Deliberative public opinion

Introduction

The concept of public opinion is deeply flawed. Demonstrating these flaws does not require much unpacking of notions of both ‘the public’ and ‘opinions’ and epistemological problems are only compounded when the two terms are combined. However, there is a sense in which the value of some notion of public opinion can be reclaimed. I am referring here to collectively achieved positions that emerge from deliberative democratic engagement or what I shall term deliberative public opinion. Perhaps ironically, the grounds for the legitimacy of this concept can be found in the same logic that undermines the more common usage of ‘public opinion’: social constructionism. In this article, I draw on criticisms from social constructionist scholarship to demonstrate problems associated with many typical uses of ‘public opinion’. I then argue that as an explicit social construction, deliberative public opinion has political legitimacy, when constituted appropriately.

The context that informs my argument is the governance of emerging and/or controversial science and technology (S&T). Science and technology dramatically shape individual lives as well as society as a whole. Science and technology shape the possibilities of employment, health care, reproduction, education and communication. Science and technology also shape our subjective experience of the world (Schraube, 2009) and our psychological and cognitive capacities and responses to the world (Danziger, 2009). Critically, emerging science and technology raise important and often unforeseen ethical challenges and controversies, such as those surrounding stem cell research and genetic modification of crops and livestock (e.g. Jasanoff, 2005; Longstaff, Schuppli et al., 2009; Nep and O’Doherty, 2013). Because of the dramatic societal impact of these technologies, it has been argued that there is a democratic deficit in the governance of science and technology (Burgess and Tansey, 2008). There is, therefore, a need to have meaningful democratic representation in the governance of science or technology of the interests, values and perspectives of those affected or potentially affected by particular technologies and scientific developments. The concept of public opinion seems to have intuitive traction in this regard. After all, how else would we represent societal interests, values and perspectives, if not through the measurement of ‘public opinion’?

Public opinion

A rich scholarly literature traces the rise of public opinion research, its changing role in society over time and its contested role in reflecting or shaping social facts (Igo, 2007; Wilson, 2013[1962]). The surveying of publics for their opinion on a vast range of topics is arguably a key component of at least the illusion of public participation in the governance of social life. Criticisms of surveys as tools for meaningful public engagement are well developed. For example, in outlining a typology of public engagement mechanisms, Rowe and Frewer (2005) argue that surveys are limited to a one-directional flow of information from public representatives to sponsors (and initiators) of the survey. As a mechanism of public participation, therefore, surveys fail to provide opportunities for dialogue, shaping agendas and questioning the framing of issues. More to the point here, however, is the particular construct surveys and polls purport to measure: public opinion.

The notion of public opinion has wide currency in popular media and policy discussions. It is also a key term in scholarly work and, of course, in political discourse. In popular media, opinion polls have been a long-standing item in magazines and other publications (Holtz-Bacha and Strömbäck, 2012). More recently, online forums have taken on a more prominent role in encouraging individuals to participate in polls to document public opinion on a wide range of topics. In policy contexts, public opinion holds a critical place in the assessment and justification of the allocation of resources, the development of programmes and the implementation and cancellation of services. In Canada, for example:

Departments and agencies of the Government of Canada conduct public opinion research studies to gather the views and opinions of the Canadian public as a means of understanding and effectively responding to the diverse needs of Canadians. Public opinion research can be used to gauge public views and satisfaction with various government policies, programs and services, as well as to improve communication between government and citizens. It also serves as a tool to determine the level of knowledge, satisfaction and engagement of citizens with various government activities and initiatives. (Public Works and Government Services Canada, 2014: 1)

Public opinion and emerging science and technology

As a topic of social relevance, specific areas of science and technology have gained prominence at different points in time. Typically because of controversy attached to a particular technology or an emerging area of science, public sentiment has then become an object of interest, with opinion polls and surveys on the topic becoming prevalent. However, the study of public opinion on science and technology manifests somewhat differently from that of many other objects of social interest. For one, many discussions of science and technology are future-oriented (MacKenzie and O’Doherty, 2011) and characterized by multiple layers of uncertainty. Second, and perhaps more important in the current context, is how scholarship on the public understanding of science has developed over the last few decades. There has been a long-standing critique of how the public is viewed and framed in relation to science and technology (Irwin and Wynne, 1996). Often termed the deficit model of public understanding of science, this view holds that ‘the public’s unwillingness to comply with scientific prescriptions is due to public ignorance and media irresponsibility’ (Wynne, 2001: 445). Focusing in particular on areas of controversial science and technology, such as responses to GMOs in the UK in the 1990s and early 2000s, Wynne argues that such public responses are better understood as reactions to failures of institutional governance in general, as well as in relation to particular areas of science and technology.

Recently, Stilgoe, Lock and Wilsdon (2014) and Burgess (2014) have observed that there has been an important shift from ‘deficit to dialogue’, that is, from viewing the broader public as deficient and therefore in need of being informed about science and technology, to viewing lay publics as key stakeholders to be involved in dialogue about the governance of science and technology. This shift also reflects a change in emphasis in scholarship. Previous emphasis on the gauging of public understanding of science has increasingly given way to experimentation with models that recognize the relevant expertise of diverse publics and seek to incorporate these in policy decisions. Although largely in support of this shift, Stilgoe, Lock and Wilsdon identify important shortcomings. First, in spite of nominal rejection of deficit models of public understanding of science, many public dialogues implicitly incorporate deficit-like assumptions (see also Wynne, 2006). Second, Stilgoe, Lock and Wilsdon argue that while close attention to the evaluation of particular dialogue processes is warranted, what is needed is examination of the wider political context in which public engagement with science takes place. Similarly, Irwin (2006) argues that institutionalization of public participation in the governance of S&T needs to be treated with analytical scepticism, and the new role of public discourse in science policy carefully examined.

