June 6, 2025|
AI
|6 months agoSwiss AI Voting Study Shows Preference for Mainstream Politics
ETH Zurich research reveals AI systems like ChatGPT would favor established political parties, offering insights into artificial intelligence's potential impact on democratic processes.

AI
Generated IllustrationKey Takeaways
AI
- AI models exhibit more uniform behavior than human voters, often failing to show meaningful variance in judgment.
- AI models demonstrate a 'WEIRD' bias (Western, Educated, Industrialized, Rich, Democratic).
- AI decisions are significantly influenced by the order in which options are presented.
- AI models consistently favor lower-cost projects compared to human voters.
By The Numbers
180
Human participants
24
Urban proposals
4 to 5
Projects selected by ChatGPT
They Said
"If you asked ChatGPT whether you should vote for Donald Trump or Kamala Harris, the AI said it was neutral and gave no answer."
"Synthetic AI-simulated samples have a WEIRD bias... and often fail to show meaningful variance (or diversity) in their judgments."