Artificial Knowing

Gender and the Thinking Machine

Alison Adam

• Publisher: Routledge
• Year: 2006
• ISBN:
  9781134793556
• Edition: 1
• Topic: Philosophy
• Subtopic: Philosophy History & Theory

Artificial Knowing challenges the masculine slant in the Artificial Intelligence (AI) view of the world. Alison Adam admirably fills the large gap in science and technology studies by showing us that gender bias is inscribed in AI-based computer systems. Her treatment of feminist epistemology, focusing on the ideas of the knowing subject, the nature of knowledge, rationality and language, are bound to make a significant and powerful contribution to AI studies.
Drawing from theories by Donna Haraway and Sherry Turkle, and using tools of feminist epistemology, Adam provides a sustained critique of AI which interestingly re-enforces many of the traditional criticisms of the AI project. Artificial Knowing is an esential read for those interested in gender studies, science and technology studies, and philosophical debates in AI.

1. FEMINIST RESOURCES

INTRODUCTION

In looking at the way that gender is inscribed in artificial intelligence systems, what resources are available from the literature on feminist theory and practice? Put another way, how should such a study locate itself within the rich texture of contemporary feminist writing? Part of the answer, but by no means the whole story, may be found in the way in which the domain of AI itself is regarded. In other words what sort of thing is AI understood to be, both from the point of view of those who work within AI, and of those who study it, whether from a philosophical or a sociological perspective? There is a certain amount of controversy amongst AI practitioners themselves as to whether the subject matter should be regarded as science or engineering.

Nevertheless, the trend in computing disciplines, in general, has been to move towards an engineering style in recent years, a style which can be broadly characterized in terms of an interest both in building artefacts and establishing what are perceived to be professional standards for the building of artefacts.¹ This is mirrored within AI, where practitioners see themselves, at least potentially, as not only contributing to the theory of their subject, but also, crucially, as building things, which may take the form of software systems or hardware systems such as robots. For these reasons, it makes sense to locate AI within the realm of engineering and technology and to look to feminist and social science studies of technology as a starting point.

Judy Wajcman (1991:14) notes that recent social science research has witnessed a move away from traditional models which see technology as applied science, where science is seen as the creative activity and where the discovery is made only then to have the more prosaic technology apply it in an essentially uncreative way. Rather than being seen as dependent on scientific activity, she argues that technologies are better viewed as having their own cultures distinct from scientific disciplines. It is also important to understand that a discipline which falls under the rubric of technology is no less creative and imaginative than a science. A great deal of the creative impulse comes from the way in which technologies are involved in the creation of artefacts. This feature is a striking part of AI. As the following chapters show, considerable imagination goes into both the development of AI models of intelligence and also the production of software based on these models.

Wajcman (ibid.) argues that technology is best understood on at least three levels. The first level sees it as a form of knowledge which includes the know how to use the technical artefacts. This kind of know how is often difficult to capture verbally, indeed, as I shall argue later at a more general level, it may be impossible to capture completely in any form of natural or artificial language. It is, however, transmitted by the education process. There is nothing mystical or magical about the idea of know-how and it is a form of knowledge which is common to all academic disciplines and is transferred by education and apprenticeship. A graphic example of know-how can be found in Collins’s (1974; 1985) study of the building of TEA-lasers in physics, while this branch of laser physics was still developing, in the 1970s. In order to build one of these new types of laser, even experienced laser physicists had to ‘sit at the feet of the masters’, in other words they had to go to the laboratories where the lasers were being built and serve a short apprenticeship, to have any hope of building a laser which would actually work. It was not sufficient to be an experienced laser scientist, and to study the published literature; a visit was essential in order to produce a working laser. This was because the TEA-laser scientists had some ‘tacit knowledge’ i.e. knowledge they did not know they had, about how to make a laser work. This is part and parcel of the idea that what is written down in the scientific textbook and what goes on in practice in the scientific laboratory, in other words the day-to-day operation of know-how, are by no means the same things. Interestingly, as chapter two describes, Collins (1990) has elaborated his views on ‘tacit knowledge’ from the laser study to apply it in an argument as to why it will not be possible to produce expert systems which encompass all of human knowledge. Know-how, knowing how or skills knowledge will be discussed again in chapter four. It is the blurring of the distinction between this type of knowledge and more formally articulable knowledge which forms a major stumbling block for symbolic AI.