These debates signify a deep recognition of the importance of opening up societal decision-making about technically complex issues to broader public input, while acknowledging the significant challenges of achieving this in meaningful ways. One issue that has been inadequately explored is the ontological status that is accorded to public opinion. In this context, Burgess (2014) notes that one particular challenge inherent in operationalizing concepts such as ‘the public’ and ‘public interests’ is that although they are abstractions, attempts to consult publics necessarily presume or require their explicit instantiation.

Public opinion deconstructed

The notion of public opinion, as commonly used, is based on several premises. First, most fundamentally, individuals are assumed to have opinions. Opinions are therefore assumed to exist in an ontological sense and to be relatively enduring. Second, opinions are assumed to exist within individuals, and to some degree to be characteristic of that individual relative to some external referent. Opinions, therefore, are seen to be intra-psychic, bounded and specific to a particular issue or object. Third, opinions are seen to be cognitively accessible. When an individual is asked about his or her opinion, via a survey, poll, interview or focus group, he or she is presumed to be able to articulate that opinion in such a way that it is recognizable to the interlocutor. Individual opinions are thus posited to pre-exist the event in which an opinion is elicited and articulated. Fourth, the conceptual shift from individual opinion to public opinion posits a particular notion of a societal body, a ‘public’, that is characterized at any given time by a distinct sentiment towards a given issue or object. Because mechanisms for measuring public opinion rely on measuring large numbers of (appropriately sampled) individual opinions, public opinion is also implicitly a cognitive construct. Identification of a dominant public opinion therefore relies on aggregation of large numbers of individual opinions on a subject, which is then extrapolated to hold for an entire community or population.

Scholarly literature in several disciplines has shown the flaws in all of these premises. In the following, I draw selectively on this literature to demonstrate that the notion of public opinion as commonly used in the media as well as in policy and academic work is not defensible on epistemic grounds. In this I draw particularly on discursive psychological and related work.

Potter and Wetherell (1987) articulate a strong social constructionist critique of psychological research on attitudes, and in particular the ontological foundation upon which this research rests. The assumptions underlying notions of attitudes criticized by Potter and Wetherell mirror assumptions underlying notions of opinions implicit in surveys and polls. Potter and Wetherell observe the lack of theoretical coherence associated with the construct of attitude which is noted even by attitude theorists themselves. Nevertheless, a working definition of attitude seems to hold sway according to which ‘“When people are expressing attitudes they are giving responses which locate…objects of thought” on “dimensions of judgment”’ (Potter and Wetherell, 1987: 43, citing McGuire, 1985: 239). Using the example of research on racism, and focusing specifically on quantitative scales to measure racist attitudes, Potter and Wetherell demonstrate several failings of such an approach. First, the social categories used in attitude scales lack objective criteria, and this ambiguity undermines interpretation of responses. For example, the term ‘coloured immigrant’ used in Marsh’s scale (1976, cited in Potter and Wetherell, 1987) has no obvious referents with objective criteria for inclusion and exclusion. It is therefore unclear how individuals interpret the term when prompted to respond to the survey. Importantly, Marsh’s scale is not an anomaly reflecting bad research practice, but rather is typical of attitudinal scales. Second, in the translation of participants’ own expressions into categories the analyst uses to systematize and organize responses, the researcher by necessity glosses over contextual variation in the use of particular terms. The researcher therefore imposes her or his own analytical categories in interpreting participants’ responses. Third, the point of an attitude scale (and of an opinion survey) is not simply to see how individuals complete the scale or survey. Rather the point is that such a tool measures the underlying attitude (opinion) that gives rise to the response on the scale. That is, the scale is presumed to point to an unambiguous ‘object of thought’, the evaluation of which arises from an enduring individual psychological characteristic, the attitude. Importantly, this also presumes that attitudes are relatively stable objects within individuals which lead them to respond in predictable ways to the same ‘object of thought’ across different contexts, one of which is the attitude scale.

