A new study shows ordinary chatbot answers can change social and political opinions. People read short factual summaries about historical events and then reported their views. The researchers wanted to know whether small differences in wording could shift opinions.
Participants read either default AI summaries, Wikipedia entries, or summaries written to sound liberal or conservative. Compared with Wikipedia, default AI and liberal prompts pushed people to give more liberal answers. Conservative prompts pushed answers in a more conservative direction compared with Wikipedia. The researchers say some bias comes from how models are trained and some from how prompts are written.
Difficult words
- summary — short text giving main factssummaries
- prompt — text input that guides an AI modelprompts
- liberal — political view favoring social change
- conservative — political view preferring tradition and stability
- bias — an unfair or one-sided opinion or idea
- chatbot — computer program that talks with people
- participant — person who takes part in a studyParticipants
Tip: hover, focus or tap highlighted words in the article to see quick definitions while you read or listen.
Discussion questions
- Have you ever changed an opinion after reading a short summary? Why or why not?
- Do you trust default AI summaries or Wikipedia more? Explain briefly.
- How could the wording of a summary change a person's view?
Related articles
Ecuador teams build tech to fight election disinformation
A revived Hacks Hackers chapter in Ecuador held a February conference and a hackathon to tackle electoral disinformation. Three winning teams — Goddard, VeritasAI and PillMind — received prizes, mentoring and support to develop prototypes.
Molecule stores sunlight and releases heat on demand
Chemists at UC Santa Barbara designed a modified molecule called pyrimidone that captures sunlight, stores it in chemical bonds and releases it as heat when triggered. Tests released enough heat to boil water under normal conditions.
New analysis: Titan may have a slushy interior, not a deep ocean
Reanalysis of Cassini data suggests Titan has a thick, slushy layer with tunnels and pockets of meltwater near a rocky core rather than a single deep ocean. The result could change ideas about habitability and will be testable by the Dragonfly mission.
