Key Concepts

Tenets of SEE
There are four tenets of SEE and a methodological rule implied by the first two

1. Language is central to practice
2. Language contains practice
The rule is `methodological interactionalism’ that concentrates on the contribution of language in learning a practice
3. Anything goes that works
4. Be reflexive in analysing expertise

Click here for full paper

Third Wave of Science Studies and SEE
In the third wave paper Collins and Evans argue that science studies needs a `third wave’ which will treat expertise as something real rather than attributed and thus encourage the development of a normative theory of expertise. This programme is given the name Studies of Expertise and Experience, or SEE. Under SEE, those who study expertise accept their role as experts on the nature and distribution of expertise. The rest of this website records the beginnings of the working out of SEE.

Problem of Legitimacy and Problem of Extension
In the third wave paper the problem of legitimacy is defined as the problem of getting the public to accept new and technological developments when they may feel they are risky or undesirable. This can or has been be solved by allowing more public discussion of such developments. The problem of extension is how far to extend the debate to the general public: should everyone be involved? Should public expertise be taken to extend to the very heart of the technological debates usually held among experts? SEE is intended to resolve this problem.

Periodic Table of Expertises
The Periodic Table is an attempt to classify all the kinds of expertise that might be brought to bear on a technological problem. The driving idea behind the Table is tacit knowledge. The starting point is ubiquitous expertise which turns on the tacit knowledge which all of us need to live in society. The next six entries give examples of some of the more salient or typical categories found in the Periodic Table.

Tacit Knowledge
This is the tacit knowledge required to live in society. It is what we come to `know’ tacitly as we become fully socialised members of a society.

Beer Mat Knowledge
This is one of three categories of knowledge that can be gained using only ubiquitous tacit knowledge; one may read a short explanation, such as was found on a beer mat, of how a hologram works.

Interactional Expertise
This is fluency in the technical language associated with a specialism. For example, it might be fluency in the language of gravitational wave physics which can be acquired in the absence of contributory expertise — which is needed to write papers or invent and build the instruments belonging to gravitational wave physics.

Discrimination is a meta-expertise. It refers to the judgements about expert claims that non-experts can (and must) make based not on detailed technical knowledge of the expert domain but on their everyday experiences of living in a particular society. Where these experiences that inform such judgements are widely distributed, we say that the corresponding meta-expertise is ubiquitous discrimination. In contrast, where the experiences that inform the judgements are restricted to specific social groups, such as residents of a particular region, clients of a particular organisation or users of a particular technology, we say that the corresponding meta-expertise is local discrimination. The feature that links the two categories is that, in each case, the technical judgements that are made do not depend on a technical understanding of the science or expertise involved. Instead, they rely on shared understandings of the way organisational cultures work; in effect social knowledge is `transmuted’ into a technical judgement.

Referred Expertise
This is an example of a `meta-expertise’ — an expertise about other expertises. There are five categories of meta-expertise. Referred expertise is typified by the role of the manager of a technical project who has gained the experience to be able of judge between competing expertises and experts as a result of working on earlier projects of a similar kind but with different technical contents (eg gravitational waves versus building big telescopes).

Transmuted Expertise
This refers to a class of meta-expertises where judgments between experts are made on the basis of criteria which are not technical — such as demeanour, consistency, etc. Political or personal judgments are transmuted into technical judgments.

Intrinsic and Extrinsic Politics
In the third wave paper a distinction is made between the demonstrated fact that politics is intrinsic to science and the attitude of mind that the scientist should adopt. It is argued that thought politics may be intrinsic to science it should never be extrinsic, that is, it should never be explicitly endorsed. To endorse the role of politics in science is to cease to act as a scientist; it is to vitiate the `form of life’ of science. For example, though Edinburgh politics may have affected the way local scientists viewed the structure of the brain, we would never say: `brain science will advance better if we arrange for Edinburgh politics to influence it more.’ The distinction between the intrinsic and the extrinsic is useful in many cases. For example the same distinction can be used to show that locus of legitimate interpretation is in one place even though the locus of interpretation is in another; the former has to do with extrinsic actions, the latter with the intrinsic nature of interpretation. The position can be understood in terms of the idea of Formative Action Types.

