The value of practice: A critique of interactional expertise

Introduction

In a series of publications since 2002, Collins and Evans (C&E) have proposed a ‘normative theory of expertise’. One of their goals is to solve the ‘problem of demarcation’ in public debates involving technical matters: who should participate? For C&E (2002), the decision-making process in such debates can be divided into two phases: ‘political’ and ‘technical’ (p. 262). In the former, all citizens would have the right to participate, while in the latter, ‘the right to contribute technically to a technical decision is to be assessed by examining expertise’ (C&E, 2002: 260). As proposed by C&E, the normative theory of expertise is then limited to the technical domain.

C&E (2002) posit three broad types of expertise – ‘contributory’, ‘interactional’ and ‘referred’ (pp. 254–257). Contributory expertise is acquired by ‘full-scale physical immersion’ (i.e., practice) within a field and leads to the ability to make contributions to that field or to given problems within it (C&E, 2002: 254–256, 2007: 59). Contributory experts are, then, full-blown practitioners. Interactional expertise (IE) is gained solely ‘through linguistic socialization’ with practitioners in a given domain and enables one to speak a technical language fluently (Collins, 2004b: 125). Referred expertise is earned when ‘expertise in one field can be applied in another’ (C&E, 2002: 257). According to C&E (2002), the solution to the demarcation problem – who has the ‘right to contribute technically to a technical decision’ – is to find the ‘appropriate balance’ (p. 260) among individuals who possess these three types of expertise in relation to the problem at hand. While those with contributory and referred expertises can contribute to the debate based on their practical experiences, the contribution of interactional experts would come from the understanding and judgemental ability that linguistic fluency enables (C&E, 2002, 2007).

C&E’s ‘Third Wave’ normative theory of expertise provoked a tsunami. They were criticized for their ‘rhetorical strategies’ (Rip, 2003: 419) as well as for their ‘reductive’ and ‘decontextualized’ reading – or even ‘misreading’ – of science studies (Jasanoff, 2003: 391, 393, 398). The division between the ‘political’ and ‘technical’ phases was also contested. The critics stated that it is not possible to disentangle science and power or science and politics in ‘how public issues are framed’ (Wynne, 2003: 402) or in defining ‘who counts as an expert (and what counts as expertise)’ (Jasanoff, 2003: 393). Additionally, C&E were criticized for providing an ‘internalist sociology’ of science in their analysis (Rip, 2003: 420; Wynne, 2003: 402). For the critics, this led C&E to mistakenly focus on the ‘core-set as the entry-point’ (Jasanoff, 2003: 395–397; Rip, 2003: 421–424; Wynne, 2003: 409), to ‘neglect context’ (Rip, 2003: 420; Wynne, 2003: 404, 408, 410) and, ultimately, to reduce public issues to epistemic issues (Jasanoff, 2003: 394; Wynne, 2003: 403, 408).

Despite the criticisms, the ‘normative theory of expertise’ remains under-scrutinized. C&E’s theoretical proposal has been generally treated as if it were established fact. ¹ The argument for ‘limiting participation to contributory and interactional experts in technical phases of disputes’ (Durant, 2011: 710) is reproduced uncritically, and the idea of interactional expertise (C&E, 2002) has been gaining space in academic journals and books (e.g. C&E, 2007, 2014a, 2014b; Collins, 2004b, 2011, 2013; Evans and Plows, 2007; Gorman, 2010; Jenkins, 2010; Reyes-Galindo and Duarte, 2015; Ribeiro, 2007c; Weinel, 2007; Wehrens, 2015) as well as in broader journals and magazines such as in Nature (Giles, 2006) and The Atlantic (Selinger, 2012). ²

The goal of this article is to discuss C&E’s typology of expertise, especially the concept of interactional expertise (IE) and the category of interactional experts (IEs). The questions we will address are as follows: Is IE a ‘real’ expertise? If so, what is an interactional expert able to perform? Are there any differences between the contributory, interactional and referred types of expertise? What are the ‘technical’ contributions and limitations of each of them? In sum, if we could bracket out the power and issue-framing problems of public debates, are these types of expertise able to indicate whether someone is an expert or can make a technical contribution to a technical debate?

We start this article by providing the reader with some context of the development of the idea of IE and introduce some of our relevant empirical research. Such contextualization provides the basis for the second part of the article: an empirical, methodological and philosophical critique of IE. In the final section, we return to the questions above, showing the limits of C&E’s solution to the problem of demarcation.

Interactional expertise in context: Setting out the problem

The idea of IE was born of C&E’s (2002) reflections on their fieldwork experience as sociologists of scientific knowledge. It was first defined as having ‘enough expertise to interact interestingly with participants and carry out a sociological analysis’ (C&E, 2002: 254). Two years later, IE was described as ‘the ability to converse expertly about a practical skill or expertise, but without being able to practise it’ (Collins, 2004b: 125).

More simply, IE just means ‘linguistic fluency’ in a (technical) domain. Thus, a statement such as ‘we are all interactional experts’ (Collins, 2011: 274) is to be read as ‘we are all fluent in one language (or another)’. This practical definition enables any person who is fluent in any language to assess this idea and its claims because they can relate it to their own experience. This is clear, for instance, when C&E (2002) say that a ‘necessary but not sufficient condition of translation [between different social worlds] is the achievement of interactional expertise [i.e. linguistic fluency] in each of the fields between which translation is to be accomplished’ (p. 258).

The assumed novelties connected to IE can be summarized in four central claims:

Proponents of IE argue that examples of interactional experts and imitation game (IG) experiments provide proof of the idea of IE (C&E, 2007: 98, 2014a: 5). The latter are Turing-test-like experiments that are taken as proving the ‘“strong interactional hypothesis” – that those with maximal interactional expertise are indistinguishable from those with contributory expertise in linguistic tests’ (C&E, 2007: 11). This means that IEs can – in principle and in practice – answer questions posed by practitioners in discursive settings without the latter being able to identify them as non-practitioners.

