CHICAGO, July 21, 2026 /PRNewswire/ — Cognitive neuroscientist, AI builder & best-selling author, Sarah Baldeo today announced the release of a new research on PsyArXiv, an open-access preprint platform for psychology and behavioral science research. The study, from ID Quotient Advisory Group, examines how professionals use generative AI, how AI reliance relates to confidence in reasoning, and algorithm design approaches that may stimulate cognition. The expanded manuscript reports aggregate findings from 1,923 adults who completed AI-assisted reasoning tasks.
“The question is not whether people use AI,” said Baldeo. “Our research labs are seeking to understand how nuances in AI use can be both beneficial and adverse.”
The study examines generative AI reliance, cognitive offloading, confidence in reasoning, and participant-reported autonomy. Findings suggest that interaction style of AI-users matters when it comes to perceived authorship, confidence in our own reasoning, and complexity of task outputs.
Baldeo said the findings point to a larger public issue: AI may function less like a simple replacement for cognition and more like a cognitive magnifier.
“AI does not automatically strengthen or weaken human confidence,” said Baldeo. “It can magnify the cognitive habits users bring to it. The people who benefit most are not those who delegate judgment, but those who use AI to challenge, refine, and strengthen their own reasoning.”
The updated manuscript also highlights a fast-emerging governance issue: AI interaction records are not ordinary data. Prompts, outputs, edits, overrides, workflow traces, and qualitative reflections can reveal how people think, decide, hesitate, delegate, and revise. In professional populations, those records create privacy and re-identification risks even when names or direct identifiers are removed.
“AI interaction data is behavioral data,” said Baldeo. “It reveals patterns of reasoning, confidence, workflow, and professional decision-making. Ethical AI research protects the privacy of participants as perceptions around AI use continue to evolve.”
Participant-level AI interaction materials from this study are not publicly released. The study was governed by informed consent and a Canadian research-ethics framework (TCPS 2) that restricts public release, third-party transfer, or external reanalysis of participant-level AI interaction records.
The latest manuscript expands on research methodology, highlights data-governance frameworks, updates figures and narrows the interpretation of the findings. The study does not claim to show clinical impairment, neural change, cognitive decline, or causal brain effects. The manuscript is available on PsyArXiv: https://osf.io/preprints/psyarxiv/xqnkz_v2
View original content to download multimedia:https://www.prnewswire.com/news-releases/study-suggests-that-ai-may-act-as-a-cognitive-amplifier-302830015.html
SOURCE ID Quotient Advisory Group
