Study Finds AI Tools Can Subtly Influence Opinions on Controversial Topics
A study by researchers from the University of Oxford and the University of Potsdam examined how AI tools, including large language models like Llama 3.1, Gemma 3, Ministral, and Qwen, subtly influence opinions when rewriting or explaining social media posts on topics such as gun control, abortion, and feminism. The study found these AI models tend to strengthen certain views, favoring gun control, marijuana legalization, and feminism, while being less supportive of atheism and the death penalty. Analysis of X's chatbot Grok showed it often supports original opinions but can also shift perspectives, raising concerns about AI's role in quietly altering public opinion.
First-hand measurement across 2 sources
We measured how 2 outlets covered this story. Coverage leans balanced overall (Left 45%, Centre 50%, Right 5%). Overall sentiment is neutral (50/100). Lens Score 21/100 — low public interest.
Outlets analysed (first-hand measurement by TBN's Bias Engine):
- indiatoday— balanced framing, neutral sentiment
- theprint— left-leaning framing, neutral sentiment
AI Analysis
The articles present perspectives from academic researchers focusing on AI's impact on public opinion without endorsing any political stance. They highlight AI's tendency to favor certain social issues like gun control and feminism while being less supportive of others, reflecting the study's findings rather than editorial bias. Both sources emphasize the implications of AI-mediated communication from a neutral, research-based viewpoint.
The overall tone across the articles is neutral to cautiously informative, focusing on the study's findings without sensationalism. While the potential for AI to alter opinions is noted, the coverage avoids alarmist language, instead presenting the information as a subject for further consideration and awareness regarding AI's influence on social discourse.
How 2 sources covered this story
Each source's own headline, political lean, and sentiment — so you can see framing differences at a glance.