As we will see, there are distinct versions of IE, depending on how it is acquired. Its most ‘pure’ version – what we will call ‘pure-IE’ – refers to the claim that an individual can learn a domain language by talking to its practitioners solely within ‘discursive settings’ – that is, in settings that are ‘distant from or unaffected by the material and practical context of the domain’ (Collins, 2011: 294). A pure-IE would have the same understanding as practitioners about what can be spoken:

Imagine a person who has been blind and restricted to a wheelchair from birth. The claim is that such a person could acquire a practical understanding of tennis solely from extended and intensive discussion of tennis in the company of tennis players, without watching tennis or stirring from their wheelchair; such a person could, in principle, understand tennis as well as someone who had played it all their lives. (Collins, 2011: 272–273)

A philosophical claim is tightly linked to the idea of IE and applies to both natural and technical languages: ‘only the minimal bodily requirements necessary to learn any language are necessary to learn the language of any community’ (C&E, 2007: 79; emphasis added). The sociological argument is that the individual human body – or more broadly ‘embodiment’ – is not as relevant for acquiring a language as is linguistic socialization (C&E, 2007: 110). Accordingly, to become fluent in any language, one only needs a ‘minimal body’, which consists of ‘those bits of the brain to do with language-processing and those bits of the body to do with language-learning and speaking: the ears, larynx, and the rest of the vocal apparatus’ (C&E, 2007: 78). This means that ‘fluency in the language of a domain can be acquired outside of bodily engagement with the practices of the domain’ (C&E, 2007: 90).

The idea of IE establishes a social space – ‘discursive settings’ (Collins, 2011: 294) – within which practitioners and IEs are equivalent, regardless of the distinct nature of their experiences within a particular practice. This equivalence is based on the assumption that it is possible ‘to separate the notion of understanding from practical experience and practical contributions’ (Collins, 2013: 411). Fluency in a technical language would then enable a non-practitioner to make ‘practical judgements’ comparable to those of practitioners within discursive settings because technical ‘language contains practical understanding’ (Collins, 2011: 282) and is ‘loaded with tacit knowledge, Wittgensteinian rules, and [the] ability to make intuitive judgements, as any native language’ (C&E, 2007: 87).

Granting a relatively autonomous but equally powerful role to language vis-à-vis the practice of practitioners gives the impression that IE could have wide implications. IE is regarded as the reason why managers of ‘large scientific projects’ can ‘make technical decisions about sciences in which they have never made a research contribution’ (Collins and Sanders, 2007: 621); peer-reviewers in funding agencies can do their job ‘properly’ (Collins, 2011: 273) ‘when only sometimes [they are] contributors to the narrow specialism being evaluated’ (C&E, 2007: 32); and ‘non-scientists’, such science and technology studies (STS) scholars, can ‘accomplish deep and authentic analysis of sciences they do not actually practice’ (Collins, 2011: 273). As proposed, IE would alter the power balance between specialists and lay-people-turned-into-IEs, changing the conditions and forms of public participation in scientific and technological controversies. In short, if accepted by the academic community, the idea of IE would change our way of understanding the world of practices and the connection between science, technology and society. ³

Before we move on to analysing the four claims associated with IE, we present some empirical research undertaken by the first author (R.R.) of this article, which will serve as support for our arguments.

R.R. witnessed first-hand the early days of the idea of IE. He arrived at Cardiff University in 2003, one year after the publication of the ‘Third Wave’ paper (C&E, 2002) and just before the first IE experiments took place. As a graduate student supervised by C&E, he worked as a voluntary research assistant in such experiments, a support that was kindly acknowledged afterwards (Collins et al., 2006: 656). However, his own research in industrial settings, during and following his graduate programme, provides empirical data that are in conflict with the idea of IE.

Novices in a nickel plant

The research with the interpreters suggests the idea of ‘types of immersion’, a term used to describe the ‘various kinds of experience one or more individuals can go through within a form of life or collectivity’ (Ribeiro, 2007a: 16–17). Stemming from this research, Ribeiro (2007a) proposes five types of immersion (p. 76–80): non-immersion (e.g. machines), self-study (e.g. just reading), ‘linguistic socialization [alone]’ (Collins, 2004b), ‘physical contiguity’ (Ribeiro, 2007c) and ‘physical immersion [i.e. practice]’ (C&E, 2007: 59). The assumption underlying these types of immersion is that, except for machines, the type of immersion individuals undergo within a practice defines their abilities and understanding with respect to that particular practice.

A study of the pre-operational training of novice employees at a new US$3.2 billion nickel plant in a remote area of Brazil provides a test for the types of immersion framework. ⁴ Except for ‘non-immersion’, the multiple types of training the novices received perfectly matched the other four types of immersion. For instance, in the beginning of their training, novices were asked to read operational procedures and manuals (self-study) and attend classroom training (linguistic socialization). They were then sent to observe the plant assembly first-hand and to make technical visits to other plants (physical contiguity). Eventually, many of them experienced periods of on-the-job training in similar plants (physical immersion).

Each later instance of immersion generally mixed distinct experiences. For example, when the novices made technical visits, they also talked to practitioners; during the on-the-job training, they continued talking with practitioners and visited other areas of the plant. This means, for instance, that ‘linguistic socialization’ was present in all kinds of immersion from self-study onwards. Here, we will use the terms ‘linguistic socialization alone’ as a type of immersion and ‘linguistic socialization’ as an experience individuals have within physical contiguity and physical immersion. ⁵

The study involves witnessing some of novices’ training and also interviewing the nickel novices about that training; precise research goals were not mentioned. Thus, the novices’ reports on differences between the types of training they had were taken as instances of differences between types of immersion. Table 1 classifies some statements made by novices during interviews.

OPEN IN VIEWER

Table 1.

Differences between types of immersion (adapted from Ribeiro, 2013a).

