51±¾É«

Large Language Models (LLMs) such as ChatGPT have established themselves as integral components of knowledge work in a remarkably short time. Their low-barrier operation through natural language drastically lowers the threshold for entry – but at the same time entails a central risk: intuitive use can obscure the unreliability of such systems and lead users to uncritically adopt erroneous AI outputs, a phenomenon known as overreliance.

AI literacy – the understanding of how such systems work and where their limits lie — which enables users to deliberately question AI outputs is increasingly coming into focus as a key mechanism for the appropriate use of AI. Despite growing interest, two questions remain largely open: First, it is unclear whether interventions aimed at increasing AI literacy actually translate into changed behavior – or whether an attitude-behavior-gap between knowledge and action persists. Second, robust evidence is lacking on how sustainably the effects of such interventions endure over time.

The project addresses these gaps through behavioral, experimental studies. The goal is a theoretically and empirically grounded understanding of the mechanisms underlying such interventions – as a foundation for designing effective and persistent formats.

The anticipated insights are relevant beyond the corporate context and carry broader societal significance: a competent approach to AI also touches on questions of opinion formation and the engagement with AI-generated content in an increasingly AI-shaped information environment.

Funding body: Graduate and Professional Training Center Ulm (ProTrainU) - Ulm Postdoc Research Grant 2026

Project period: 2026-2027