In addition to technical know-how, Wajcman’s model also contains the more mundane practices and human activities of what people do and also the artefacts that are practised upon—the computers, bridges, washing machines and so on. Hence any definition of technology would have to encompass knowledge, practice and objects. But in addition to these, I argue that a discipline which specifically designates itself as ‘engineering’, and also which sees itself as on the frontiers of research, also encompasses the idea of designing and building the artefacts involved. So it becomes not just a question of the use, but also the ways in which technological artefacts are created. This means that I am defining AI as a kind of ‘research engineering’ domain as a subset of technology in general, and it is the design and creation of artefacts which marks the distinction.

AI AS ENGINEERING

Although I return to a more detailed historical discussion of the development of AI in the following chapter, it is important at this point to give some of the flavour of how the subject has developed as an engineering discipline. This is especially so as I want to make my point of departure, at least in feminist terms, gender and technology studies.

A survey of AI researchers in North America and Europe undertaken by Daniel Bobrow and Patrick Hayes (1985—quoted in Narayanan 1986), two well-respected US AI researchers, obtained some interesting results. They found a tension between, on the one hand, AI viewed as a science, essentially part of cognitive science, with its focus on theories of intelligence; and, on the other hand, AI seen as engineering technology, part of computer science, with its focus on the design and building of computer systems which simulate some aspect of intelligent human behaviour. This offers a picture of a domain which designates itself as pragmatic and functional in outlook, testing its products against human competences rather than defining what it is to be intelligent, let alone artificially intelligent. AI, in common with other aspects of computing, has moved from thinking of itself as a science towards reconstituting itself as an engineering subject and this dichotomy has some interesting implications.

The balance between AI as technology and AI as cognitive science is a particular manifestation of a more general tension in computing between a view of its subject matter as science on one hand and as engineering on the other. This can be witnessed across computing as a whole, where the setting of standards and the introduction of mathematical formality are both part of the rhetoric which forms a major part of the strategy adopted in computing, in striving to become what may be regarded as a proper engineering discipline (Shapiro 1993). The story also concerns the process of professionalization and the apparent desire of computer professionals to be seen as engineers, as opposed to something else potentially less prestigious, although this acknowledges that the status of ‘engineer’ is variable across different countries and is not particularly high in the UK. This is an understandable part of professionalization, but it is interesting that, in prestige terms, engineering might be favoured over, say, science. Some authors certainly see the move towards engineering within computing as problematic. For instance, Mike Hales (1994) argues that the CSCW (computer-supported co-operative work) community, as the radical wing of the usability movement in computing, is hovering on the brink of an acceptance that ‘engineering’ is not the way to deal with design problems in the production of computer systems which are becoming defined as social rather than narrowly technical.

As a final analysis, we might ask what point there is in a discussion of engineering when it may make very little difference to the actual practice of those who work in AI. But I believe that it is important; the culture of engineering has a profound effect both on what is done in AI, and on the rhetoric employed by its practitioners to justify what they do. It permits AI researchers to concentrate on the design process and the building of artefacts and to leave to one side more difficult philosophical questions as to the nature of intelligence. In some respects this is quite understandable, as too much consideration of difficult philosophical questions may be immobilizing from the point of view of producing working AI systems. Such a view is clearly reflected both in the design of Cyc, and in the work of the roboticist, Brooks (1991). This further reinforces an assertion which I wish to make explicit in the following chapter, namely that philosophers and other critics of AI who focus on AI as the creation of an artificial mind are looking at things on the wrong level. At least as a starting position we need to keep in mind the engineering or technical focus of AI in the building of computing systems as artefacts.