Formative Action Types
Philosophers’ analysis of actions has usually been informed by the need to understand such things as guilt and innocence and responsibility. The standard object of discussion is, then, the `action token,’ a particular thing done on a particular occasion. Societies and their `forms of life,’ however, are best described in terms of the types of action that can make sense within them. For example, one society may be constituted by acts such as witch divination whereas another might be constituted by actions such as taking out mortgages. To understand the society one does not need to think about particular instances only the types of action that go to constitute the society. What one thinks about are `Formative Action Types.’ Thus, it is formative of the form of life of science to try to minimise the influence of politics on one’s findings and to try to minimise the scope for interpretation of the meaning of a claim and keep the locus of interpretation close to the producer of knowledge. In the arts, the opposite might be the case and often is.

Locus of Legitimate Interpretation
Where the outputs of an expertise are made to be consumed, such as in the arts, or consumer technology, the locus of legitimate interpretation lies with the critic or consumer; it is the critic or the consumer rather than the producer who has the responsibility for making judgments of value. Where, as in science, the outputs are directed at the truth, the locus of legitimate interpretation — the responsibility for meaning — lies with the producer. This is a normative idea and is true irrespective of where the locus of interpretation (as opposed to the locus of legitimate interpretation) is found in practice.

Tacit Knowledge
The term `tacit knowledge’ was used my Michael Polanyi to refer to those things we can do without being able to explain how. It is also associated with phenomenological philosophers such as Heidegger and Merleau-Ponty who were particularly concerned with the way the `body’ enabled one to find one’s way through the world in the absence of explicit rules or calculative procedures. Hubert Dreyfus used these ideas to criticise artificial intelligence and to develop a five-stage model of learning a physical skill.

The idea of tacit knowledge underpins the Periodic Table of Expertises but it is not quite the same idea as discussed by the earlier generation of philosophers. At Cardiff we are much more influenced by Wittgenstein’s notion of form-of-life which we read as a philosophical underpinning to the sociologist’s idea of the `social collectivity.’ We claim that lots of tacit knowledge is tacit because the true location of the knowledge is society not the individual; the individual draws on socially located, or collective, tacit knowledge. The obvious example is natural language, whose form, maintenance, and evolution is a matter of society not the individual; the individual gains fluency in natural language by immersion in the language-speaking community. Language is both an example of `collective tacit knowledge’ but also an important topic within SEE. Language, like physical activity, is tacit-knowledge laden and it is the idea that there are tacit-knowledge laden specialist languages associated with particular practical expertises that gives rise to the idea of `interactional expertise.’

A number of important distinctions follow from the emphasis on the social location of knowledge. These distinctions are not made and, perhaps, cannot be made, within Polanyi’s, Heidegger’s, Merleau-Ponty’s or Dreyfus’s approaches. The distinctions are as follows:

  • Polimorphic vs mimeomorphic actions: this turns on whether successful understanding/execution of an action depends on social understanding. Some skilful actions do not depend on tacit knowledge.
  • Somatic-limit vs collective tacit knowledge: this turns on whether the tacitness of the knowledge is due to the limitations of the human body (as in balancing on a bike), or the social location of the knowledge (as in riding a bike in traffic).
  • Social vs minimal embodiment theses: this turns on the influence of the bodily form of the members of a society on the language developed by that society, as opposed to the influence of an individual’s body on the language of an individual who is embedded in that society (which is minimal).
  • Interactional vs contributory expertise: this cannot be understood without recognising that language is collective and tacit-knowledge laden.

In the absence of these distinctions the earlier approaches to tacit knowledge tend to exaggerate the role of the human body, use practical skills as the paradigm of all expertises, not deal in enough detail with the demarcation between those human actions that can be accomplished by properly programmed computers and those that cannot, do not explain the way that `intelligent computers’ perform as `social prostheses’ within social groups and do not deal with the full range of activities that are properly counted as expertises. Another problem is misunderstanding of the Turing Test, which is often rejected because it deals with language alone. But a well designed Turing Test would probe for the possession of the collective tacit knowledge that is part and parcel of linguistic fluency and is therefore quite adequate for indicating whether the `problem of artificial intelligence’ — acquiring the social sensitivity associated with collective tacit knowledge — has been solved.

What computers can do: In retrospect, all critiques of AI that have the ambition of properly understanding the abilities of computers must be able to explain their successes as well as their failures. For example, Terry Winograd explains successful systems as operating within `micro-worlds’ but the trouble is that micro-worlds are only one small example of the success of computers. Hubert Dreyfus copes with the problem by positing scientific and mathematical knowledge as being context-free. Computers, then, can handle mathematics and science because these are not subject to the problems of sensitivity to social context. But sociology of scientific knowledge has shown that science is far from context free — scientific knowledge is invested with the social: it is embedded in the micro-social life of the sciences and the social influences of society as a whole and computers cannot handle it <Computers and the Sociology of Scientific Knowledge>. Given this we need another way of explaining the success of computers. For example, if arithmetic is essentially a social activity, how does my pocket calculator work? <Artificial Experts> The answer comes in three parts: first, a close examination of what, say, a pocket calculator does shows that it does something different to a human calculator — for example, when approximation is called for; second, humans make up for what the calculator fails to accomplish, often without noticing that they are doing it — the computer is a social prosthesis; third, to explain what the calculator really is good at, a far more detailed disaggregation of types of action is needed than the crude `science vs non-science.’ This can be done with the idea of polimormorphic and mimeomorpic actions. To see how this works in a different kind of context see <The Bread-making Machine, Tacit Knowledge and Two Types of Action>

