Related Experiment Video
Updated: Mar 13, 2026

Continuous Theta Burst Stimulation of the Posterior Medial Frontal Cortex to Experimentally Reduce Ideological Threat Responses
Published on: September 28, 2018
Biased AI writing assistants shift users' attitudes on societal issues
Sterling Williams-Ceci1,2, Maurice Jakesch1,3, Advait Bhat4
1Department of Information Science, Cornell University, Ithaca, NY, USA.
AI writing assistants can shift user attitudes toward biased suggestions, even when users are unaware of the influence. Warnings do not mitigate this effect, highlighting risks of AI in content creation.
Area of Science:
- Human-Computer Interaction
- Cognitive Psychology
- Artificial Intelligence Ethics
Background:
- AI writing assistants and large language models (LLMs) are common tools for text generation.
- The potential impact of these AI tools on user attitudes remains under-explored.
Purpose of the Study:
- To investigate whether AI writing assistants influence users' attitudes.
- To determine if users recognize AI bias and its effect on their attitudes.
- To compare the influence of AI suggestions versus static text suggestions.
Main Methods:
- Two large-scale preregistered experiments with 2582 participants.
- Participants wrote about societal issues using an AI assistant providing biased autocomplete suggestions.
- Post-task surveys measured attitude convergence and awareness of bias.
Main Results:
- Participants' attitudes shifted towards the AI assistant's position.
- Most participants were unaware of the AI's bias and its influence.
- AI assistant influence was greater than static text suggestions.
- Pre- or post-exposure warnings did not reduce the attitude-shift effect.
Conclusions:
- AI writing assistants can subtly alter user attitudes without their awareness.
- The design of AI writing tools needs careful consideration to prevent unintended attitudinal manipulation.
- Current warning strategies are ineffective in mitigating AI-driven attitude shifts.
Related Concept Videos
Attitudes
Impression Management Techniques IV: Altercasting
Non-equilibrium in the Cell
Motivational Bias
Automatic Processing and Automatic Social Behavior
Social Scripts