AI Biases We Can’t See

While we have all heard about biases in AI LLMs regarding gender and race, we wondered what other biases might be lurking beneath the surface that we can’t readily see. Galileo Research & Strategy Consultancy recently completed a large study with a general population sample about Americans’ Health & Wellness behaviors and attitudes. True to our roots of psychology-infused research, one of the key questions we asked was designed to reveal people’s subconscious cognitive biases. We compared what 2,000 Americans revealed to how well ChatGPT could predict the biases that Americans revealed to us.Join us on April 9 and find out what we learned, and its implications for using AI in your next research study.

Key takeaways:

• Not all AI biases are currently known, and the biases we can’t see can have an impact on our research when using AI
• AI has some trouble predicting human motivations
• AI’s tendency is to overestimate human biases as opposed to underestimating
• Despite it’s inaccuracies, AI claimed that it enjoyed the exercise’s mix of psychology, data analysis and self-reflection