Cognitive neuroscientist Sarah Baldeo spearheaded the research, which examines the interplay between AI reliance, confidence in reasoning, and perceived autonomy. The findings, published on the PsyArXiv preprint platform, suggest that users who employ AI to refine and test their thought processes gain the most utility, while those who use it for passive delegation risk undermining their own cognitive independence. The study explicitly avoids claims of clinical impairment or brain-level changes, focusing instead on behavioral patterns and the nuance of human-AI interaction.

Beyond individual performance, the manuscript raises significant concerns regarding data privacy. Baldeo argues that AI interaction logs—comprising prompts, edits, and workflow traces—constitute sensitive behavioral data that can expose a user’s decision-making style, hesitation, and professional logic. Even when names are redacted, these traces present re-identification risks. Consequently, the researchers restricted the release of raw interaction materials to comply with the Canadian TCPS 2 research-ethics framework, emphasizing that the governance of these records must evolve alongside the technology itself.