Of course in AI, as in other computing disciplines, the artefact has a curious non-physical existence; it is usually a software system, where the programs that go to make it up have been written and assembled to run on a particular type of computer (or perhaps a robot). But whereas a bridge is still there as a bridge when no one is using it, when a computer is switched off the software system has a curious way of disappearing. Even opening up the casing and inspecting the hard disk will not reveal what has just been viewed on the computer’s screen. Computer people do not usually speak in these terms; we are so used to these things that we must now distance ourselves from such familiar objects to see them in this way. Our acceptance of technological artefacts which are virtual as having a materiality then becomes interesting. Although their materiality is of a different type, we seem able to see computer software systems, bridges, washing machines and the boxes that are called computers themselves as all equally material. Nevertheless, although computer software systems are clearly subject to some sort of design process which can be construed as a type of engineering design, I think that it is the ethereal quality of computer systems which has made it difficult, in some circumstances, to convince computing professionals themselves, let alone others within more traditional engineering subjects, to accept computing’s credentials as an engineering discipline.²

GENDER AND TECHNOLOGY

If AI is to fall under the rubric of engineering and technology rather than of science, then the most obvious place to locate the present study would be within the corpus of writing on gender and technology. As will be apparent from what follows, although I start from gender and technology research, as I am certainly describing the way in which gender is inscribed in a particular technology, I cannot wholeheartedly locate my study centrally within that tradition. Part of the reason for this lies in the lack of general feminist theoretical attention to technology. However, as I discuss below, there are some excellent empirical studies and evidence that theoretical interest in gender and technology is now gaining ground. This, in turn, gives a compelling rationale for drawing on more overtly philosophical resources, especially in the shape of feminist epistemology, than might be usual in a study which centres itself within a ‘gender and technology’ paradigm. But there are no hard and fast boundaries. Transgressions and overlappings are important in forming new ways of thinking in feminist theory. And gender and technology studies and feminist epistemology are most certainly linked through their direct relationship with gender and science writings.

Until a decade or so ago, it would have barely been possible to define a separate area of gender and technology, however in recent years there has been a considerable increase in interest, including several detailed case studies, as well as continued growth on the theoretical front. There are a number of excellent introductions to gender and technology which set as at least part of their task an explanation of the distinction between science and technology in relation to gender (see Cockburn and Ormrod 1993; Faulkner and Arnold 1985; Grint and Gill 1995; Wajcman 1991: chapter one; 1995). I agree with authors such as Wajcman (1991) that there is a distinction to be made between gender and technology, and gender and science, as feminist disciplines. But because my study draws on a range of sources from both areas, I do not see a need to labour the nature of that distinction. There is, however, a particular aspect to the relationship of gender and technology and gender and science which I think needs to be made explicit for the present study. There are some circumstances where there is very little distinction to be made between technology and science. For instance, in the discussion that follows I can see no difference between the rhetoric used to get more women into science and that used to get more women into engineering/technology—indeed they are often lumped together as in the acronym WISE (women into science and engineering). But where there is a distinction in my argument and where a discussion on gender and AI needs to fall within the rubric of gender and technology, is within the notion of designing, building and using of technological artefacts which I have just described. The design/build/use triad does not appear to be a particular feature of research in gender and science.

What are the main areas of interest in contemporary gender and technology studies? The name of Cynthia Cockburn (1983; 1985; Cockburn and Ormrod 1993; Cockburn and Fürst Dilić 1994) springs immediately to mind for her well-researched and clearly argued studies of the 1980s and early 1990s. Indeed her research has been enormously influential in spawning a distinct area which can be designated as ‘gender and technology’. Cockburn’s work has been instrumental in unpacking the gendered relations in engineering technology. It would be easy to imagine that the story of the relationship of gender and technology is purely one where engineering and technology are ‘masculine’ in some unanalysed way and where women simply reject engineering technology because of its perceived masculinity. But Cockburn’s research shows the processes at work are much more subtle. Her studies involve detailed analyses of the ways in which hierarchies of skills come about and the ways in which women’s and men’s skills become defined in relation to technology. These are not fixed nor are they absolute. Women’s relationships to technology are not determinate and often are ambivalent; indeed women may desire to acquire ‘technical’ skills because of their perceived status rather than shun them because of their apparent masculinity (Henwood 1993).

