The debate about AI and cognitive decline has been loud.
New peer-reviewed research published by the American Psychological Association offers data over another opinion.
The study followed 1,923 workers across North America through 10 simulated work tasks involving planning, decision-making, and reflective reasoning.
Participants who modified, challenged, or rejected AI suggestions reported greater confidence in their own reasoning than those who accepted output with minimal revision.
“The strongest outcomes were not linked to avoiding AI, but to staying cognitively engaged while using it,” says study lead Sarah Baldeo, CEO of IT/digital transformation consultants ID Quotient and a PhD candidate in NeuroAI at Middlesex University.
“The users who questioned, refined, and overrode AI responses reported greater confidence in their own thinking because of tool usage.”
It turns out the key is using AI as a “thought partner,” with 58% of participants agreeing that “AI did most of the thinking” during their tasks.
If your organization’s workflows are built to remove people from the loop as efficiently as possible, the research suggests you’re trading short-term output for long-term capability.
The paper also pushes back on cognitive decline claims directly, finding no evidence that AI causes reduced intelligence. Active oversight builds confidence, while heavy reliance erodes it.
“AI tools that push you to question the output, weigh your options, and think out loud could make you sharper, not lazier.”
That’s a design and governance question.
Final shots
- Whether your people come out of AI-assisted work more capable or less is a leadership question. This research gives you the evidence base to make it a real one.
- The goal for AI governance in 2026 is structured engagement. Prompting guidelines, onboarding design, and workflow architecture all have a role.
