Phenomenology vs Symbolic AI: Hubert Dreyfus’s Philosophy of Skill Acquisition

Abstract

A conflict between artificial intelligence (AI) researchers and phenomenologist Hubert Dreyfus arose in the 1960s and continued until the 2000s. The creators of the first AI programs believed that skill acquisition is a matter of solving problems by using particular mental representations,or heuristics. Dreyfus set out to prove that heuristics are not needed for skill acquisition because the human mind and body are capable of reacting to problematic situations in a flexible way without any mental representations. By clarifying the backstory of the conflict and analyzing the fundamental contradictions between the two theories of skill, the article shows how the phenomenology of skill acquisition originated from a critique of symbolic AI. Dreyfus developed his understanding of interconnections between mind and body in opposition to the associationism in the theories of Herbert Simon, Allen Newell and Edward Feigenbaum. He maintained that human beings have fringe consciousness, insight and tolerance of ambiguity and that they have a specific body structure and needs which make it possible to discriminate between relevant and irrelevant features in the environment and get a maximum grip of it. The author analyzes how theories of learning created within symbolic AI influenced Dreyfus’s five-stage model of skill acquisition. That model explained why programs by Simon and his colleagues achieved initial success, but it also exposed their limitations. To clarify the teleology of skill, Dreyfus explored how the idea of motor intentionality is connected with neural network modelling. Two perspectives on the role of Dreyfus in the history of AI are outlined together with the reasons why his philosophy had almost no effect on the AI community even though it was influential in the social sciences and humanities. Finally, current challenges facing the phenomenology of skill acquisition are explored.