Adam’s view on AI and its specific problems

 

Adam’s key argument is that other criticisms of AI, although sophisticated, fail to look into the problem of the male dominated power structure that AI bases itself on. Therefore, Adam privileges the discussion of the problem of knowledge construction in AI and not the issue of AI’s possible success or failure.

She addresses the fact that feminist epistemology, as a science, has looked at knowledge itself, what constitutes knowledge, and who knowers are. Using these ideas, she criticises AI for she argues that "knowledge and representation of knowledge [are] at the heart of AI" (Adam, 1995, p. 356). This is a very logical argument for if AI ‘models’ human knowledge it is only reasonable to wonder which knowledge and whose knowledge is it modelling. As a consequence Adam reinforces the importance of examining AI from a cultural and not from the philosophical aspect. Questions asked by feminist epistemologists around knowledge can be used to build cultural criticism of AI. How are AI systems used to represent knowledge? Whose knowledge do they reflect? What kind of knowledge is represented?

Adam considers gender to be inscribed in AI systems. She argues that women’s knowledge is left out of AI’s thinking machines (Adam, 1998, p. 1). The knowledge in ‘symbolic’ AI — the branch of AI which believes that intelligence can be represented symbolically on a computer — consists in that which is valued by the traditional epistomological definition of knowledge. This traditional definition considers knowledge to be composed of logical propositions or rules. Adam addresses the two ‘knowledges’ as knowing how — the one knowledge which can only be acquired by practising — and knowing that — the one considered as ‘knowledge’ on AI systems which is that of propositional knowledge. Adam contends that part of women’s knowledge is common sense — the main notion of knowing how — and skills — the cultural realm of knowledge — which the symbolic AI fails to represent. Thus, the status of knowledge is denied to traditional women’s skills. Another consequence is that considering propositional knowledge as the only form of valid knowledge creates the illusion that this kind of knowledge represents the truth (Adam, 1995, p. 358). The outcome is that it becomes more difficult to criticise a system such as AI when it argues for the portrayal of truth.

In addition to arguing that propositional knowledge is the only valid form of knowledge, traditional epistemology also assumes the existence of a universal knower. Thus, this universal knower’s expertise is the only form of knowledge from which AI systems derive their data. Adam believes that there is no such thing as a universal knower and that such concept leads to the ignoring by AI systems of inequalities that exist between genders. Traditional epistemology defines the universal knower as "a rational individual, in a ‘normal’ situation, perceiving the world through his senses with no dependence on others for knowledge of the world" (Adam, 1998, p. 69). The absurdity of such an assumption lies not only in the fact that this definition depends on highly abstract terms — such as ‘normal situation’ or even ‘knowledge’ — but also in the fact that being social beings that we are gives us little chance of acquiring knowledge ‘independently’ of others. Assuming a universal knower is also quite comfortable for AI designers since it gives no one the responsibility over the knowledge that AI systems will perpetuate. Therefore, by assuming a universal knower, symbolic AI functions to perpetuate male dominated concepts around who has and who does not have knowledge.

Despite such criticism Adam is not against the idea of AI. She accepts the possibility of AI’s success and exemplifies by stating not only the existence of various AI systems in use today but also an example of how AI is to be used by women (Adam, 1995, p. 356). Adam and Furnival in their internal report Designing Intersections: Designing Subjectivity: Feminist Theory and Praxis in a Sex Discrimination Legislation System (1995) of an AI system that would inform women of their rights. It would also allow for feminist legal theory to find room in including women’s experience in the development of legal knowledge (Adam, 1995, p.356). It is important to note that Adam recognises that such use of AI by women is quite contradictory in that they are built "just as disembodied and unsituated" as traditional AI projects (Adam, 1998, pp. 163-164).

 

~ Intro ~ Alison Adam ~ Artificial Intelligence ~ Adam's views ~ Feminist Epistemology ~ Expert Systems ~

~ Strengths ~ Weaknesses ~ Traditional Epistemology ~ Works Cited ~

 

 

 

 


Valentina Mello Ferreira Pinto
Communication Studies/Humanities Double Major

Communication Studies Program, Social Science Division
York University, Toronto, Ontario, Canada, M3J 1P3