Active and latent social groups and their interactional expertise

Introduction

We set ourselves the task of exploring the ways in which interactional expertise (see Collins and Evans, 2014; Collins et al., 2006; Evans and Collins, 2010) is applied in interaction. Given this aim, we initiated a study of interactional expertise in practice, opening up the role of interactional expertise for the formation of social groups and their composition. As such, we expand and strengthen the existing studies in interactional expertise by showing that it is relevant not only for displaying membership in a social group, but also for their construction.

The term ‘interactional expertise’ was originally coined by Collins and Evans (2002). It refers to the capacity to grasp the conceptual structure of another’s social world and speak fluently in that world (Collins and Evans, 2014). Their understanding of interactional expertise quintessentially distinguishes between discursive performance and practical expertise (Collins and Evans, 2015). They maintain that interactional expertise does not require the ability to manage and execute the corresponding practices of the social realm; it is thus a third kind of knowledge, placed between formal propositional knowledge and embodied skills (Collins, 2004).

Collins and his colleagues proposed the imitation game (henceforth IG) as a method to test empirically the notion of interactional expertise and its distribution between social groups. The IG is inspired by Alan Turing’s (1950) idea of testing whether machines can be intelligent. According to Turing, if a person cannot distinguish a human from a computer through a five-minute interrogation, it would be reasonable to call the machine intelligent, because we judge other people’s intelligence in a similar manner.

The IG was then developed into a standardized method for exploring the level of interactional expertise between groups in society. The founding hypothesis of the IG is that interactional expertise improves the chances of passing the test. Furthermore, because minorities are extensively immersed in majority/dominant populations, thereby acquiring the expertise to interact with them, majority/dominant population members are expected to be less successful at imitating minority members (Collins and Evans, 2014). This analysis considers the pretenders’ success as a direct proxy for the pretenders’ amount of interactional expertise.

In this article we reveal what lies behind the success rates of the IG. Drawing upon ethnomethodology, we look at the methods participants employ in an imitation game on religion in order to distinguish group members from non-members. This allows us to focus on sociologically relevant issues, such as participants’ reasoning, criteria and categorizations of social groups and their respective management of memberships. The results indicate a variation in the management of membership by active Christians, who use knowledge-based questions, and non-religious individuals, who mainly use opinion-based questions. The success rates thus depend on more variables than the expertise of the pretender groups. Our analysis also points towards further uses of the IG and interactional expertise: the studies on identification and recognition of group members allow us to address the formation and maintenance of social groups.

The imitation game as a method

Initially, Collins and Evans (2014) aimed at developing a standardized procedure for studying interactional expertise. However, IG experiments can be organized in many ways. Here, we present a widely used version of the IG that consists of three players, none of whom knows the identity of the others, interacting via computers. The players are drawn from two social groups, with the three roles allocated to players as follows: the judge and non-pretender are always from the same social group, representing the ‘target expertise’ that is being tested; while the pretender is from the other social group (see Figure 1). The judge designs questions for the other two participants, one of whom is an outsider but is asked to answer as if a member of the target group. The judge’s task is then to determine who is pretending and who is answering naturally.

OPEN IN VIEWER

Figure 1.

The concept of the imitation game (Collins and Evans, 2014).

Because interactional expertise (IE) is acquired through socialization, two hypotheses regarding interactions between members of different social groups can be proposed. In the Identify Condition, the pretender group (e.g. non-religious) is expected to have relatively little exposure to the target group (e.g. active Christians), and these limited interactions make it relatively easy for the target group member to identify the pretender (Collins and Evans, 2014).

The second configuration is the Chance Condition. Here, the pretender group members (e.g. active Christians) are expected to be deeply immersed in the life of the target group (e.g. non-religious); therefore, they have interactional expertise (IE) that makes it difficult for target group members to differentiate them from the majority group members. As a result, identifications in this configuration are mere guesses (i.e. chance). Our analysis will pose an alternative interpretation to these hypotheses.