Potter and Wetherell demonstrate the flaws in these assumptions using interview data on immigration in New Zealand. In particular, they show that what could be taken as individuals’ expressions of attitudes are normally embedded in complex conversational contexts. Across these contexts it is not possible to infer a particular attitude in a straightforward manner. Because individuals’ expressions are never simply intended to describe an internal psychological state, but rather organized to accomplish discursive acts, linking actual discourse to specific attitudes held by the speaker is usually not possible. In this sense, therefore, the short decontextualized sentences represented in attitude surveys to elicit responses do not reflect language that actually occurs between people. Potter and Wetherell show further that the assumption that individuals give similar responses with respect to the same ‘object of thought’ is incorrect as systematic analysis of discourse reveals striking variation in the attitudinal expressions of even the same person. And finally, Potter and Wetherell argue that the foundational logic of attitude research requires an ontological separation of ‘attitude’ from the ‘object of thought’. The only way that it makes sense to study different people’s attitudes to an object is to assume that there is one object, toward which different people have different attitudes. The same can be said of opinion research: it implicitly requires us to consider that the object that is the topic of the opinion survey is separate from the opinions that people have of it. Potter and Wetherell show that in the interviews on immigration in New Zealand, participants do not simply express an attitude toward an unambiguous referent. Rather, it is in the course of developing an evaluation of something that speakers constitute the object they are evaluating. For example, one speaker who is giving his view about ‘Polynesian immigrants’ is shown to be formulating the nature of ‘Polynesian immigrants’ as he is expressing his evaluation. Because categories such as ‘Polynesian immigrant’ have flexibility with regard to the referent that is implied, statements in which attitudes towards Polynesian immigrants are expressed come to constitute that referent. It is thus not possible to distinguish categorically the ‘object of thought’ from the attitude held towards it. Similarly, when the opinions of different individuals’ responses to a survey item are aggregated, responses to different ontological objects are treated as if they referred to the same object.

Puchta and Potter (2002) extend this line of reasoning to attitudes and opinions that are expressed and measured through market research focus groups. They note that in spite of discursive social psychological and social constructionist research that points to the epistemological flaws in methods attempting to measure attitudes and opinions, much research continues to presuppose enduring underlying attitudes. Puchta and Potter argue that focus group methods, just like quantitative methods used to ‘measure’ attitudes and opinions, are structured in a way that leads to the appearance of attitudes and opinions as intra-psychic phenomena, and masks the flexible and rhetorical organization of evaluations. They provide a highly detailed turn-by-turn analysis of focus group transcripts to demonstrate that far from eliciting pre-existing individual opinions, focus group moderators meticulously guide participants toward the articulation of free-standing opinion packages. Their analysis shows that focus group participants engage in much talk that is evaluative, and that is embedded in particular contexts and rhetorically organized. They also show that the moderator responds to these contextual evaluations in systematic ways that ultimately lead to participants’ articulation of context-independent ‘opinions’ that are no longer linked to any particular rhetorical purpose. The ways in which the moderator achieves this include: (1) ignoring certain statements by participants; (2) explicitly articulating the kind of contributions that are welcomed; and (3) stripping away the rhetorical context in which an evaluation was articulated. Thus, Puchta and Potter argue that as is the case with quantitative scales, market research focus groups do not elicit and measure opinions that already exist as ontological objects within individuals, but rather produce free-standing opinion packages that come to resemble commonly held notions of individual opinions.

Another level of critique of public opinion research focuses less on the ontological status of opinions, and more on the kind of aggregation that is entailed in making claims about the opinions measured across many individuals. Because polls intended to measure public opinion are typically conducted on only a small sample of the population of interest, they inevitably require an extrapolation from the opinions measured in that sample to the opinion of the larger public. This extrapolation assumes that an appropriately selected sample will show the same characteristics as those of the population. These characteristics are statistical aggregates, in the form of mean values, which are often presumed to have a relative stability beyond the point in time measure of the survey.

The knowledge value of a survey aiming to measure public opinion therefore relies fundamentally on the notion of aggregation of individual responses to the survey. The epistemic foundation of this process of aggregation in turn relies on the notion of l’homme moyen, the ‘average man’ (Gigerenzer et al., 1989). The average man is an abstraction introduced by Quetelet in the 1830s that was critical to the development of statistical social science. Through it, extremes in individual behaviour could be ignored as incomprehensible deviations from the average man [sic] in whom ‘everything particular or exceptional balances out’ (ibid.: 41). Quetelet’s emphasis on statistical regularities instead of causal explanations of particularities went against prevailing methods in astronomy and mechanics. The pre-eminence of these disciplines required justification on the part of Quetelet, which he based on the concept of society. That is, he attributed regularities in the statistical measurements of people not to the behaviour of an assortment of individuals, but rather to ‘the particular, if fictitious, individual, the average man, representative of the “social body”’ (ibid.: 42).

When Quetelet introduced it, the notion of the average man was controversial. Given the ubiquity of opinion polling now, Quetelet arguably won the day. Indeed, the notion of ‘the average man’ has been so successful in the long run that it has now receded into the taken-for-granted background of what constitutes knowledge about society. Conversationally and conceptually, we rely on notions of ‘the everyday person’, ‘Average Joe’, ‘common man’, ‘man in the street’, ‘woman in the street’, ‘Jane Doe’, ‘John Doe’, etc. Even with more sophisticated polling methods that allow for the identification of different types within a population – for example, southern Baptists versus other Baptists (Igo, 2007) – the logic of making claims about the population of interest still relies on the notion that characterizing a fictitious average person as a reference person for a population is a meaningful activity of knowledge production. Suggestions that surveys reporting statistical regularities are not meaningful measures of social facts would seem alien or even absurd to many contemporary consumers of survey data. Although we no longer rely on the average man as a formal construct, public and expert discourse has fully incorporated the implications of building a science of public opinion measurement on a fictitious representative of ‘the public’, an arguably equally fictitious entity.

