When do researchers trust AI? Trust, accuracy, and experience in the use of generative AI

Mon, 20 Apr 2026 00:00:00 +0000

Generative AI (GenAI) is increasingly part of academic work – from literature searches to coding and writing. At the same time, its processes are opaque and its outputs can be misleading, so relying on it is inherently risky. How do researchers decide whether to use these tools?

What we did

Building on trust theory and a stimulus–organism–response logic, we tested whether perceived accuracy shapes researchers’ intention to use GenAI via trust, and whether this depends on their experience with GenAI. We conducted two online studies with a total of 402 researchers:

Study 1 (N = 232) used a three-wave longitudinal design with one-week lags, measuring perceived accuracy, trust, and usage intention at separate time points.
Study 2 (N = 170) was a vignette experiment in which we manipulated the accuracy of ChatGPT’s recommendations.

What we found

Researchers who perceived GenAI as more accurate trusted it more, and trust in turn predicted their intention to use it. In the experiment, more accurate recommendations increased trust.
Trust thus explained (mediated) the link between accuracy and usage intention.
Experience was a boundary condition: the more familiar researchers were with GenAI, the weaker the link between trust and usage intention.

What this means

For novices, trust works more like a heuristic – a shortcut for deciding whether to rely on a tool – and it becomes less relevant as experience grows. This suggests that, especially for less experienced users, it matters whether their trust is well-founded: helping researchers evaluate the accuracy of AI output supports calibrated rather than simply high trust.

Read the paper

Kleine, A.-K., Schaffernak, I., & Lermer, E. (2026). When do researchers trust AI? Trust as a mediator and experience as a boundary condition in researchers’ intentions to use generative AI. Computers in Human Behavior: Artificial Humans, 8, 100310. https://doi.org/10.1016/j.chbah.2026.100310

Trust in AI: Two talks and a symposium at the DGPs Congress 2024

Mon, 16 Sep 2024 09:00:00 +0200

Symposium (chair)

Navigating Trust in the Digital Realm: Trust in AI and Customer Relations.

Talk 1

Exploring the landscape of AI-based clinical decision support systems in mental healthcare: An analysis of patents and commercial products. Kleine, A.-K., Kokje, E., Cecil, J., Lermer, E., Hummelsberger, P., Heinrich, A., & Gaube, S.

Building on our patent review, we compared what is being patented with what is commercially available. The product analysis has since been published in Artificial Intelligence in Medicine ( summary).

Talk 2

The relationship between performance, trust, and the intention to use generative AI in academia.

How do researchers decide whether to rely on tools such as ChatGPT? This talk examined how the perceived performance of generative AI and trust in it shape researchers’ intentions to use it. Results from this line of research are now published in Computers in Human Behavior: Artificial Humans ( summary).

Contributions by the team

Cecil, J., Kleine, A.-K., Lermer, E., & Gaube, S. Exploring attitudes and adoption intentions of AI-enabled technologies among mental health practitioners.
Gaube, S., Biebl, I., Engelmann, M. K. M., Kleine, A.-K., & Lermer, E. Comparing preferences for skin cancer screening: A conjoint experiment.

generative AI on Anne-Kathrin Kleine