Process-Constituted Intelligence: A Shared Criterion for Humans and Machines

A recent study published on arXiv (2608.16213) suggests that intelligence should be viewed as a process—an iterative activity from which results arise—rather than the results themselves. The researchers contend that generative AI (GenAI) learns from remnants of human cognitive activities, generating outputs that reflect those remnants. Although GenAI's outputs mimic reasoning, problem-solving, and creativity, the essential processes that lead to these outputs in humans are largely missing, rendering current GenAI only weakly comparable to genuine cognition. The paper utilizes cognitive science's framework of weak versus strong equivalence, outlining strong equivalence through seven process characteristics applicable to both human and machine cognition. It warns of a symmetric risk: GenAI tools that replace human generative processes could hinder the development of vital skills. Design principles are proposed to alleviate this concern. This research is significant for the fields of artificial intelligence, cognitive science, and education, influencing the development and application of AI in creative and intellectual endeavors.