The shift that is required from opinion as an intra-psychic construct to public opinion as a societal property thus raises several awkward questions. Leaving aside for a moment problems associated with the ontology of individual opinions and attitudes, one might posit that the aggregation of many individual opinions across society might be useful for various purposes. Indeed, if we assume that opinions do exist as intra-psychic constructs, and if we further assume that they are cognitively accessible in such a way that surveys can act as tools to measure them, then it seems to be meaningful to characterize populations using descriptive statistics to describe the relative frequencies of opinions held by individuals in that population. But statements regarding public opinion typically claim to do more than that. The term ‘public opinion’ arguably implies the existence of a social body, in the sense of Quetelet’s assertions, with which that opinion is associated. And this is where the logic breaks down. It is not at all clear what kind of ontology to associate with public opinion, since it relies on a simplistic anthropomorphization of the ‘social body’. In contemporary discourse, we do not use the term ‘social body’, but we do use the term ‘public’ in analogous ways to imply a homogenous body about which knowledge can be obtained, such as the opinions it holds (i.e. public opinion). Given the heterogeneity of society, plurality of cultures, religious convictions and interest groups, it is strange to think that one might characterize all of these using a universalizing construct like public opinion which lumps all of this into a single analytical unit – the public. And even if we ignore these problems and allow for the awkward leap of thinking of the public as a homogenous body, and even if we allow for this public body having ‘opinions’, they would certainly not be opinions of the type envisaged by those who measure opinions and attitudes on the individual level (i.e. as cognitive constructs). That is, the measurement of public opinion cannot rely on the same epistemological foundations as the measurement of individual opinions because they are incompatible ontological constructs.

In short, surveys and polls cannot measure public opinion because the object they seek to characterize, public opinion, simply does not exist in the way that is presumed. But if surveys and polls do not measure public opinion, what is it that they produce? Drawing on critical and constructionist perspectives of public opinion research, Lipari (2000) argues that far from reflecting pre-existing cognitive phenomena, polls are constitutive of the opinions they purport to measure. Lipari continues to develop a discursive approach to polling that allows for analysis of the linguistic properties of poll questions (as opposed to presuming that these elicit psychological responses). Resulting analysis of poll questions enables identification of such elements as the information (or misinformation) that respondents must implicitly accept in providing a response, the range of positionings made available to respondents that constrains their expression of opinions, and the framing of relevant agents and actions. For example, Lipari considers the following poll question:

I’m going to read you two statements about the welfare system. Please tell me which of these statements about this issue comes closer to your point of view. A: The welfare system does more good than harm because it provides assistance and training for those who are without jobs and live in poverty. B: The welfare system does more harm than good, because it encourages the breakup of the family and discourages the work ethic. (Hart and Teeter for NBC and the Wall Street Journal, 1994, cited in Lipari, 2000: 206)

Other features of polls that implicitly constitute the opinions they nominally measure include concealing actors, attributing certain actions to particular agents, strategic vagueness to make particular propositions within questions unchallengeable, and designating different agents in different parts of the question.

What Lipari’s analysis demonstrates is not so much that an individual poll respondent’s opinion is shaped through the act of responding to the poll (though this of course may be the case as well). It is that the results of the poll are constituted in large part through the discursive construction of the poll question. Elements of poll questions such as propositions that have to be accepted irrespective of the response option chosen are not only communicated to poll respondents, but also, and arguably primarily, to the consumers of poll results. The results of the poll, that is, the public opinion that is characterized and subsequently mobilized for further policy (or other) action, does not pre-exist in the social body of the fictitious subject of the public; it is constituted in large part in the formulation and administration of the poll question. Importantly, this is not the result of ‘bad’ poll questions since all possible poll questions contain some presuppositions that need to be accepted when responding.

A final important critique of survey research recognizes the reflexive nature of public opinion studies. Nominally intended to measure psychological and social characteristics of the public, these methods simultaneously provide the results of these measures for consumption of that same public. Igo (2007) provides a detailed account of the ways in which public opinion science and data enter into American consciousness and the everyday lives of Americans. In observing public responses to information generated through polls and surveys, Igo comes to the same conclusion outlined above: such instruments cannot simply be understood as objective measures of independently existing objects. This therefore raises questions about how awareness of the results of surveys affects the individuals being surveyed, and implies that surveys are themselves active constituents of the phenomena they purport to measure. Very much in line with Gigerenzer et al.’s (1989) arguments, Igo therefore suggests that ‘the public’ as a recognizable body, as well as the individuals inhabiting it, is shaped by the availability of data produced through modern survey methods.

Deconstructing the notion of public opinion, in this context, allows us to question the ability of polls and surveys to act as measurement tools. If the ontological status of public opinion is questionable, then certainly any tools that have been devised to measure its characteristics are at best representing something that has not been defined properly, and at worst measuring nothing at all, but simply producing discursively the object they purport to represent. The legitimacy of public opinion polls therefore falters on a classical realist fallacy, namely that social objects can be characterized using epistemology of the natural sciences from a previous century (O’Doherty, 2015). That is, social and psychological phenomena are presumed to exist in ways similar to objects and phenomena studied in the natural sciences, and are therefore seen to be measurable in similar ways. ‘Opinions’, ‘attitudes’, ‘values’ and the like are presumed to be no different from, say, the properties of a metal, such as ‘density’, ‘length’ and ‘conductivity’. Public opinion can be said to represent something we might call a will or voice of the people only if it is ‘real’ in the sense that it has ontological status independent of the tools measuring it. Therefore, questioning not only our ability to measure public opinion, but also its very existence, brings with it important challenges for democratic governance. If the public does not have an opinion, then how can ‘the will’ of the public be known and acted upon by government?