The relative success rates of judges over a series of IGs can be compared by using a standardized indicator called the ‘Identification Ratio’ (IR), which is derived by dividing the number of excess right guesses (right guesses minus wrong guesses) by the total number of guesses. In the Identify Condition, the expectation is that the IR would be closer to 1 (i.e. almost all guesses are correct), whilst in the Chance Condition, the IR would be closer to 0 (i.e. there are roughly equal numbers of right and wrong guesses). Where IGs are played in both conditions, the difference in IR between Identify and Chance Conditions gives the effect size, which is a standardized measure for how much easier it is for a member of a group X to pretend to be a member of a group Y than vice versa. For instance, in a secular society we might expect members of a religious group to be better at pretending to be non-religious than non-religious persons pretending to be religious. We may interpret the finding also as a measure of the exclusiveness of the target group. If the IR is 1, the one group has absolute accuracy in differentiating non-members. If, on the other hand, it is zero, then the group is open and non-members pass the test as successfully as members themselves.

Previous IG results showed that social groups tend to fall between the extremes above (Collins and Evans, 2014). For instance, individuals who are colour blind, blind or have perfect pitch communicate fluently with outsiders, but for outsiders it can difficult to integrate in these groups. Similarly, religious groups (e.g. active Christians) and gays can assimilate well in mainstream society: gays can successfully pretend to be straight, and active Christians can successfully pretend to be non-religious.

IGs can facilitate three kinds of comparisons: between different social groups; between different societies; and diachronic, historical comparisons. In general, the results indicate that people from minority/marginal social groups or backgrounds have enough IE to successfully blend in with the surrounding social world. IGs also provide for comparisons between countries: games on religion and/or sexuality have been organized in Finland, Hungary, Italy, the Netherlands, Norway, Poland and Spain. Finally, diachronic comparisons provide for examinations of the changes in the distribution of interactional expertise (for instance, changes in the spread of IE between women and men if participation in society becomes more equal). Similarly, the public prominence of a religious group would diminish its exclusiveness and improve its diffusion.

In Helsinki, IGs have been organized since 2012. In 2012 and 2013, religion games involved the social groups of active Christians and non-religious participants by self-identification. In both religion games, the IR was 0.65, which came very close to the result in Cardiff (0.68; Collins and Evans, 2014). The materials of the Helsinki religion games will be studied qualitatively here. In addition, a politics game and several ethnicity games were organized in Helsinki in 2014–2017.

In their statistical analyses on the distribution of IE between social groups, Collins and Evans did not analyse what was going on in the interactions between players in the game. In this article, we will open the ‘black box’ of IG experiments with the help of ethnomethodology and conversation analysis (EMCA) (Heritage, 1984; Sidnell and Stivers, 2013). We will address the game interaction and explore how people ‘recognize whether somebody is a member of one or another category’ (Sacks, 1979), and show that the players’ reflexive assignment of categories is consequential for the formation of group boundaries. We argue that interactional expertise is not a quality that social groups possess but, rather, is a practice through which a social group emerges (Berger and Luckmann, 1966). We will study participants’ reasoning practices and categorizations as well as the respective management of group membership. We will shift the focus from the study of the distribution of IE between social groups to its use in the formation of group boundaries (Brubaker and Cooper, 2000).

Social interaction and interactional expertise

Collins’ (2010) interest in interactional expertise grew out of an explication of the role of tacit knowledge for social practices. Expertise is usually tacit rather than explicit (Collins and Evans, 2007); however, Collins (2011) also supports a strong linguistic socialization hypothesis, according to which an understanding of tacit expertise can be articulated and taught linguistically without embodiment. Consequently, interactional expertise is a purely discursive phenomenon that allows articulation of embodied tacit practices (Collins and Evans, 2007; see Ribeiro and Lima, 2016). Accordingly, IGs are an adequate tool for analysing discursively articulated interactional expertise.