We build our list of publications on tacit knowledge with the `embedded-in-society approach’ to tacit knowledge in mind. The publication list, then, cannot be used as an exhaustive bibliography or the basis for a history of the idea of tacit knowledge but as an indication of how to use the idea of tacit knowledge for a more detailed understanding of the full range of expertises.

Polimorphic and Mimeomorphic Actions
To execute a polimorphic action successfully one must be able to draw on socially located tacit knowledge. For example, the action of greeting must take account of the social circumstances and the nuances of what it means to repeat a greeting on different occasions. Polimorphic actions are usually executed with different sets of behaviours on different occasions. Mimeomorphic actions are always executed with the same behaviours so they can be mimicked by machines. The successful execution of a mimeomorphic action does not depend on social understanding though when it is carried out `intentionally’ its meaning will be understood. Crucially, to the outsider, the action appears identical whether carried out with or without understanding.

Social and Minimal Embodiment
The social embodiment thesis holds that the language of a social group is partly a function of its collective bodily form. We take it that this is what Wittgenstein had in mind when he said `If a lion could speak we would not understand what it said.’ The minimal embodiment thesis holds that an individual does not have to share the bodily form of the social group to acquire its language. In principle, a `talking lion’ could lean human language if socialised into human society, and a human could learn `lion language.’ This also explains why the physically disadvantaged in our society are not linguistically disadvantaged (a thesis which has been tested by examining the linguistic abilities of the colour blind). To acquire a language, then, one requires only a minimal body — those parts of the body associated with language-speaking. Those that do not have those bodily components, such as the congenitally deaf, are much more disadvantaged when it comes to mastering the language.

Somatic-limit and Collective Tacit Knowledge
Some of the knowledge that humans possess has to be tacit knowledge purely because of the limitations of the human body. For example, in spite of Polanyi using it as his central example, there is nothing especially tacit about riding a bicycle, if by riding we mean balancing as we ride along. The physics of bike-balancing is understood and there are machines that can ride bikes; the problem is that humans cannot do the calculations fast enough to stay upright so they use their tacit skills. If we could calculate a billion times faster we could probably ride a bike using the rules of physics. Hence this kind of tacit knowledge is tacit only because of our somatic-limits. Tacit knowledge that is tacit because it is a matter of embedding in society (such as the informal `rules of the road’ that vary from country to country), is more than a matter of the limits of the body.

The Imitation Game and the Turing Test
In the 1950s Alan Turing based what became known as the `Turing Test’ on the imitation game. In the imitation game questions are asked of hidden persons, for example a woman and a man pretending to be a woman, and the judge tries to distinguish between them. The protocol of both Turing Test and imitation game are much more complicated than they look at first sight as explained in Chapters 13 and 14 of Artificial Experts. We now use the imitation game and variants of it to test for the possession of interactional expertise. For example, we have found that colour blind persons cannot be distinguished from colour perceivers when the two are compared in an imitation game even though the judge knows that one of the respondents is not a colour perceiver. In the same way, Collins, after years immersed in the gravitational wave physics community, was able to pass as a gravitational wave physicist. These results have been written up in Experiments with Interactional Expertise.

The Folk-wisdom View
This is the mistaken idea that because some lay persons can gain expertise, lay persons in general are experts. It is the romantic notion that `the people’ are technologically wise.

Technological Populism
Exploitation of popular sentiments by citing the existence of a scientific controversy where there is no real controversy. In the early 2000s in the UK, certain politicians’ exploitation of parents’ fears about the MMR vaccine to boost the notion of `choice’ is an example. Much more sinister political agendas, such as those that have to do with genetics and race, could be advanced under the banner of technological populism.

The explosive growth of belief in the reality of some technical danger, often bolstered by the mass media, when the true size of the technical problem is quite small. An example is the sudden panic in 2007 in the UK over bad fuel sold by supermarkets.

Comments are closed.