Self-study	Linguistic socialization alone	Physical contiguity	Physical immersion
1. ‘If you only read, at times you imagine something that is completely different … And this can happen, you may have an interpretation’ 2. ‘There were things that we would read, read and imagine, [but] sometimes we would imagine in a more vague way than the reality actually is’ 3. ‘[If you do not go there] you can only imagine what it is like’ 4. ‘The problem [of going to the plant without the supervisor] is that you look but do not see’ 5. ‘[The specialist] always provided us with technical material that helped us to visualize the process’	6. ‘The discussions forced us to go back [to theory] and understand [the reasons behind a given event]’ 7. ‘The discussions sped up the absorption [of the material]. One thing is just reading; another is to discuss with an experienced person. It helps you to short-cut the way’ 8. ‘[Classes] do not show how difficult [the task] is’ 9. ‘A film does not present [any] risks’ 10. ‘There will be things that the camera will not catch’ 11. ‘Looking at a picture you cannot feel the heat [of the calcinate − 950°C–1,742°F]’	12. ‘[You see the] real scale [of the equipment] within the context’ 13. ‘[You gain a] notion of the temperatures, distances, risks’ 14. ‘You see what else constitutes the equipment, what is connected to it’ 15. ‘[You get] to know from inside out, fragile points … When you operate [in the future], it helps, you know the limits’ 16. ‘[During technical visits, you] witness unforeseen events’ 17. ‘[To follow plant assembling] raise questions … interest’	18. ‘[Repetition:] 5 months looking at the calcinate’ 19. ‘Sensitivity in operating the button switch, given that not all [machines] are the same … the oldest had a delay’ 20. ‘When you see the relevance of your action … you can see the outcome of your work, whether positive or negative, see what to adjust 21. ‘If you do not pass through intense situations, you learn nothing’ 22. ‘[I had a quasi] accident … an explosion, with smoke, air movement. [I was] sent back 10 metres. It raised my interest in safety’

To ‘imagine’ was the novices’ only option when they were experiencing ‘self-study’, that is, when they could not witness the practices about which they were reading (quotations 1–3). Most of the novices came from farms or small cities nearby and had never worked in any kind of industrial plant before joining the company. Like the Japanese–Portuguese interpreters, they were learning the language of a form of life they had never heard of before. This indicates that those who go through self-study – or even ‘linguistic socialization alone’ (Collins, 2004b, 2011) – in fields completely new to them may reproduce what they have been reading or listening to, but they will not really know what they are talking about.

In moving from ‘self-study’ to ‘linguistic socialization alone’, the novices started interacting with the members of the given form of life, that is, with its practitioners. Talking to old-timers helped novices to know what was relevant to read or to look at when following the construction of the plant (quotations 4 and 5). ⁶ It also helped them to speed up their learning and conceptual understanding of the processes by discussing papers they had read previously (quotations 6 and 7).

Sociologically, linguistic socialization alone means getting access to the field, which is a significant move. However, this is still a limited experience, as it is confined to what newcomers can make sense out of their out-of-context conversations with practitioners. To help novices understand the production process, old-timers would sometimes even show novices a picture of, for instance, the furnace used to melt nickel or a film of workers removing metal and slag from it. ⁷ In any case, all of these strategies were insufficient for novices to get a grip on the scale of equipment and its contexts or on how difficult and dangerous were the tasks they would perform in the future (quotations 8–11).

Physical contiguity allows not only for grasping the language used to describe what one is witnessing but also for acquiring a perceptual, lived experience of the overall context and of what newcomers have been reading or listening to (quotations 12–15). As a trainee explained, ‘[it is] on the shop floor that [a novice] will make the link between his imagination and reality. He will be able to connect one thing to another and make associations: “Ah! Now I know!”’. ⁸

Finally, physical immersion is a means of developing the abilities, skills and understanding to cope with a given set of practices (quotations 18 and 19). It allows for ‘involvement’ to take place (Dreyfus and Dreyfus, 1988: 26) along with the emotions and learning that accompany it (Dreyfus, 2009) (quotations 20–22). In the case of IEs, it is not expected that they will develop any of the motor and perceptual skills practitioners display when working. The issue here is whether this practical experience gives practitioners distinct linguistic fluency, understanding and judgemental abilities compared with IEs.

The relevance of the study with the nickel novices for our discussion is twofold. First, it corroborates the findings of the research with the Japanese–Portuguese interpreters: the limits of ‘linguistic socialization alone’ for acquiring an understanding of a set of practices that goes beyond ‘imagination’ – and how ‘physical contiguity’ facilitates it. The second point is that the ‘linguistic socialization’ that takes place within ‘physical contiguity’ and ‘physical immersion’ is distinct from ‘linguistic socialization alone’ (Collins, 2004b, 2011), both in its depth and breadth.

To be close to the practices leads newcomers to see tools or equipment and witness events that, in turn, raise questions and provide more opportunities for eliciting the experience of old-timers (quotations 16 and 17). Moreover, the professionals met when visiting an industrial plant are different from those met inside classrooms or other meetings, as experienced supervisors, technicians and workers do not usually become formal instructors. This increases the chances for talking more about the practice about which one is learning when experiencing physical contiguity. Finally, additional questions always appear when individuals have to do something by themselves – that is, undergo physical immersion – for the first time, which enhances the opportunities for posing questions. Altogether, the research with the interpreters and the nickel novices highlights the importance of looking at the types of immersion individuals have experienced when analysing their linguistic fluency, understanding and judgemental abilities.

Putting IE in context calls for further investigation of the main claims underlying it. We divide this exploration into four parts, in the next four sections. In the ‘Interactional expertise in practice’ section, we discuss the empirical claim concerning the existence of IEs and argue that there is not a single case of pure-IE in all the examples found in the literature. In the ‘Researching Interactional Expertise’ section, an in-depth analysis of the IG experiments uncovers an intrinsic flaw in their design, showing that this method cannot be treated as a ‘proof of the concept’ of IE. In the ‘Interactional expertise in principle’ section, we analyse the philosophical assumptions underlying pure-IE and present in-principle arguments against this concept and its supporting ‘minimal embodiment hypothesis’. Finally, in the ‘Types of immersion, perspectives and understanding’ section, we show that non-practitioners who are fluent in the language of a technical domain and practitioners of such domain do not share the same linguistic fluency, understanding and judgemental abilities within discursive settings.

Interactional expertise in practice

Until 2004, six professions had been analysed and two main cases utilized to support the claim of IE. These were as follows: ‘art critics’ (C&E, 2002: 244), ‘sociologists of scientific knowledge’ (C&E, 2002: 254; Collins, 2004b: 125), ‘scientists themselves’ (Collins, 2004b: 125), ‘high class [“high level” or “specialist”] journalists’ (Collins, 2004b: 129), ‘certain types of science administrators’ (Collins, 2004b: 129), ‘colour-blind persons’ (Collins, 2004b: 138) and the cases of ‘Madeleine’ (Sacks, 1985) (Collins, 2004b: 131–132) and of ‘AIDS activists’ (Epstein, 1995) (C&E, 2002: 262).