Our IG analysis employs EMCA to explicate the participants’ work to display and account for their membership in a social group, which establishes their interactional expertise in passing as members of the group. As with Collins and Evans (2007; 2014), EMCA considers immersion in social interaction as a basis for linguistic socialization (e.g. Levinson, 1992), and it has grown into a systematic field of study for structures and procedures of social interaction.

We draw on those aspects of EMCA that account for people’s group membership. Previously, Harvey Sacks (Sacks, 1979: 11) studied ‘hotrodders’, and he saw, in their category use, a revolution from the bottom up, a self-governance of group identities, such that ‘they will recognize whether somebody is a member of one or another category, and what that membership takes, and they can do the sanctioning’ (Sacks, 1979:11). Furthermore, this is an example of an emergent social group that can define who is counted among its members: ‘and what is known about hotrodders – what they do with their cars, how they look, how they behave – these are things that hotrodders can enforce on each other and defend against non-members’ (Sacks, 1979: 12).

The analysis of membership categories (MCA) has developed into a research field of its own (Jayyusi, 1984; Rautajoki, 2016; Silverman, 1998). Of particular interest has been the dynamic use of categories: these are not simply labels but, rather, a way to explicate a meaning of social actions. Furthermore, the flexibility of categories makes them suitable for moral work, as speakers can expand or narrow the scope of their categories to match their ongoing social action (Pihlaja, 2014).

Another development of Sacks’ ideas is conversation analysis (CA), in which studies of epistemics (knowledge in interaction) and affective alignments are relevant for studies of group membership. In recent years, epistemics has grown into a focal area of CA (Heritage, 2012, 2013). The central notion of epistemics is epistemic status, which refers to relative knowledge between two or more speakers: a speaker may present themself as being knowledgeable to a greater or lesser extent than the recipient(s). Often, epistemics uses Kamio’s (1997) theory of information territories (though this has also been criticized; see Lynch and Macbeth 2016). According to Kamio (1997), the closer the speaker is to the target area of knowledge in terms of immediacy, the higher is the speaker’s epistemic status. Kamio’s theory calls for inspection of the relationship of interactants’ territories of knowledge (i.e. epistemic domains). We have used this theory to analyse the judges’ assessments as to whether another player’s epistemic domain corresponds to the player’s domain as a way of identifying group membership. We have called this ‘epistemic correspondence’. In contrast, the lack of correspondence provides the judge with an opportunity to recognize a pretender.

There is also a tradition for studies on emotional or affective alignment between interactants (Peräkylä and Sorjonen, 2012). When participants recognize the similarity of their experiences, they can identify with the recipient, thereby forming the basis of group membership (Arminen, 1998). We have called this ‘experiential correspondence’. The lack of affective reciprocity (i.e. experiential correspondence) may in turn provide a basis for exclusion (Goodwin, 2006).

Finally, the granularity of descriptions is relevant for both epistemics and experiential group boundary management, as suggested by Sacks (1979: 8): ‘why do kids go about making up all those typologies of cars – and the typologies they have are really enormously elaborate, and they use those typologies to make assessments of other drivers’. Sacks pointed out that group categorizations are the members’ resource (i.e. new categories can be invoked for establishing new boundaries if existing group categories are not sufficient).

Data and method

The data consist of computer-mediated conversations from IGs on religion that were played at the University of Helsinki in 2013, organized in collaboration with Harry Collins’ team. In total 70 active-Christian students and 195 non-religious students participated in the games that were played in the labs for two days. Participants were recruited through email lists and on the spot, and they received free movie tickets in return for their participation. The designated categories of ‘active Christians’ or ‘non-religious’ were not perceived as problematic and participants were able to choose their category. However, without further study it is impossible to know how many students decided not to participate in the study. That said, the students’ use of the categories was somewhat flexible during the game: we will return to that issue in the analysis. While the original data are in Finnish (see Arminen and Simonen, 2015) and the analysis is based on that, it can be followed through English translation.

