Triage Against the Machine: Can AI reason deliberatively

In 1997 the computer Deep Blue defeated world champion chess player Garry Kasparov, heralding the "coming of age" for artificial intelligence, and presaged ever more sophisticated AI models such as ChatGPT, Llama, and Claude. However, the development of AI, often valorised as triumph over flawed human cognition, represents more a leap in computing power operating via improved algorithms using basic Boolean logic than eclipsing of human reason. More we argue such a comparison is fundamentally misplaced, particularly as it applies to political deliberation on matters of common concern. In contrast to the classic (syllogistic) model deployed by AI, recent evolutionary accounts of human reason refer to intuitive inferential modalities that function optimally in group contexts via deliberative mechanisms (Mercier and Sperber 2017). Our account of deliberative reason extends this work into discourse ethics, where mutual construction of understanding of the world informs collective decisions via mutual understanding of relevant considerations and inferential (representational) framework via situationally activated deliberative capabilities (Niemeyer et al. 2023). We contrast these capabilities with those of AI using the Deliberative Reason Index (DRI) applied to regulation of human genome editing as an ideal case—involving both scientific and discursive complexity. While AI performs well regarding specific matters, it dramatically fails to match human deliberation, either failing to anticipate the shared inferential (representational) architecture or (more significantly) the integrative capacity of human deliberators regarding relevant considerations. We discuss the implications of these findings for the potential limitations of existing AI, development of the technology, and appropriate domains of its application. References: Mercier, H. and D. Sperber (2017). The Enigma of Reason. Cambridge, MA, Harvard University Press. Niemeyer, S. J., F. Veri, J. S. Dryzek and A. Bächtiger (2023). "How Deliberation Happens: Enabling and Activating Deliberative Reasoning." American Political Science Review.