From 2004 to 2007, many other cases were said to be instances of interactional experts or of those who needed IE to do their jobs. They were as follows: ‘all kinds of participant observers, including ethnographers and social anthropologists … [and] reviewers [of funding agencies and journal editing]’ as well as ‘pitch perceivers’ (Collins et al., 2006: 659), ‘human coaches’, ‘salespersons’, ‘managers’ (C&E, 2007: 30–32), ‘technical connoisseurs’ (such as ‘architects’ and ‘wine buffs’) (C&E, 2007: 58–59) and ‘midwives’ (Schilhab et al., 2010).

Collins (2011) coined a new term – ‘special interactional expert’ – to designate ‘the special group of interactional experts who are not contributory experts [i.e., practitioners]’ (p. 274). Collins (2011) does not discuss how ‘special-IE’ is acquired in this case – if by ‘linguistic socialization alone’ or by ‘physical contiguity’. However, according to the given examples – ‘sociologists, anthropologists, high-level journalists and so forth’ (Collins, 2011: 283) – we can assume that special-IEs develop their abilities through ‘physical contiguity’.

In order to better comprehend all of the cases above, it is necessary to distinguish, in the context of types of immersion, the existing types of IEs that have been discussed so far. The qualifiers ‘pure’ and ‘special’ will refer to those who have acquired IE through ‘linguistic socialization alone’ and ‘physical contiguity’, respectively. Finally, there are contributory experts who learn the language of their domain while working; we will refer to them as ‘typical’-interactional experts. In sum, we will adopt the terms pure-IE, special-IE and typical-IE here in order to link IE to how it is developed, linking them to ‘linguistic socialization alone’, ‘physical contiguity’ and ‘physical immersion’, respectively.

Cases of IEs found in the literature are revisited below through the lens of types of immersion. Methodologically, this entails looking for the simplest type of immersion these individuals needed to have experienced in order to have acquired their versions of IE. For instance, architects were once classified as IEs as they could not tile bathrooms, but supposedly could judge the work of bricklayers based on their linguistic fluency in the ‘tiling’ language. ⁹ The concept of types of immersion, however, allows us to trace abilities to the experiences IEs need to have had. This makes it easy to realize that ‘judging tiling’ requires previous physical immersion: one needs to see and assess many tiling jobs – and not simply talk about tiling – until one is able to visually distinguish, for instance, whether the tiles are aligned or even.

Table 2 displays the most representative cases of IEs according to their types of immersion. ¹⁰ Graphically, it shows that there is no empirical evidence – not a single example – of pure-IE. All cases fall under either the ‘physical contiguity’ or ‘physical immersion’ categories, with the exception of two that are classified under ‘natural language’. We argue that the claims underlying IE have been overstated by treating distinct cases of language acquisition as though they were similar phenomena. In addition, the category of immersion under which a case should be placed depends on which activity or ability is under analysis (Ribeiro, 2013a). This means that what is to be considered ‘practice’ changes from one activity to another. Thus, the experience of ‘observing’ is practice for architects assessing a completed tiling job, and therefore, they fall under the physical immersion category for this activity. On the other hand, observation is not practice for architects who can only talk about doing tiling, and they would then fall into the category of physical contiguity for this second activity. ¹¹

OPEN IN VIEWER

Table 2.

Cases of ‘interactional experts’ by types of immersion.

Linguistic socialization alone	Physical contiguity	Physical immersion
A means to develop Pure-IE (e.g. Collins, 2004a, 2011)	A means to develop Special-IE (e.g. Collins, 2011)	A means to develop Typical-IE (e.g. C&E, 2007; Collins, 2011)
	‘Sociologists of scientific knowledge’(C&E, 2002; Collins, 2004a, 2011)‘High-level journalists’(Collins, 2004a, 2011)‘All kinds of participant observers … ethnographers, social anthropologists’(Collins, 2011)‘Japanese–Portuguese interpreters’(Ribeiro, 2007c)	‘Technical connoisseurs’, ‘art critics’, ‘architects’, ‘wine buffs’(C&E, 2002, 2007)‘Pitch-perceivers’(Collins et al., 2006)‘Human coaches’(C&E, 2007)‘Scientists themselves’(Collins, 2004b)
Specialist language	‘Aids activists’(C&E, 2002)’Midwives’(Schilhab et al., 2010)‘Managers of scientific projects’ and ‘Managers’(C&E, 2007; Collins and Sanders, 2007)
Natural language		‘Madeleine’(Collins, 2004b; C&E, 2007)‘Colour-blind people’(C&E, 2007)

The following discussion will then focus on one activity/ability at a time. The analysis starts with the cases at the right side of Table 2.

Technical connoisseurs, art critics, architects, wine buffs and all situations in which the ability of making perceptual discriminations is necessary (such as seeing a fully finished tiling job, viewing artwork or tasting wine) call for physical immersion. Technical connoisseurs are then practitioners/typical-IEs with regard to ‘perceptual discriminations’ because they can both make and talk about them. ¹² These very same professionals, however, would have been classified as special-IEs if the activity of ‘doing’ were the focus of analysis. In this situation, they would have fallen under the category of physical contiguity if we assume that they are fluent in the language of ‘doing’ but cannot make wine, art and so on. ¹³

The problem of considering professions ‘as a whole’ rather than focusing on specific activities or abilities applies to all examples displayed in Table 2. Pitch-perceiver musicians, for example, may have been born with the perceptual ability to distinguish pitches, but this only makes them practitioners/typical-IEs in two artificially delimited domains: the musical field and pitch-perceiver musical field. Human coaches and scientists also make perceptual discriminations when watching a game or examining differences between samples under a microscope. All activities involving perceptual discrimination fall under physical immersion, and thus are cases of practitioners/typical-IEs. This point will not be discussed further. In the following analysis, we will stress only what is new in each case.

Human coaches and scientists also think about the best strategy to adopt next season or about the problems faced in their experiments. The difficulty here, which is the problem underlying the whole discussion, is C&E’s attempt to neglect the contributions of embodiment in the genesis of ‘conceptual judgements’ (Todes, 2001: 8). For example, when a maintenance supervisor makes an accurate estimate of the time that it will take to repair a machine, it is not the language through which he or she communicates the information to the manager that led to the estimation but the embodied experience of (at least) witnessing several repair jobs (i.e. of experiencing physical contiguity).