Democracy and public deliberation

Danziger (1997: 149) explains that by the 1920s no one would have questioned that what was being measured by surveys on various political topics was, in fact, opinions. However, this peculiar way of conceiving of opinions, still so prevalent today, relied on a historical trajectory of developments within psychology and beyond that is explored by other articles in this special issue. Notably, it was specifically psychologists who were devoted to developing individualistic conceptualizations of the related concept of attitudes, in which the social was reduced to a particular category of stimulus acting on the individual, in contrast to sociologists who understood attitudes as inherently social phenomena to be examined in the interactions between people (Danziger, 1997). Danziger suggests that this peculiarly individualistic psychological way of operationalizing and measuring attitudes can be understood as emerging in the historical context and socio-political culture of liberal democracies. However, critical scholarship over the last few decades has increasingly challenged the idea that plurality of interests, values and positions in society can be understood as quantitative variation across known variables, and managed simply through aggregation of preferences.

In democratic societies, there is a fundamental link between the legitimacy of decisions that affect society as a whole, and the degree to which these decisions are informed by individuals and communities within that society. The notion of public opinion has been argued to be the ‘main currency’ of representative democracies (Holtz-Bacha and Strömbäck, 2012). However, as outlined above, critics of opinion polling have pointed out that as a mechanism for providing a meaningful opportunity for public input into collective societal decisions, polls and surveys fail. Nevertheless, the failure of polls and surveys in this regard does not detract from the fundamental need of decision-makers to be cognizant of the values, interests and positions of their constituents on any particular decision that will affect them. Thus, we not only require mechanisms to convey ‘public opinion’ to decision-makers; we also require mechanisms to articulate that ‘public opinion’ in the first place. I have argued above that the implicit construct underlying common and expert uses of the term ‘public opinion’ is flawed. Not only is the epistemology underlying its measurement problematic, the ontological assumptions of a singular opinion attributable to a cohesive social body are troubling. In contrast to common conceptualizations of public opinion, deliberative democratic theory offers a radically different way to think about mechanisms for articulating public will and linking it to societal decisions.

Theorists have observed the strong rise to prominence of deliberative democracy in many domains (Dryzek, 2010). However, as noted by Levine, Fung and Gastil (2005), there have been bursts of public participation and deliberation in the recent and distant past. What makes the recent deliberative turn different, and why is it occurring now? Arguably, what attracts many advocates to the idea of deliberation is, on the one hand, the distance many experience between themselves and meaningful contribution to collective decisions and, on the other, an increasingly strong reliance on ‘thin description’ of public interests in the form of attitude surveys and similar instruments (see Young, 2017). Deliberation also offers a radical alternative to dominant societal trends towards streamlining, reducing and simplifying social knowledge and facts. Instead, deliberation aims to nuance, complicate and add richness to public discourse.

At its core deliberative democratic theory aims to enhance democracy to lead to more just ways of addressing pluralism in society (Chambers, 2003). Dryzek (2010) proposes that deliberative democracy is primarily a theory of legitimacy. Deliberative democrats, then, hold that for a collective decision to be legitimate it needs to be achieved through deliberative participation of those who are affected by that decision. This necessitates that ‘ordinary’ citizens are involved in meaningful conversations about policy decisions that affect them, beyond just casting an occasional vote. The ways that deliberative democrats have suggested to achieve this vary, though there are common elements. Gutmann and Thompson (1997) argue that while disagreement is an inevitable component of democracies, it is possible for members of a society to deliberate about their disagreements in ways that contribute to the overall improvement of society. They suggest that deliberation improves collective decision-making through enhancing decisions’ legitimacy, basing them on questions of common rather than exclusively individual interest, through encouraging mutual respect among individuals even in the face of strong moral disagreement, and by basing decisions on greater individual and collective understanding.

Practical implementation of deliberative democratic principles takes many forms. I am here most interested in the use of mini-publics as mechanisms to provide a forum for public discourse on an issue, and the possibility of conveying the content and conclusions of those deliberations to decision-makers. Goodin and Dryzek (2006) describe mini-publics as groups of people that are small enough to be genuinely deliberative and large enough to be representative in some sense of that term. They include in their consideration of mini-publics citizen juries, planning cells, consensus conferences, deliberative polls, and several other designs, with a particular emphasis on forums involving lay citizens and non-partisans. Though mini-publics are rarely representative in an electoral or statistical sense, they can be representative in the sense of incorporating the diversity of perspectives in a given population on a particular issue (Longstaff and Burgess, 2010). With respect to the discussion above, it is important to note that a mini-public is not an abstraction in the way that the term ‘public’ is. Whereas ‘public’ does not denote an unambiguous referent, the term ‘mini-public’ points to a specific group of individuals who are assembled for a particular purpose at a particular point in time.