The data contain the participants’ questions, answers and judgments. In the first stage of the experiment, 20 active Christians and 20 non-religious persons (age 19–37 years, mean 24 years) assumed the roles of judges, pretenders and non-pretenders in turns. The total game time for each student ranged from 60 to 120 minutes. Our analysis concerns this first stage, in which the question–answer sequences occurred intact in real time. For statistical purposes, more participants were involved with the help of various online procedures, but from a game interaction perspective the varying online conditions caused a variation, the effect of which is hard to estimate; we therefore restricted our analysis to the first stage, which occurred in real time.

The IG contains adjacency pairs (Schegloff and Sacks, 1973): the judge’s question and two players’ answers. The judge decides the questions; the two other players independently decide how they answer. Each adjacency pair follows the structure: the judge writes and submits the question (line 1 in the following examples) and the software delivers it to the players. The pretender and non-pretender receive the question and write their answers. The software delivers the answers back to the judge, who receives them simultaneously. The judge then selects the alleged pretender, and chooses his/her level of confidence, as follows: (1) ‘I have little or no idea who is who’; (2) ‘I am more unsure than sure’; (3) ‘I am more sure than unsure’; and (4) ‘I am pretty sure I know who is who’. The judge then writes an account for the judgment.

The IG experiment is based on a fundamental transformation of constitutive social structures. Participants are selected according to their acknowledged membership category (e.g. religious and secular), information hidden from the judge. This provides for an examination of the participants’ ability to pass from one social group to another and to identify others’ attempts to do the same. We paid close attention to the social group’s in-group knowledge and experiences of their group membership. Social groups appear to set varying conditions for interactions with them, thereby making the display of interactional expertise contingent on those conditions – something we will demonstrate in the following sections.

Epistemic correspondence

Active Christians have theological knowledge, which they also apply for testing the respondents’ group membership by identifying whether recipients share an epistemic domain that corresponds to their own. We start the analysis from the Identify Condition, where judges and non-pretenders are active Christians, and non-religious pretenders attempt to imitate active Christians. Later, we analyse the Chance Condition, where judges and non-pretenders are non-religious, unlike the pretenders who imitate them.

In Q1, an active Christian judge poses a two-part question (line 1). The judge refers to ‘grace’, thereby presupposing a certain level of religious expertise. Moreover, the question is presented as a test or an exam question, in which the judge can assess the appropriateness of answers (Arminen, 2016; Heritage, 1984). Test questions help identify who shares the same knowledge or experience.

Q1. The Christian judge gains certainty (uid = 2885)

Respondent 1 (line 2) shows special expertise by treating ‘grace’ as an abbreviation conveying special content: the capital letters (U-L-F-M) in the English words ‘Undeserved Love For Me’ correspond to the letters in the word ‘grace’ in Finnish (armo), which were given in upper case (A-R-M-O). Only after that does the player answer the questions (lines 2–5). Respondent 2 answers briefly and directly (line 6). The judge asserts (lines 7–8) that Respondent 1 has relevant knowledge, while Respondent 2 offers inaccurate and false information.

The judge in Q1 has a certain theological understanding of grace. The test question allows the judge to measure epistemic distances between the judge’s own, and the respondents’, theological domains. The comparison of epistemic correspondence enables the judge to establish who is who.

Next, in Q2, another Christian judge uses a test question, this time framed more in terms of personal experience.

Q2. The Christian judge gains certainty (uid = 2881)

In Q2, the test concerns Letter to the Romans, a theological topic that allows the judge to measure the respondents’ epistemic distance from the target area. The respondents’ turns (lines 2–5 and 6–7) reveal their unequal access to the domain, as in Q1. Respondent 1 presents her understanding of the topic by a reference to the ‘doctrine of righteousness’. According to the judge (lines 8–10), respondent 2 is inaccurate and is not able to link the topic to her own experiences. The judge’s account accentuates the factors that makes the recognition of group membership possible: the ability to summarize the theological topic and discuss it personally. In contrast, sketchiness and an inability to prove theological knowledge speak against membership. The judge is sure of her (the judge’s own) recognition.