Moving to the middle column of Table 2, sociologists of scientific knowledge, high-level journalists, ethnographers, anthropologists and interpreters are perhaps the best examples of IEs as imagined when this concept was first coined (C&E, 2002; Collins, 2004b). These professionals have developed specialist languages in order to talk with experts in a specialist domain or to act as ‘translators’ for those outside it. That granted, it is not correct to state that ‘linguistic socialization alone, that is, via the possession of [pure] interactional expertise has been argued to be of great importance in science studies’ (Collins, 2011: 273). Much research in science studies is based on detailed ethnographic work, which requires long periods of physical contiguity and does not rely on linguistic socialization alone – as Collins’ (2004a) own study of gravitational wave physics exemplifies. Contrary to the claim above, most of those who undertake science studies and are not scientists themselves are, at the most, special-IEs and not pure-IEs – and this analysis extends to the other professions stated above. ¹⁴

Similarly, previous analyses of AIDS activists and midwives suffer from the same problem of the lack of discernment between the types of immersion experienced in order to develop each ability. For instance, midwives can be placed under distinct types of immersion depending on their genders and, in the case of female midwives, on whether they are mothers. In any case, midwives can be special- or typical-IEs with regard to the language of ‘giving birth to a baby’, but actual midwives can never be pure-IEs!

The case of scientific project managers is slightly different because here Collins and Sanders (2007) made an effort to separate the abilities that may be based on IE from those that managers may have brought from other experiences – that is, ‘referred expertise’ (C&E, 2002: 257) – or that they must possess as managers. However, the data analysis – of an interview with one manager (Collins and Sanders, 2007) – did not factor types of immersion into the equation, limiting the breadth of data it could provide.

For instance, Collins states that it was the IE acquired by Sanders that enabled him to decide on a certain type of ‘adaptive optics’ in the Thirty Meter Telescope Project: ‘You said something very interesting … You said you could not actually design the thing but nevertheless you had the interactional expertise to sit down and argue through’ (Collins and Sanders, 2007: 629). However, the explanation given by Sanders, before Collins’ comment, illustrates the opposite:

I couldn’t design an adaptive optics system but I really do, after six to nine months in the field, I really do understand the different kinds of adaptive optics in the field … The interesting thing is that in retrospect it was easier to understand [adaptive optics] than I thought because of what I learned at LIGO about optics and control systems … From high-energy physics I brought some things; from LIGO I brought different expertises, and having that broader palette of experiences made a difference. (Collins and Sanders, 2007: 629; emphasis added)

Sanders is, at best, another case of special-IE. It took him ‘six to nine months in the field’ to acquire it, and certainly he did more than simply talking to domain specialists during this period. With respect to his ‘referred expertise’, it is clear that this pertains to activities performed in the two fields in which he worked previously. Collins and Sanders (2007) do not discuss how similar these activities are to those of the Thirty Meter Telescope Project and, therefore, how much Sanders could simply ‘transfer’ to the new field, how much he had to adapt and how much he had to forget. ¹⁵ In any case, given the above quote, we can consider Sanders to be a practitioner/typical-IE regarding these activities.

To summarize, up to this point, there is no empirical evidence of pure-IE. All examples of IEs in the literature are actually instances of special- and typical-IEs. Moreover, the need to define ‘practice’ in each case and to focus on one given activity or ability are overlooked. Thus, it is difficult to use with any precision such examples of IEs to claim the separation between embodied experiences and the acquisition of a language.

It can be argued, however, that the philosophical, in-principle claims underlying the idea of IE remain untouched. We also have yet to discuss the cases of Madeleine and the colour-blind people, which fall under the category of ‘natural language’ and will be shown to be typical-IEs with regard to ‘natural language’ (Table 2).

Interactional expertise in principle

The minimal embodiment thesis … argues that … only the minimal bodily requirements necessary to learn any language are necessary to learn the language of any community in which the organism is embedded. (C&E, 2007: 79; emphasis added)

The importance of the minimal embodiment hypothesis for our project is that it supports the strong interactional hypothesis – that fluency in the language of a domain can be acquired outside of bodily engagement with the practices of the domain. (C&E, 2007: 90; emphasis added)

The minimal embodiment hypothesis is the foundation of IE as originally conceived. If shown to be impossible in principle, the novelty of ‘linguistic socialization alone’ (Collins, 2004b: 135, 2011: 272) underlying pure-IE vanishes. Phenomenology can fulfil this role, although previous attempts (e.g. Selinger et al., 2007) have only partially succeeded, with some critics changing their minds (Selinger, 2012). At this point, we approach the problem from within: we will start from C&E’s (2007) own premises in order to show that the minimal embodiment hypothesis does not hold.

For C&E (2007), language is inseparable from the practices and bodies of those who participate in its genesis: ‘the social embodiment thesis holds that the particular language developed by any social group is related to the bodily form (or practices) of its members because bodily form affects the things they can do in the world’ (p. 79). On this point, there is no disagreement. Our point of departure from C&E is that, for us, the individual human body – or more generally ‘embodiment’ – is necessary for learning any kind of language. Moreover, C&E only focus on one single, descriptive use of language, while we also focus on its performative role. ¹⁶ These points of divergence are made clear when we look at how Collins (2004b) and C&E (2007) analyse the case of ‘Madeleine’ (Sacks, 1985), compared with our own analysis.

Madeleine is described by Collins (2004b) as a person who ‘was born blind and disabled, being unable even to use her hands to read braille’ (p. 132); on the other hand, she is taken ‘to be a person who “spoke freely indeed eloquently … revealing herself to be a high-spirited woman of exceptional intelligence and literacy”’ (quoting Sacks, 1985: 56). Accordingly, C&E (2007: 82) use her case as an empirical confirmation of pure-IE and the minimal embodiment hypothesis: Madeleine was fluent in her natural language although she ‘had a minimal “body” with almost no ability to take part in the normal activities of the members of the surrounding society’ (Collins, 2004b: 132).

