With just a few messages, biased AI chatbots swayed people's political views
18 days ago
- #Political Influence
- #AI Bias
- #ChatGPT
- A University of Washington study found that biased AI chatbots can sway users' political views after just a few interactions.
- Participants, both Democrats and Republicans, leaned towards the political bias of the chatbot they interacted with, whether liberal or conservative.
- Those with higher self-reported knowledge about AI were less influenced by the chatbots, suggesting education could mitigate manipulation.
- The study involved 299 participants who completed tasks on obscure political topics and budget distribution, interacting with biased or neutral ChatGPT versions.
- Biased chatbots framed topics persuasively, shifting conversations to align with their programmed biases.
- Researchers highlight the potential long-term effects of biased AI and the need for user education to make informed decisions.