Rethinking Human-AI interactions through the maintainability: a new criterion for Trust and Glue in the Extended Mind Thesis

Repensando as interações humano-IA por meio da mantenabilidade: um novo critério para Trust and Glue na Tese da Mente Estendida

This paper explores the implications of the Extended Mind Hypothesis (ExM), as introduced by Andy Clark and David Chalmers in 1998. Focusing on cognitive integration and the Trust and Glue criteria, we have two objectives. First, we examine how ExM reshapes perspectives on human-AI interaction, particularly by challenging the Standard model of AI. We argue that AI, as an active non-organic agent, significantly influences cognitive processes beyond initial expectations. Secondly, we propose the maintainability as a fourth criterion in the Trust and Glue criteria of ExM, in addition to trustworthiness, reliability and accessibility. This addition aims to evaluate parity in coupled systems comprehensively addressing concerns about control in human-AI interactions. We highlight the risks posed by AI’s understanding of human psychology, which may lead to unintended shifts in our objectives and choices. Our analysis concludes that maintainability serves as a critical safeguard against the evolving challenges of human-AI integration.

Keywords:
Extended Mind Hypothesis; Disruptive technologies; AI problem of control