What practices we single out in this case are relevant for our discussion. For C&E (2007), ‘the minimal embodiment thesis follows from the idea that the language of a domain can be learned without full physical involvement in the domain’ (pp. 79–80). That is, for them, ‘full physical involvement’ – or ‘practice’ – is involvement in a set of activities in which some individuals cannot participate because of the constitution of their bodies and the form of the practices. This also explains C&E’s focus on the descriptive use of language: if individuals with certain bodies simply cannot perform some activities, they can at best talk about them. The problem in C&E’s analysis, however, is that they disregard various daily activities in which Madeleine could participate as a perceiver and a speaker.

Madeleine was full-fledged social being. Her blindness and disability did not impede her from playing roles as a daughter, a friend and perhaps as a sister or an aunt. She had enough of a body to feel loved, hated, sad or happy. She also participated, in one way or another, when being cared for, fed, washed and carried around. Moreover, we can infer that she was able to ask for food or water when she was hungry and thirsty, to talk about the food and juice people offered her, to appreciate those who were kind and to thank them, to complain about her bad days, to participate in other people’s lives by listening and giving her opinion, to request a doctor when feeling sick and so forth. In short, Madeleine knew ‘how to do things with words’ (Austin, 1976 [1962]).

Two views of ‘embodiment’ underlie these distinct analyses of the case of Madeleine. Once again, for C&E (2007), the ‘minimal body’ one needs in order to learn a language consists of ‘those bits of the brain to do with language-processing and those bits of the body to do with language-learning and speaking: the ears, larynx, and the rest of the vocal apparatus’ (p. 78). However, what Madeleine experienced and was able to do with her body cannot be reduced to ‘those bits’ she used to interact with the world and others. This points out that C&E do not distinguish between the objective, biological body and the lived, phenomenal body.

To be human requires a phenomenal body that transcends the physiological body and enables one to ‘synchronize’ with the world (Merleau-Ponty, 2012 [1945]; Ribeiro, 2014). This is the process by which quality acquires a ‘living value’, a ‘signification for us’ and ‘my body gives a sense not only to the natural object, but moreover to cultural objects such as words’ (Merleau-Ponty, 2012 [1945]: 52). As a result, for human beings to be synchronized and cope with their lives, ‘sensing always includes a reference to the body’ (Merleau-Ponty, 2012 [1945]: 244), to what matters to us. This means that Madeleine could not have become fluent in her natural language and engage in her form of life without being an embodied human being. By using her senses and her (natural) language Madeleine could indeed participate in several of ‘the normal activities of the members of the surrounding society’.

The assumption that ‘fluency in the language of a domain can be acquired outside of bodily engagement with the practices of the domain’ (C&E, 2007: 90) does not consider distinct ways of being in and coping with the world and others and the role of the phenomenal body. It is, then, a mistaken conclusion that ‘[Madeleine] has learned the language through immersion in the world of language alone’ (C&E, 2007: 82). Collins (2004b: 140) is right in an earlier point: ‘to master interactional expertise [i.e. fluency] in a [natural] language is to master contributory expertise’; but this makes it impossible to say that ‘Madeleine … [exemplifies] the minimal embodiment thesis and the power of interactional expertise’ (C&E, 2007: 82). ¹⁷

The discussion above shows that there is no ontological difference between being fluent in natural or technical languages. The performative use of language can only be made by those who undergo physical immersion. As Collins points out, he is unable to write a paper contributing to gravitational wave physics although he considers himself to be as fluent as physicists of this domain. That which does not allow him to make such a contribution is the kind of intuition and understanding that appears only in practitioners’ language when they write a paper, but not in the language of special-IEs. ¹⁸ No one can, therefore, become fully fluent in any language through ‘linguistic socialization alone’ (Collins, 2004b: 135, 2011: 272) or even through ‘physical contiguity’ (Ribeiro, 2007a, 2007c). This is to say that only practitioners/typical-IEs can be fully fluent in any natural or technical language (Table 2).

Finally, if cases of distinctly embodied people are to be used in order to discuss expertise, the way they engage and make sense of the world must be researched in detail. For instance, colour-blind people typically are ‘close’ to or actually participate in many activities where colours are present and used; they undergo either physical contiguity or physical immersion depending on the activity. The difference is that they have a distinct way of ‘matching’ what is being said with what they can actually see and do. For instance, at a traffic light, colour-blind people see differences in hue, saturation and lightness. Although they do not actually see red, yellow and green, they ‘do see different colors at the traffic light’ (Flück, 2007: original emphasis), which are called ‘red’, ‘yellow’ and ‘green’ by colour-perceiving people.

There are also situations in which colours are so intertwined with other institutionalized practices and devices that it really does not matter whether colour-blind people see colours because they can perform the act correctly on the basis of other cues and contexts. This is so in the case of ‘Walk/Don’t Walk’ pedestrian signs, which have different pictures associated with red and green colours. These factors so sufficiently lend towards the success of colour-blind people in language use and in many daily activities that some of them only learn about their condition later in life, when very specific, practical situations reveal that their linguistic codes and visual abilities are not equivalent to those of the majority of people. In short, the ways in which distinctly embodied people cope with the world cannot be attributed solely to their linguistic fluency without further analysis. ¹⁹ However, C&E’s (2002, 2007, 2014b) claims do exactly that.

Researching interactional expertise

We now turn to whether the IG experiments can be used as ‘proof of concept’ (C&E, 2007: 98) in order to identify who has or has not developed any kind of IE in a technical domain.

The IG experiment has its origins in the Turing Test, where a ‘judge’ poses a question to both a machine and a person, each hidden from the judge, and then tries to identify which answer came from the machine and which from the person. If the judge cannot identify the machine as a machine, this means that it can be deemed intelligent. In IG, a contributory-expert judge poses questions through computers to a pair of participants, out of which one is always a contributory expert in the same field as the judge. They are all located in separate rooms, and the role of the judge is to distinguish the contributory expert from the other person. There are two experimental configurations, one in which the second participant is assumed to be an IE and another in which he or she is not; they are called, respectively, the ‘chance’ and ‘identify’ configurations (Table 3). The former is the one in which the IE will supposedly pass as a contributory expert in the linguistic exchange given his or her fluency in the technical language; here, the judge would not be able to guess who is who correctly and the result would be ‘chance’. On the other hand, the judge would be able to point out the contributory expert in the identify configuration because the other participant, who is not an IE, will give herself or himself away as lacking fluency in the language used by the judge in the questions posed.