In everyday life, speakers are sensitive to each other’s epistemic statuses and try to match the granularity of their descriptions to the expected level of their recipients (for example, those revealed by the level of a previous turn, such as a question (Schegloff, 2000). Moreover, the granularity gives an indication of the speakers’ degree of expertise, competence and ways of perceiving reality (Fasulo and Fiore, 2007). In Q1 and Q2, the judges appear similarly sensitive to the variation of granularity, and they interpret the lack of detail as a lack of expertise, which for its part becomes consequential for the judgement of group membership.

Experiential correspondence

We now turn to experiential correspondence and its depth. In this section we describe how participants try to determine whether they share the same experiences and feel about them in the same way (Herlin and Visapää, 2016). In Q3, the judge recognizes the pretender in the very first question by comparing the recipients’ personal experiences with her own.

Q3. The Christian judge gains certainty (uid = 2981)

The question (line 1) conveys the visibility of Christianity as a presupposition (cf. ‘how does x’ or ‘does x’; Hayano, 2013). The judge’s assessment (lines 9–10) reveals that the question preference had been deceptive; that is, the judge agrees that the visibility – the presupposition – is debatable (lines 9–10). Respondent 1 (in line 2) recognizes the controversial nature of this and answers accordingly. In turn, respondent 2 (lines 5–8), lists features that aim at demonstrating the visibility of her faith, and she appears not to have realized the controversial presupposition. For the judge (lines 9–10), the first respondent’s ability to recognize the concealed controversy provides a point of identification, and the judge agrees with the (first) respondent. The judge then provides a more elaborate account of the respondent 2 (lines 10–16). The second respondent attempts to provide an image of Christian life, but according to the judge she offers inaccurate and implausible details that give away her identity (Holstein, 1993). The judge claims that respondent 2 is ignorant, has stereotypical thoughts, and ‘tries too hard’. Two factors – the experiential correspondence between the judge and respondent 1, and the contrast between the judge and respondent 2 – facilitated the identification.

The example of Q3 demonstrates how the judge accesses her own experience of Christianity as a resource for interpretation of the responses, to establish who sees and understands things in the same way as she, and thus facilitates identification. In this way, we have moved from epistemics of theological knowledge to experiential correspondence among Christians sharing experiences. The judge acts as if she had access to the other players’ minds (cf. sharing experiences in AA; Arminen, 1998). The identification establishes experiential correspondence that comes close to an agreement of opinions. How uniform social groups are in terms of their experiences or opinions remains an empirical question.

In Q4, an attempt to identify with the respondent establishes a permanent and unrepairable criterion (see Garfinkel, 1967: 76–103; Heritage, 1984: 84–97). Having identified the respondent, the judge considers the case closed. Here the judge receives nearly identical answers to the first question. The judge thus poses a question on the experiences of Christianity and assesses the answers in terms of like-mindedness.

Q4. (a) The Christian judge’s second question (uid = 2883)

In her assessment (lines 8–10), the judge focuses on respondent 2, pointing out contrasts with her own way of thinking. First, she refers (lines 8–9) to societal values as ‘external pressure’, which to her mind is not so powerful. Second, she refers to homosexuality as the ‘gay card’, and she considers bringing that up as an indication of not being a Christian (lines 9–10). Her description of respondent 2 is built as a complaint against a person who is ‘always willing to discuss’ homosexuality (Pomerantz, 1984). This difference in the preference of discussion topics moulds her understanding of respondent 2’s group membership. By contrast, the judge does not mention respondent 1 at all in her account; apparently she does not experience a similar clash with respondent 1. Also, in this case, the experiential correspondence is the decisive criterion by means of which the pretender is recognised.

However, the judge needs to reconsider her interpretation after the fourth question.

