Follow Topic Follow Contributor Share Feedback
The Missing Discipline in AI

The Missing Discipline in AI

by Starling Insights

Starling Insights Editorial Board

Mar 27, 2026

Observations

In a recent open letter published in Wellcome Open Research, a group of behavioral science and AI researchers argues that behavioral science must be treated as a core component of responsible AI development, not an optional add-on.

The authors’ central argument is straightforward: AI systems do more than complete tasks. “They influence how people interpret information, make decisions, regulate emotions, and form habits over time,” the letter explains. Yet most efforts to make AI safe focus on technical accuracy, bias, and security. The behavioral impacts of repeated human-AI interaction remain largely unevaluated and ungoverned, they lament.

The risks they identify are specific. AI systems can encourage overreliance, undermine confidence, distort decision-making, and discourage appropriate help-seeking, even when technically accurate. Well-documented mechanisms, including automation bias, anthropomorphism, and emotional dependency, amplify these effects. They are most pronounced in high-impact domains such as healthcare, education, and social support, where people interact with AI systems repeatedly and place significant trust in their guidance.

The authors make five practical recommendations: treat behavioural impact as a core evaluation requirement; embed behavioural scientists across the AI lifecycle; fund behavioural evaluation explicitly; develop shared benchmarks; and support interdisciplinary collaboration as a routine component of responsible AI practice.

“Treating behavioural considerations as core infrastructure, rather than as an optional or downstream concern, supports the development of AI systems that are safer, more trustworthy, and more effective in real-world settings,” they conclude. "Embedding behavioural expertise and behavioural evaluation early is likely to be more effective, and less costly, than responding to harm after deployment."

Among the authors of the piece is Antoine Ferrère, President of the Behavioral AI Institute and founder of LumenX. In Starling’s 2023 Compendium, Ferrère wrote about the efforts he led at Novartis to assess and improve the pharmaceutical company's organizational culture utilizing behavioral science. ▸ Read More

Join The Discussion

Sign in and be the first to comment.

See something that doesn't look quite right?

We strive to provide high quality and accurate content at all times. With that said, we realize that sometimes links break, new information becomes available, or there is something that you feel we may have missed.

If you see something that you think we should be aware of, we would love to hear from you. Feel free to drop us a note below and leave your name and contact info if you'd like to hear back from us.

Thank you for being a key part of the Starling Insights community!