Why Trust in AI Declines Once It Becomes Familiar

Researchers followed 222 Dutch university of applied sciences students for eight months as they used ChatGPT for their studies. What stands out is not that usage declined, but why it happened.

Trust, the sense of control, and what the researchers call the creepy factor all decreased as time went on. Together, these three factors explained nearly a quarter of the decline in usage, with trust being the strongest predictor.

The most striking figure is the drop in sense of control, from nearly 5 to 3.5 on a scale of up to 7. Students increasingly felt less able to decide for themselves how and when they used the chatbot, especially when institutions later introduced policies and restrictions.

Exactly the problem EduGPT solves

That is precisely where EduGPT takes a different approach. Not a generic chatbot that has rules imposed afterwards, but an AI study coach that the school itself sets up and manages, with clear boundaries from day one. The teacher remains in control, and the pupil knows exactly where they stand.

The researchers advocate for continuous monitoring and refinement of chatbot functionality instead of one-off adoption. That is exactly how we build EduGPT, with quarterly AI literacy training and a product that evolves rather than becoming overgrown with restrictions.

Source

Polyportis, A. (2024). Riding the digital wave, examining ChatGPT's role in students academic life: a longitudinal analysis. Frontiers in Artificial Intelligence, 6, 1324398. Read the full study.