led to a contribution in the plenary from the whole group who argued that "you could design a better system than just a flat registration fee. It is a bit of a blunt instrument and … it ultimately wasn't workable for that reason." Consequently, two alternatives -"congestion charging" and "dynamic tolling"generated more support as better alternatives in this regard. The blunt instrument view was also present in the online discussion. Indeed, one of the most highly ranked synthetrons in this deliberation explicitly made the case that "this is a blunt instrument and that an array of responses will be needed" (82% Reach and 24% agreement) while others stated that they were "not convinced that it would be able to be implemented. Too complex" and that the deliberation was missing arguments about "how complex implementing this policy would be." Although the vehicle registration discussion is the best example of the capacity of citizens to analyze policy mix issues, the other deliberations we observed were also replete with similar arguments. Indeed a "blunt instrument" code was created in both the automation tax and the working week discussions. In the former that code came from a coalescence of views that suggested that the taxation system was not dynamic enough to keep pace with workforce changes, and that a range of shorter-term transition policies would represent a better mix, while in the latter, a range of arguments were recorded around a central view that a "one size fits all" approach to economic productivity was inadequate. This was evident in the transcripts of every small group discussion and the final plenary discussion, in the working week discussion and was best expressed by Participant E who, when reporting his small group's consensus to the plenary, made the point that "we eventually came to the view that we don't think its desirable to force a particular set number of hours for every person to work a week … we also thought it would be very, very difficult to enforce" (Participant E). However, an important sub-theme within this code reflected a view that implementing a single instrument based on a static number of hours "limits flexibility" in terms of the agency that employers and employees need in a twenty-first century economy. A second code related to the analysis of apposite policy mixes related to arguments about the way in which a reduced working week could reduce income. Consequently, the face-to-face deliberations spent some time discussing the associated policies required in an apposite policy mix. These included minimum wage, income tax, goods and services tax and policies that could potentially deliver a living wage (such as universal basic income). These views also represented the strongest synthetron in the online discussion (with a reach of 73% and a strong support among 59% of the sample), which was communicated in relation to a simple statement that "the income deliberation is critical … The impact depends on what assumptions one makes to income reduction as a result of hour reduction. Some people would be happy to reduce hours for less income whereas others would need to seek an additional job to earn income required."

Discussion and practice recommendations
What our experiment has shown is that deliberation changes opinion and that this takes place through a shared process of sensemaking. We, therefore, have two aspects to discuss and make recommendations about for those interested in the policy analysis-deliberation relationship.
The first is that the existence of preference transformation casts doubt on the evidence that snapshot consultation tools generate. In this experiment, exposure to expert information from different perspectives and a process of listening to others "talk policy" led to people who supported a proposal rejecting it completely and it also encouraged important changes in opinion about specific aspects of policy, both positive and negative, underneath the headline of a "yes" or "no" outcome generated via survey. We, therefore, argue that there is value in observing how policy arguments change via deliberation prior to the pursuit of a course of action simply because it can give a prospective insight into the dynamics of public opinion that might play a role in the politics of policy implementation. This leads to a specific recommendation that: Large-scale policy consultation ought to include a citizen-deliberation component even if decision makers do not wish to share decision making power with citizens. Data from a deliberative component can be triangulated with "snapshot" forms of consultation in order to give policy analysts a fuller understanding of public opinion.
Second, the specific justifications that change opinion in a citizen-led deliberation offer a source of data from target users about "what matters" to them. In all these discussions, discursive themes emerged which influenced how citizens perceived aspects of public policy. The analysis of these themes offers a window into a range of issues that are germane to policy design. We reported on one clear theme, which was significant across all our deliberations, which showed how these citizens thought about policy mixes and the (in)capacity of singular policy instruments to work effectively. This, we argue, is proof of the potential relevance of deliberation to policy designers. However, the discursive themes which influence discussions could represent an opportunity for policymakers to understand a variety of different areas of relevance to policy design. For example, in contentious policy areas, such as that involving climate change observing justifications that generate consensus across difference offers a valuable insight that could help anticipate and avoid policy failure. Calls have already been made for policy designers to get to grips with the more 'irrational' side of target user thinking that does not correspond with rationalistic models (Howlett 2019). Here again analysis of the sensemaking processes that citizens use to understand policy seems pertinent. Therefore, we recommend that: Deliberative mechanisms should be employed in policy design processes when analysts need data about policy mixes, prospective political problems and the potential ways in which public opinion might undermine the legitimacy of a policy instrument.

Conclusion
Citizen-led forms of deliberation in relation to policy can have value in democratic terms but all too often they are rejected because of the perceived dangers of sharing decision-making power. In recognition of this issue, this article asked a different question. Namely, is there value in simply letting citizens debate policy even if they are not given decision making authority? We conclude that there is value in this proposition, based on evidence that shows how analysis of citizen sensemaking and citizen justifications can deliver insights for policy designers. Those interested in thinking prospectively about the fate of policy once it leaves the design table would certainly benefit from a better understanding of the shifting nature of public opinion around policy, and the need to consider the citizen as an agent and not simply a target of policy. The evidence presented here suggests that deliberative mechanisms have the potential to help in both regards.