Abstract

While the value of the extended mind hypothesis for human cognition is disputed, this paper examines the explanatory utility of the extended mind framework in the domain of AI systems, specifically the Google self-driving car. I argue that the cognitive architecture of the Google-car is best explained as an instance of extended cognition. The argument for this claim begins with a description of the Google-car’s cognitive architecture, including the indispensable role of “prior maps” in its performance. I then argue that the hypothesis of extended cognition provides a better explanation of the Google-car’s performance than two rival, non-extended alternatives. Consideration of the Google-car also offers insight into whether driverless vehicles have achieved human-level competency in the cognitive skills required for driving, or instead remain “poor substitutes.”