Q4. (b) The Christian judge loses certainty (uid = 2883)

In a test-like manner, the judge expects that the Christian player would emphasize grace in her response (line 10). However, it is respondent 2, whom the judge has already identified as a pretender, who introduces the subject of grace (line 5). The judge resolves this reality disjuncture (Pollner, 1987) by inventing a new category: a non-Christian with knowledge of Christianity. This new category explains the anomaly and supports the judge’s prevailing interpretation (Pihlaja, 2014). This example shows that the identification, the experiential tie, may prevail despite some epistemic contradictions. The identification gained – ‘I am like her’ – remains, even though the participants may differ in terms of knowledge-related issues. The judge notes that the responses are unexpected, but she is able to explain away this deviance. According to the judge, the non-group member may have knowledge about the group (i.e. a non-religious person may have theological knowledge), but this individual piece of knowledge is not considered sufficient to alter her already established view on the group membership of the players. Similarly, a group member may not have perfect in-group knowledge; an individual missing a piece of knowledge is not yet a sufficient basis for reconsideration. This example supports the view that identification based on empathy and sharing of experiences is considered deeper and more embodied than knowledge and epistemic issues, which are considered more transient and more easily learnable (Zlatev and Blomberg 2016).

The judge underlines the salience of experiential correspondence in her final assessment.

Q4. (c) The Christian judge’s final judgment (uid = 2883)

Only the answer to question Q4 led to the idea that person 1 would be the

Category-bound opinions

Next, we analysed the Chance Condition, in which judges are non-religious and Christian players try to imitate the non-religious group. Here the questioning strategies of non-religious judges appear to be very different from those of Christian judges, who can make use of their theological knowledge and pose related epistemic questions. Non-religious judges simply do not possess a parallel field of knowledge; there is no institutionalized or conventionalized secular ‘A-theology’ (or whatever one might call it). Consequently, there are no parallel epistemic questions, such as Q1 and Q2, posed by the non-religious judges. There are some experiential questions, such as Q3 and Q4, but overwhelmingly non-religious judges seem to pose what we call ‘opinion questions’ (i.e. they aim to find out how respondents regard a particular issue), and they seem to think that opinions separate Christians and non-religious persons. Moreover, non-religious judges often specify and reformulate the category of the recipient whose opinion they ask. This can be called a ‘search for category-bound opinions’.

Let us consider the example of Q5, where the judge uses the contrast pair ‘Christian’ and ‘atheist’.

Q5. The non-religious judge invokes the categories ‘Christian’ and ‘atheist’ (uid = 2890)

The question in Q5 does not reformulate the group division of the experiment from ‘active Christians’ and ‘non-religious persons’ but portrays the relevant line of opposition between Christians and atheists – and (in line 8) it is revealed that this is also how the judge perceives the groups of the experiment. The account further reveals that she is searching for a like-minded respondent (lines 9–10). This allows us to reconstruct the judge’s understanding of the non-religious group. For the judge, it is a group of like-minded people who are devotedly non-religious. Here, as well, it is an empirical question as to whether the players’ understanding of the social group is adequate. The methodological consequence for the IG is that the adequacy of the judges’ conception of their social group is critical for the game result. If the judges’ accounts of the group are biased, then the assessments suffer from the same bias.

Notably, many opinion questions recategorize the membership groups of the IG. In Q6 a question that is on the borderline between opinion and personal experience is posed to ‘non-active church members’. Thus, the social groups of the game are recategorized into active church members vs. others.

Q6. The non-religious judge invokes a category ‘non-active Church member’ (uid = 2892)

The question (lines 1–2) shows that the judge assumes that reluctance to be an active Church member is a characteristic feature of the non-religious group. The judge’s categorization renders the non-religious social group as broad and diffuse; that is, all those who are not active members in the Church. Notably, the respondents give very different accounts: blatantly utilitarian (costs and benefits) vs. value-based (resonance with the message of the Church). The account by the respondent 1 (line 3) appears to cover only a small sub-category of the group the judge has invoked – that is, those who are not church members – and gives an account for that. The judge only considers the answer of the respondent 1, which to her is stereotypical (lines 6–7). We can conclude that, for the judge, non-active Christians represent diffuse social groups who are not distinguished by their relationship to materialism. Consequently, the judge is misaligned with respondent 1, whose categorization of the non-religious group differs from hers. If the social group is broad and diffuse, the level of agreement between the group members may be weak and the identification of group membership based on category-bound opinions will remain uncertain.