OPEN IN VIEWER

Table 3.

The IG experimental design: expected outcomes of socialization experiments (adapted from C&E, 2007: 96).

First participant: pretends to be			Second participant: has the target expertise	Expected outcome	Judge: has the target expertise
A	Colour-blind	Imitates	Colour-perceiving	Chance	Colour-perceiver
B	Colour-perceiver	Imitates	Colour-blind	Identify	Colour-blind
C	Pitch-perceiver	Imitates	Pitch-blind	Chance	Pitch-blind
D	Pitch-blind	Imitates	Pitch-perceiving	Identify	Pitch-perceiver

There are two problems with the IG. First, it narrows down the concept of language to what can be said about a given practice. Thus, ‘participants [of the IG experiments] were asked … to think out the answers rapidly from their existing stocks of knowledge’ (Collins et al., 2006: 666; emphasis added). This issue is also present in the argument that the IG judges must be contributory experts for they must be able to assess ‘the technical contents of discourse’ (C&E, 2007: 96; emphasis added). The issue here is that language is also part of practices – as seen in Madeleine’s use of natural language.

The linguistic fluency of practitioners/typical-IEs is broader than that of pure- or special-IEs, but the IG’s decontextualized, experimental setting does not allow for testing what one does with language – besides describing a practice. For example, in the experiment conducted to check Collins’ linguistic fluency in gravitational wave physics, the judges were asked not to pose ‘mathematical questions’ because ‘mathematics … is not used much in ordinary discourse among scientists’ (C&E, 2007: 109). We may even accept this statement at face value, but the point is that one practises mathematics not only with pencil and paper but also through language. If this use of mathematical language had been included in the IG, even C&E (2007) agree that ‘Collins would have had no chance of passing the test’ (p. 109).

By delimiting the rules of engagement and the subset of language to be used by the judges (to that which fits the only language that can possibly be acquired by outsiders), the IG experimental design establishes tautological conditions for its own success. Therefore, the IG experiments cannot be used to make the general claim that ‘those with maximal interactional expertise are indistinguishable from those with contributory expertise in linguistic tests’ (C&E, 2007: 11).

The second problem with the IG is an intrinsic flaw in its design. This has to do with the ability to ‘create contrast’ – which, in turn, comes with the ability to make judgements of similarity and difference, and relevance and irrelevance (Ribeiro, 2013b: 345). If it is up to an individual to create a contrast between two forms of life and come up with adequate discriminatory questions about their specifics, he or she must be able to ‘alternate’ (Berger, 1973 [1963]) between them and, as a result, to identify points that distinguish one from another. ²⁰ For instance, because colour-blind people socialize with colour-perceiving people, they already know that the latter have the erroneous idea that colour-blind people ‘read’ traffic lights based on the location of the lights – top, middle or bottom. Because of this, colour-blind people know that this is a discriminating question posed in order to discern who is pretending to be colour-blind. In contrast, when an individual is not subjected to alternation, the difficulty in realizing the taken-for-granted practices of one single form of life translates into a difficulty in articulating such experience.

The ability of those who have experienced two or more practices to pose discriminatory questions as well as the inability of those immersed within only one practice to talk about its taken-for-granted aspects lead to an alternative way of analysing the outcomes of the IG experiments. It is therefore crucial to see first how the experiment is currently explained.

Types of immersion, perspectives and understandings

The last point of our analysis is to discuss the assumption that non-practitioners, who are pure- or special-IEs, can display the same understanding and judgemental abilities of practitioners within discursive settings (Collins, 2011). ²¹ Collins (2013), in addition to stating that ‘the very notion of what it is to “understand” is hard to grasp’ (p. 410), does not explain the similarities and differences between the understanding that a pure-IE has, compared with the understanding of those who experience physical contiguity or physical immersion. This requires readers to believe, as a matter of faith, that ‘language contains practical understanding’ and that IEs can therefore make ‘practical judgements’ (Collins, 2011: 282) – although Collins additionally does not elucidate the term ‘practical judgements’ (Ribeiro, 2013a: 389).

The arguments against the assumption that all kinds of IEs have the same linguistic skills within discursive settings change according to one’s idea of what it is to ‘understand’ a practice. Although not explicitly, C&E (2007) hold (what we may call) an ‘explicatory’ view; for them, understanding a practice is cognitive and limited to that which is amenable to explication: ‘It is the contributory experts … who define and develop the content of the language that the interactional expert tries to master’ (p. 39). Within this view, there is a purely quantitative rebuttal to the linguistic equivalence between pure- and special-IEs on one hand and practitioners on the other.

As shown, the opportunities to talk with distinct actors in different situations and develop a broader linguistic fluency increase as one moves from ‘linguistic socialization alone’ to ‘physical contiguity’ and then to ‘physical immersion’. Moreover, pure- and special-IEs do not develop the performative part of language, which is acquired only through ‘physical immersion’. Finally, even if we stick to the language necessary for describing a given practice, it is not reasonable to expect that pure- or special-IEs will ever be able to evoke from practitioners everything that can be elicited.

There is a philosophical difference between the in-principle possible claim that one (i.e. a non-practitioner) can know everything that has been said about a domain and the in-principle impossible claim that one can know everything that can be said about a domain. This means, for instance, that practitioners can draw on their experience in order to answer to new questions – ones that they have never thought of or discussed before given their taken-for-granted nature – while presumably pure- or special-IEs cannot. ²²

As a result of the above analysis, pure- or special-IEs and practitioners will not be equivalently able within discursive settings. If the understanding and the ability to make judgements depend on one’s linguistic fluency in a given practice, and the linguistic fluency of pure- and special-IEs will always lag behind that of practitioners, it follows that, within an explicatory view of understanding, full-blown practitioners will always understand more of their practice and make better technical judgements than pure- and special-IEs. Our position, however, is that there are multiple ways of ‘understanding’ a practice.

The understanding of a practice varies according to at least five criteria: (1) the type of immersion individuals experience; (2) whether they come from another ‘perspective’ and its relation to the practice (e.g. that of analysts or co-workers); (3) whether the understanding is about on-going or past practices (e.g. work analysts and historians); (4) how close or distant it is the form of life of those who want to understand the practice from that of practitioners; and (5) the purpose of such understanding.