Wajcman’s (1991: chapter one; 1995) writing gives a comprehensive account of the growth of gender and technology research, especially in relation to its development from gender and science studies. Her analyses point up the differences between different technologies rather than trying to impose an artificial uniformity. Separate technological disciplines are stamped by all sorts of varying interests, for example the male interests that shape reproductive technologies are not the same as those relating to technologies in the workplace (Wajcman 1995:190). Importantly, she notes that the most heavily researched areas to date are production and reproductive and domestic technologies (ibid.). Information technologies have received much less attention, although there is evidence that this is rapidly changing (see e.g. Adam et al. 1994; Green et al. 1993). In her own research, Wajcman is anxious to eschew deterministic models of technological change. She argues that technologies do not have in-built, pre-determined trajectories.

A crucial point is that the relationship between technological and social change is fundamentally indeterminate. The designers and promoters of a technology cannot completely predict or control its final uses…. For example, when, as a result of the organized movement of people with physical disabilities in the United States, buildings and pavements were redesigned to improve mobility, it was not envisaged that these reforms would help women manoeuvering prams around cities. It is important not to underestimate women’s capacity to subvert the intended purposes of technology and turn it to their collective advantage.

(Wajcman 1995:199)

In their edited collection on gender and technology, Keith Grint and Rosalind Gill (1995) argue that, in the development of feminist studies of technology, the concentration on empirical research means that the theoretical underpinnings of the gender-technology relation tend to remain underdeveloped. They argue that feminist analyses see-saw between a view which tries to sever the link between masculinity and technology on the one hand, and a view which acknowledges the force of the relationship between masculinity and technology on the other. The problem, they argue, is that studies which assume definitions of masculinity and patriarchy without explanation, tend to drift towards an essentialist position. But even so, there is evidence that this is changing. Research such as Anne Balsamo’s (1996) study of the technologized gendered body, shows that the process of grounding empirical material in a stronger theoretical base is now well under way.

THE RELATIONSHIP TO GENDER AND SCIENCE

Gender and technology research is rapidly developing, both in its range of empirical studies and in its theoretical platform, and I am aware that much of the feminist theory that I address relates to work on gender and science. This is the case both for the particular strands of feminist theory that I wish to describe in this chapter—liberal feminism, standpoint theory, postmodern feminism—and the more philosophically inclined feminism which has most directly informed my study. Despite this, it is clear that there are very large parts of feminist theory which do not have a critique of science as their basis; a major part of French feminism, for instance, revolves round a critique of language (Sellers 1991). In turn, all these theoretical stances have a considerable bearing on the development of feminist epistemologies. Once again, although feminist epistemology is by no means coincident with feminist studies of science, there is a large overlap. But there appears to be very little material which explicitly discusses epistemological matters in relation to technology, at least from a feminist position. This means that in looking at discussions of gender and epistemology, we are led, inevitably, to a discussion of gender and science research for inspiration.

A further reason to keep in mind the parallels of the twin gender and science/technology traditions is to be found in the way that the wider disciplines of social science studies of science and technology have developed in parallel and learned from each other. All this adds to the argument that the development of gender and technology studies should be seen against the backdrop of the longer established tradition of gender and science. This tradition includes the studies of Carloyn Merchant (1980), Evelyn Fox Keller (1983; 1985; 1992), Sandra Harding (1986; 1991) and Hilary Rose (1994). I found Brian Easlea’s books (1981; 1983) extraordinarily stimulating in raising consciousness about the relations of gender, science and technology in the early 1980s. It is interesting that he was possibly the only prominent male academic to publish on the subject in that period, yet he held an almost astonishingly radical feminist position. In their use of aggressive sexual and birth metaphors, he suggests that scientists’ desire to control nature, particularly in developing the atomic bomb, is a result of their inability to give birth, for which they are doomed eternally to envy women. In its enthusiasm for an idealized form of femininity his writing is akin to the ‘eco-feminism’ I describe below.

He argues that if a feminine view were spread more uniformly through men and women alike, especially in relation to child-rearing, this would bring about a solution to the aggressive domination of nature which scientists display. There is something rather appealing about Easlea’s fundamental and singular explanation, which ultimately gives women the upper hand. Yet for all the excitement of his books, feminists may feel that they have not stood the test of time. This is partly because it becomes increasingly hard to sustain a unitary explanation in the teeth of a burgeoning interest in postmodernism, making itself felt everywhere in the social sciences no less than in feminist research. Secondly, there has been an explosive growth in feminist theory offering more subtle explanations of the relationship between masculinity and femininity (which in all fairness has m...