In some cases, the judge’s group categorization is explicitly dispersed. In Q7 the judge defines the target area with three different groups; therefore, the question per se offers three possible group memberships to choose from. The further subdivisions erase the boundary between the two original groups.

Q7. The non-religious judge invokes three new categories (uid = 2900)

The judge takes the internal disunity of non-religious persons as his starting point and invokes three group categories: ‘non-active Christians’, ‘non-Christians’ and ‘those whose relationship to religion has changed’ (lines 1–2). Hence, the question is addressed to members of three distinct social groups. The respondents assign themselves to the different alternative group memberships offered (lines 3–8). The judge makes a choice in terms of plausibility, but admits being unable to justify it (line 9). Methodologically, the IG requires a judge to have sufficient competence (interactional expertise) to draw a clear and distinct boundary for her/his own social group. If the judge has only a latent understanding of her/his social group, then s/he has also limited chances to assess and enforce membership and defend the group against non-members (Collins and Evans, 2007; Sacks, 1979).

The Chance Condition, where non-religious judges attempted to identify pretenders (i.e. Christians who imitated non-religious individuals), was difficult for judges: they were unable to present test questions that would reveal the pretender. Analysis of the question sequences suggests that the difficulty did not arise out of the interactional expertise of Christians (i.e. their very good answers), but from the non-religious people’s latent, weak understanding of their own group. Because the group members lacked a reflexive understanding of their social group, the group dissolved into disunity. The non-religious group becomes defined in dispersed ways, either as persons not belonging to the Church, non-active Christians, agnostics or downright atheists. Group membership appears to be a reflexive accomplishment, dependent on the maintenance of and orientation regarding group boundaries. When group boundary maintenance is weak, the group becomes porous, boundaries become blurred and social existence is threatened (Antaki and Widdicombe, 1998). Latent social groups lack solid boundaries against non-members, which influences the IG results.

Summary

The Christian judges (n = 20) were good at identifying the pretender in the Identify Condition (correct: 14; undecided: 2; incorrect: 4), while the non-religious judges (n = 20) made correct identifications in fewer than half of the games (correct: 8; undecided: 6; incorrect 6). The standard interpretation has been that Christians are better pretenders than non-religious people, because they have been immersed in the life of a non-religious majority (e.g. Collins and Evans, 2014). Our analysis also considered the question sequences and showed that they varied between the groups. Christians have an epistemic and experiential access to a religious domain (e.g. grace, Letter to the Romans, the God relationship), enabling the use of test questions. In particular, the epistemic questions are specific to the Christian group, while there is no conventional in-group knowledge for secular people. Furthermore, the Chance Condition game shows that non-religious individuals are internally disunited, which hinders the use of epistemic and experiential test questions. As a result, the non-religious judges attempted to recognize their group members on the basis of their alleged like-mindedness by making category-bound opinion questions, whose results barely exceeded random chance.

Discussion and conclusions

We were able to identify factors that contribute to the results of the IG. The opposing groups, Christians and non-religious persons, posed systematically different types of questions when trying to recognize the group membership of participants, and these differences appeared to have a bearing on the results. Christian judges (Identify Condition) tried to seek out group membership with epistemic and experiential (test) questions. Conversely, non-religious judges (Chance Condition) tried to invoke group membership by seeking like-mindedness and (re)formulating categories.

The observation on the distribution of question types between groups puts earlier interpretations of IG results in a new light. Earlier interpretations suggested that greater success in pretending group membership equals a greater degree of interactional expertise (Collins and Evans, 2014, 2015). Because different groups use different questioning strategies, however, they pose varying difficulties for pretending group membership. Members of active groups, such as religious groups, can test others’ group membership using test questions concerning knowledge and experiences. Latent groups lack self-conscious identity and reflexivity and are thus prone to dispersion, which hinders the use of test questions. In this way, the IG makes the distinction between active and latent groups relevant.

