Guest and van Rooij: AI in Psychology Is Research Misconduct

A preprint published on PsyArXiv in October 2025 stakes out unusually blunt ground: the uncritical adoption of AI in psychological research is not a methodological shortcut or a grey area. According to Olivia Guest and Iris van Rooij, it is research misconduct. The authors frame the paper as a literacy intervention for a discipline they see as increasingly vulnerable to inflated claims about what AI systems can do.

The paper identifies five areas where psychologists are, in the authors' view, ceding too much to opaque AI systems: replacing human research participants with AI, outsourcing programming and writing tasks, delegating scientific theorizing, and treating human-AI interactions as equivalent to genuine human relationships in therapeutic, educational, medical, or personal contexts. Each represents a transfer of a core scientific function to a system whose outputs resist independent scrutiny.

Opacity is central to the critique. Guest and van Rooij warn that "formal models...have a potential obfuscatory power that is weaponisable," meaning the apparent rigor of AI outputs can substitute for genuine scientific understanding while appearing credible. They also argue that the term "AI" itself lacks clear meaning, a premise that frames the rest of the argument: if a field cannot define what it is adopting, it cannot evaluate whether it should.

The scope is intentionally broad. The concern applies to psychologists across subdisciplines, from computational modellers and social and personality researchers to cognitive neuroscientists, experimentalists, methodologists, and theoreticians.

The honest caveat is that what was retrievable from this preprint does not spell out which AI applications, if any, the authors regard as acceptable in research, or how the "research misconduct" standard would be operationalized by journals, ethics boards, or funding bodies. That is the work this paper appears to be opening rather than closing, and the question major psychology associations will eventually have to answer.