Mini-publics certainly do not embody the full potential of deliberative democratic theory (Felicetti, Niemeyer and Curato, 2015), but they are an important instantiation of deliberative democratic principles and at very least offer an opportunity for the empirical study of face-to-face conversations conducted within the constraints of deliberative democratic norms. They have also been argued to hold an important role in the development of public trust (MacKenzie and Warren, 2012). Mini-publics have been used globally to provide input on a variety of issues including electoral reform (Warren and Pearse, 2008), electricity services (Fishkin, 2003) and genetically modified foods (Nep and O’Doherty, 2013) (for a wide range of participatory projects including formal mini-public deliberations, see http://www.participedia.net/). They have also been used increasingly on issues relating to social and ethical implications of emerging or controversial science and technology (Cobb, 2012) and policy decisions relating to health and ethics (Abelson et al., 2013; Blacksher et al., 2012). The mandate according to which mini-publics operate can vary. They may be provided with strong mandates, committing decision-makers to enact directly conclusions of the forum; they may be provided with an advisory mandate, with decision-makers committing to taking into account recommendations of the forum, but not necessarily to enacting them directly; or they may be constituted from outside the policy sphere with no particular mandate, but with a commitment of the organizers to convey conclusions to policy-makers.

One important aspect of deliberation about S&T in which it differs from deliberation on some other topics, is that individuals participating in the deliberation are not likely to be familiar with, or are only superficially aware of, the issues under consideration. And the more future-oriented the topic (MacKenzie and O’Doherty, 2011), the more care needs to be taken in the provision and framing of materials for informing members of a deliberative forum (see O’Doherty et al., 2012, for methodological details such as avoiding selectivity in preparing materials). As noted above, in part following calls to reject deficit models of public understanding of science, it has been argued widely that policy decisions stand to benefit from the inclusion of non-expert voices. Critically, as Hamlett explains, the idea is not that non-expert citizens displace elected and appointed individuals and institutions, but rather that ordinary citizens are ‘given regular opportunities to review and consider various policy options and present their collective, deliberative opinions to policy makers’ (2003: 118). The practice of public deliberation has thus been accepted as an important mechanism to attempt to create concrete and meaningful avenues both for the review of policy options surrounding particular areas of science and technology, and for developing and conveying the outcomes of these deliberations to decision-makers.

Deliberative public opinion

Deliberation, in the context of democratic engagement, is not simply conversation. Indeed, Dryzek (2010) laments that the object of study of social scientists claiming to study ‘deliberation’ often does not resemble what deliberative democrats associate with that label. Although competing definitions for deliberation have been advocated, even the broadest definitions require that particular norms of interaction are followed. For example, in an attempt to extend the concept of deliberation beyond the typical contexts considered by political communication scholars, Gastil (2008: 8) offers the following definition: ‘when people deliberate, they carefully examine a problem and arrive at a well-reasoned solution after a period of inclusive, respectful consideration of diverse points of view’. Other definitions are often more specific. In the context of a comprehensive review of deliberative democratic theory, Chambers (2003: 309) provides the following description:

…deliberation is debate and discussion aimed at producing reasonable, well-informed opinions in which participants are willing to revise preferences in light of discussion, new information, and claims made by fellow participants. Although consensus need not be the ultimate aim of deliberation, and participants are expected to pursue their interests, an overarching interest in the legitimacy of outcomes (understood as justification to all affected) ideally characterizes deliberation.

Criteria 1–3 relate mainly to the organization of the forum and the provision by organizers of structural elements to enable deliberation. Criteria 4–6 relate to the deliberative norms that guide the conversation. Criterion 7 relates to participants moving beyond exclusively pursuing their own interests to conceptualizing over-arching collective positions that accommodate plural and even divergent individual positions.

Adherence to norms of deliberation (criteria 4–7) allows for the development of what Niemeyer and Dryzek (2007) call intersubjective rationality. Through the ongoing exchange of perspectives, participants in deliberation have the opportunity to expand the breadth of their assessment of the issues under discussion limited only by their collective experience and creativity. Because perspectives are backed by reasoning and ongoing expansion of the collective knowledge base, participants’ subjective development throughout the process of deliberation does not occur in isolation; rather, each participant’s engagement with the issues under consideration becomes progressively inter-connected with the subjective experience of the other participants. Because opinions and reasoning are publicly expressed and constituted, their evaluation occurs in the context of other perspectives expressed in the deliberative space. Opinions that do not withstand the scrutiny of the deliberation ideally fall away, and collective opinions develop that are accepted and inter-subjectively considered to be valid.

There has been some discussion with regard to the status of consensus as the normative outcome of deliberation (Moore and O’Doherty, 2014). In contrast to the theoretical frameworks of both Rawls and Habermas, stipulating complete consensus as an ideal outcome for deliberation is now widely rejected. Niemeyer and Dryzek, for example, advocate meta-consensus, in which participants agree about ‘the nature of the issues at hand, not necessarily on the actual outcome’ (2007: 500) as an alternative, and more satisfactory, outcome. Another outcome of deliberation can be persistent disagreement on a particular issue, despite prolonged deliberation (Burgess, O’Doherty and Secko, 2008). Assuming that this polarization is not simply a reflection of vested interests or of participants’ unwillingness to shift from initial entrenched positions, such persistent disagreement can provide important insights into fundamental value differences and critical points for policy- makers to consider.