It is not in the scope of this article to discuss the differences between the many ways of understanding a practice. We will restrict ourselves to showing that, within the bounded realm of established practices, there will always exist a difference between practitioners and non-practitioners – be they lay citizens, analysts or special-IEs. ²³ Specifically, we will show that the understanding and judgemental abilities of pure- and special-IEs with regard to a practice are not the same as those of its practitioners in two situations: when pure- and special-IEs are not analysts and when they are. Within this approach, the Japanese–Portuguese interpreters and the nickel novices are not analysts; although they experienced distinct types of immersion, they all shared the same perspective of coming to learn about the steel and nickel practices ‘from within’. On the other hand, sociologists, anthropologists, ethnographers and so forth experience ‘linguistic socialization alone’ or ‘physical contiguity’, but they do so from analysts’ perspectives.

Conclusion: Dealing with differences

We are now in a position to address the questions posed at the beginning of this article and go back to the problem that IE was meant to solve: who has the ‘right to contribute technically to a technical decision’ within public debates involving technical matters (C&E, 2002: 260).

Is interactional expertise a ‘real’ expertise? If yes, what is an interactional expert able to perform? It has been demonstrated throughout this article that special-IE is the only type of IE that a non-practitioner can develop in a technical domain. In this sense, special-IE is ‘real’ but is neither something new nor ‘can be acquired outside of [any kind of] bodily engagement with the practices of the domain’ (C&E, 2007: 90). Moreover, the linguistic fluency as well as the understanding and judgemental ability of special-IEs always lag behind that of genuine practitioners. Finally, to be fluent in a technical language, per se, does not imply having adequate discrimination and translation abilities to participate in public debates involving technical matters.

Are there any differences between contributory, interactional and referred types of expertise? What are the ‘technical’ contributions and limitations of each of them? We prefer to set aside C&E’s (2002) typology of expertise and instead say that all kinds of expertise are experienced-based. However, as experiences vary, the idea of ‘types of immersion’ (Ribeiro, 2007a, 2013a) has been used as a way to show that individuals’ abilities and understanding are both enabled by and limited to the nature of the experiences they have within a particular practice and how such experiences combine with each other. Accordingly, we have shown that the set of experiences needed to develop special-IE (physical contiguity) and contributory expertise (physical immersion) are not the same and do not lead to equivalent results. In the case of referred and contributory expertise, the nature of the experience is the same (i.e. physical immersion), but the locus in which it is developed and is to be ‘used’ are different. This implies that, even when it is possible, a ‘transfer’ of skills, understanding or even knowledge is not a straightforward matter (e.g., Lave and Wenger, 1991) and does not apply to all aspects of the new field, affecting the abilities of those changing fields. Acquiring the language of a field or having experience in similar fields is, then, not sufficient for completely opening the ‘black box’ of practitioners to non-practitioners or to those coming from other areas, even if we restrict the analysis to linguistic settings.

If we could ‘bracket out’ the power and issue-framing problems of public debates, are the three types of expertise proposed by C&E (2002, 2007) able to indicate whether someone is an expert or can make a technical contribution to a technical debate? The answer to this question is negative for two reasons. First, we could see that the abilities of IEs were overestimated because this variable was not isolated from that of sociological skills. The skills of referred experts, in turn, were normalized, as though changing fields did not affect individuals’ abilities to act in a situated way. Second, it is not possible to define what the ‘actual’ problem is and who the experts are in ‘relatively new fields’ (Wynne, 2003: 402) or when ‘quiet areas’ of science become ‘destabilized’ (Jasanoff, 2003: 395) without engaging in previous discursive activities – which takes the argument back to the starting point that C&E had attempted to avoid. Social phenomena are reflexive processes (Giddens, 1990) in which neither the actors nor the rules are defined in advance. The possibility of demarcating who can contribute to technical debates within the public sphere is then found in the permanent flux of social life in its totality. To prioritize the participation of any specific group from the very beginning of a debate limits the process of establishing the possible questions to be posed and answered. In sum, regarding public debates involving technical matters, there are no pre-established epistemic criteria to legitimize those who will have the right to speak in the first place.

The responses to the above questions show that the ‘appropriate balance’ between those who possess contributory, interactional and referred expertise (C&E, 2002: 260) cannot solve the problem of demarcation. This does not mean that there are no criteria to identify who can contribute to a technical debate. In this sense, we agree with C&E (2007) that ‘if there is to be a general criterion of expertise, experience is the leading candidate’ (p. 53). Accordingly, within the framework of types of immersion, there will always be a difference in the understanding and abilities of those who have had physical immersion within a given practice compared with those who have experienced it in other ways. This view, however, does not solve the problem of demarcation either. Even if we agree that from now on only practitioners can contribute to public debates involving technical matters, the question remains of how to define which practices and practitioners are worth including in the debates.

The central reaction to the ‘normative theory of expertise’ was not a reaction to expertise per se, but to the normative aspect of it: the idea that ‘science and technology … are the best way to distill the human experience of an uncertain world’ (C&E, 2007: 2; emphasis added). Within this view, the ‘deficit’ is on the side of the laypeople, who must become IEs in order to understand science and talk to scientists. In contrast, when the ‘deficit’ is placed on the other side, distinct proposals appear in which it is the scientists and technologists who must come closer to the everyday lives of laypeople, in order to understand the limits of their ‘laboratory worlds’, theories and abstractions in dealing with the complexity of the ‘wider world’ (Rip, 2003: 422). ²⁸

The point is that there is no ‘deficit’ in any experience because there is no completeness in any other. Every experience is perspectival and cannot handle, alone, the intertwined and complex issues found in public debates involving technical matters. Moreover, any experience is bounded by the practice from which it was born: its singularity is the source of both its power and its weakness. The problem, then, is not about acknowledging the singularity of people’s experiences – be they scientists, citizens, activists and so forth – but about giving more power to a group to the detriment of others. Within this view, the solution to the problem of demarcation is not to find a set of a priori rules to ban participants from the start. Instead, it lies in how to handle interactions between actors with distinct perspectives, experiences and abilities, without allowing these differences to engender or degenerate into power struggles – a problem that is always present when the task of identifying differences is entangled with value judgements.