Active groups are communities of practice that have epistemic foundations and a set of conventionalized or institutionalized practices to articulate and realize their knowledge base (Lave and Wenger, 1991). In this way, group members may orient to their common knowledge and have a joint understanding of their boundary objects (Huvila, 2011). In-group views on boundary objects allow the group to enforce its own boundaries. The types of group ties may vary between different social groups. Segersven (2016) has suggested a distinction between epistemic and experiential groups on the basis of their group foundations. Indeed, colour-blind people or persons with perfect pitch test their group membership with experiential questions, whereas gravitational wave physicists use epistemic questions (Collins and Evans, 2007: 99–106).

Because latent groups – such as non-religious persons or non-Finns (immigrants) – do not orient to their group characterization per se, they need to try to invoke their group characterization, but may have limited chances to do so. According to Collins and Evans (2007: 35–40), interactional expertise has a parasitic nature, being dependent on the contributory practices upon which the social group is based. Without any contributory expertise, such as the religious life of a congregation or gravitational wave physics experiments, there cannot be any interactional expertise, which is always learned from contributory experts, such as priests or scientists (Collins and Evans, 2007: 35–40). In that respect, latent social groups do not have contributory expertise: they do not have organized, specialized practices that can be taught to those not participating in them. These groups also lack epistemic foundations (e.g. secular A-theology) and have a limited amount of shared experiences; consequently, they lack epistemic or experiential questions (Collins and Evans, 2007: 99–106). In the absence of contributory expertise, latent groups cannot pose questions concerning it. They are therefore forced to establish identification almost solely through like-mindedness of opinions. According to the typology of expertise by Collins and Evans (2007), latent groups appear only to have ubiquitous tacit knowledge and can only make inquiries concerning it. Opinion questions used by the latent group serve that purpose perfectly. They are posed from the standpoint of what everybody knows, and they are addressed to everybody. To sum up, constellations of latent social group members being judges do not reveal interactional expertise but, rather, ubiquitous expertise of what everyone knows.

Furthermore, IG experiments prove Brubaker’s (2002) distinction between groups and categories. Groups are bounded collectivities with a sense of solidarity and identity, entailing a potential for concerted action. Categories are no more than latent groups, in relation to which groupness can be invested. Investment requires social capital, which has been shown to be epistemic or experiential. We may also need to distinguish between categories and their incumbents. Even if a non-religious category/latent group lacks contributory practices, it does not mean that those who belong to that latent group lack other types of contributory or interactional expertise; they may be invested in other categories, because each person is an incumbent of multiple categories. Investment in categories is the contributory practice through which groupness can be found. This also makes interactional expertise relational. To communicate fluently with a member of an active group, such as an ice-hole fishing community, you need to have interactional expertise related to the contributory practices of the community. Equally, to communicate with members of a latent group, who do not have contributory practices, one can simply use one’s common sense and there is no need to have any specific interactional expertise.

The attempt to build a quantified measure for the distribution of interactional expertise by comparing the results of Identify and Chance Condition games may have been naïve, because latent groups do not possess the capability to discriminate members from non-members. Categories are sociologically interesting, but in-group specific interactional expertise can only be found in active social groups with epistemic and/or experiential foundations. In fact, the idea of interactional expertise as such does not need to be questioned. Quite the opposite applies: our analysis has re-specified the notion of interactional expertise (Collins and Evans, 2002, 2014, 2015). We conclude that interactional expertise is the relational ability to form an epistemologically and experientially relevant relationship with the recipient.

Even though the quantification of interactional expertise still requires work, the results of an IG can be used for revealing people’s methodical practices of group-making (Brubaker and Cooper, 2000; Brubaker, 2002, 2016). IGs make visible not only the epistemic and experiential practices of group formation but also their flaws, such as the use of stereotypes. These characteristics open educational possibilities for the IG. Imitation games could also be used in teaching participants to understand group processes of inclusion and exclusion. Enhancing understanding about the ways the ‘us’/’them’ distinction is formed in social practices is an endeavour worth attempting.