What is evident across these ideal outcomes of deliberation (i.e. consensus, meta-consensus, persistent disagreement, and others not discussed here) is an intersubjective convergence toward a collective position (or positions) and an explicit articulation of that position. The outcome of deliberation in the form of a convergence of opinion is described by Hamlett as the ‘reasoned, informed, consensual judgment forged out of the initially disparate knowledge, values, and preferences of the participants, as these have evolved through the deliberative experience itself’ (2003: 122).

The development of a collective position is well illustrated empirically by Walmsley (2011) in the context of a deliberation about the social and ethical implications of human tissue biobanks. A group of deliberants encountered very strong disagreement among some of its members regarding the issue of whether donors to tissue biobanks should be compensated. After intense discussion, several positions emerged including an ‘entrepreneur’ position, arguing for direct compensation of donors, a ‘rights advocate’ position, which argued against compensation based on ‘respect for humanity and life’ and the commodification of body parts that would be entailed in compensation agreements, and a ‘communitarian’ position, which argued that biobanks should be seen as a public good and thus be associated with altruistic donor motivations. Walmsley notes the strong and seemingly intractable disagreement that developed among deliberants across these positions. In a detailed analysis of deliberation transcripts, Walmsley identifies pivotal moments in the discussions that allowed deliberation to move forwards. These moments involved individuals articulating novel examples that reconfigured the debate and advancing creative and often innovative suggestions that collectively led to a well-developed consensus position. For example, Walmsley describes that consideration of the case of ‘special donors’, such as women in Africa who have AIDS but are not adversely affected, and of whether they should be paid for providing samples to a research biobank, led to novel insights regarding the implications of compensating donors. The consensus position ultimately developed by the group was that biobank donors should be compensated with tax credits and with health information, and not with direct payments or special health benefits or privileges. Importantly, this consensus was at the level of support for the position as well as the reasoning that had led to it. This reasoning was not a linear string of arguments advanced by one individual. Rather, it was the product of intense interaction and scrutiny of both individually advanced and collectively developed arguments.

Even in the case of a persistent disagreement that cannot be resolved in spite of the deliberative process, the positions about which there is disagreement ideally were arrived at through deliberation. For example, in a public deliberation about the use of microbial genomics technology for remediation of pollution caused by RDX (an explosive commonly used by the military that is also a neurotoxin), the deliberative forum was split on whether to support a mandated reduction and ultimate ban of RDX. Two dominant positions emerged. Of the 25 deliberants, 9 supported the statement: ‘There is currently not sufficient evidence that the harms of RDX justify a mandated reduction or ban’; 14 supported the statement: ‘Based on current information, we should aim for a phased reduction of RDX’ (Moore and O’Doherty, 2014; O’Doherty et al., 2013). These two positions were not present at the outset of the group discussions, but rather emerged in a manner similar to that described by Walmsley (2011). In spite of the disagreement on this issue, each of the two dominant positions constitutes an outcome of a deliberative process in its own right, supported by a consensus among the group supporting the respective statement.

We can therefore define deliberative public opinion ¹ as an explicitly articulated consensual position that has emerged from a deliberative process. Deliberative public opinion is thus characterized by intersubjective agreement on a particular statement, collective acceptance of the legitimacy of the process whereby agreement on the position was reached, and agreement on the acceptance and rejection of particular ideas and arguments that were put forward during the deliberative process. Deliberative public opinion is an acknowledged social construct, in the sense that it very obviously emerges from social processes. It is not an aggregation of several individuals’ putative opinions or even expressed statements. And while the deliberative public opinion expressed by a particular forum is likely to be informed by the values, preferences and experiences of individual deliberants, no pretence is made that the stated deliberative public opinion is a measurement of the opinions individuals might have held before the deliberative process. As an acknowledged social construct, deliberative public opinion has an explicitly recognized historical trajectory and is situated in a particular social and institutional context. Because the political legitimacy of a given deliberative public opinion is associated with the particular forum that produced it, it becomes disconnected from any one individual once the deliberative process is concluded. Individuals may subsequently change their positions without affecting the status of the deliberative public opinion as the product of the process that produced it.

As a historically situated object, the political legitimacy of a deliberative public opinion is also associated with the historical moment that produced it. For example, the results of a public deliberative process about xenotransplantation in Canada (Einsiedel, 2002) must be understood in the context of political debates, dominant discourses and prominent cultural, religious and other values at that point in time. If the outcomes of a public deliberation process are incorporated into policy, ideally this should be done with cognizance of the historical moment that produced that particular deliberative public opinion. Particularly in the context of emerging science and technology, what is technologically possible can change quickly and dramatically (see, for example, changes in information technology leading to the necessary reconsideration of what kind of health data count as ‘anonymized’ or ‘identified’, Homer et al., 2008). Similarly, public experiences of and responses to the technologies can change quickly and dramatically (see, for example, recent public responses to efforts to collect, collate and share information collected by the National Health Services in the UK, Sterckx et al., 2016). For the governance of these technologies to be ethically sound, an updating of the deliberative public opinion that is used to inform the norms and values of the governing principles may therefore be required.

It should be very evident, therefore, that deliberative public opinion cannot be measured, at least not in the sense that polls and surveys typically intend to measure public opinion on an issue. Traditional public opinion polls purport to measure something that is seen to exist independently of the measurement tool. In contrast, deliberative public opinion, as developed here, is seen to emerge from a particular process and, hence, not to have pre-existed the moment of its articulation. In contrasting the outcomes of deliberative polling ² to the outcomes of a traditional opinion poll, Fishkin and Luskin suggest that the results of a deliberative poll can be understood to be ‘what people would think if they thought, knew, and talked about the issues’ (2005: 294). While intuitively plausible, and philosophically attractive, this description can only ever be an interpretation of the outcome of a deliberative poll, not an operational definition. Because this description is not falsifiable (i.e. the counterfactual of what people would think if they were exposed to information and processes cannot be tested), it cannot be used as a foundation for measurement for such a hypothetical construct. For these reasons, deliberative processes cannot produce measurements of the kind that would satisfy those seeking the certainty of a positivist philosophy of science. What deliberative processes can do is to produce social and political objects, deliberative public opinions that can hold important substantive content to enhance the quality of decisions and critical legitimating power, providing they are constituted appropriately.

Discussion and conclusion

The purpose of this article has been to develop an ontological and epistemological foundation for the construct of deliberative group opinion. Although I have drawn on key literature from political theory, the core argument is more of a contribution to theoretical psychology than it is to political theory. In particular, I have built on criticisms of public opinion as an ontological category to develop an argument for a social ontology of the notion of deliberative public opinion. However, in contrast to previous critiques I have argued that social constructionist principles can be used not only to deconstruct accepted notions of public opinion, but also to advance theoretical articulation of deliberative public opinion.

There are too many aspects of my argument that I have left undeveloped for it to be a proper contribution to the political theory of deliberative democracy. One area, in particular, that requires further development relates to recent criticisms from the literature of the public understanding of science. As described above, scholars such as Irwin (2006) point out that while there has been a proliferation of deliberative forums on a range of S&T topics, there has been insufficient attention to the ways in which such activities are situated within or relate to institutional governance of S&T. While Irwin argues against being dismissive of current and past attempts of engaging publics in the governance of science and technology, he does advocate for a sceptical response. No matter how well designed any given deliberative forum is, therefore, its ability to challenge technocratic ways of governing S&T is constrained by the particular institutional and social relations that give rise to the forum, and within which it operates. This observation is reflected in criticisms from deliberative democrats who argue that too much attention has focused on mini-publics as instantiations of deliberative democratic principles (see Dryzek, 2010: 6), and who question whether the empirical turn to mini-publics has come with a loss of criticality for deliberative democracy (see Böker and Elstub, 2015). One response to these criticisms is work by political theory scholars on frameworks for situating deliberative mini-publics in larger political systems (Dryzek, 2015; Felicetti, Niemeyer and Curato, 2015; Mackenzie and Warren, 2012; Niemeyer, 2014; Warren, 2013). Similarly, S&T scholars have begun working on understanding the fragmentary nature of the policy sphere within which decisions about S&T are made (Irwin and Horst, 2014). Although mini-public deliberation is seen to have a role in these larger political systems and regulatory environments, the precise nature of this role is still emerging. Articulating an ontological foundation for deliberative public opinion should therefore not be read as naïve advocacy for particular models of public deliberation as an endpoint to discussions about decision-making on public issues, but rather as an essential component in being able to experiment with and assess instantiations of deliberative democratic governance.

My contribution to these broader debates is peripheral, in that theoretical articulation of deliberative public opinion does not help directly to address questions about (macro) political structure. Instead, the arguments presented here are intended to bolster the epistemological foundation upon which deliberative processes rely. Analysis of many published and unpublished instantiations of mini-public deliberation suggests that they involve little or no reflection on the ontological status of psychological constructs that are relied upon in documenting proceedings and results of deliberation. This is most notable in quantitative analyses of deliberation; for example, in measuring changes in individuals’ opinions pre- and post-deliberation. A failure to reflect on the theoretical foundations of deliberative phenomena is also evident in qualitative analyses that report results of deliberation in the form of thematic analyses, without reflection on how the analytical construct of a theme relates to psychological and social constructs, such as opinions and values. It is this lacuna that I have attempted to address.

In summary, I have argued that the epistemic foundation of an opinion in the sense of a stable intra-psychic object that is cognitively accessible to individuals and measurable via simple (e.g. Likert) scales is suspect. The notion of a public opinion in the sense of a measurable property of a social body is based on faulty reasoning and does not have a meaningfully articulated ontological foundation. Public opinion as it is used commonly in lay and expert discourse is thus a social construct. It is a discursive object that is broadly presumed to be a natural rather than a social phenomenon. It has the taken-for-granted status of an object that exists independently of the discursive processes that constitute it. As such, public opinion is a prime candidate for scholarly deconstruction, which demonstrates that public opinion is not, after all, a natural object, but a socially constituted object. As is typical for many social constructionist analyses, demonstrating the social contingency of public opinion undermines the legitimacy of discourse in which it is measured, talked about and used as if it had ontological status independent of the social processes that constituted it in the first place. In contrast, deliberative public opinion is an acknowledged social construct. Demonstrating the social and discursive processes that constitute deliberative public opinion does not detract from the legitimacy of the construct, nor does it detract from its social utility. To the contrary, it is precisely through demonstrating the concordance with democratic principles of social and discursive processes that constitute deliberative public opinion in any given instance that social legitimacy and utility are claimed.