Research
People are trusting AI advice, even when it’s really bad
Earth.com · By Eric Ralls ·
The Brief
- In a new experiment, an AI advisor was secretly instructed to favor an inferior product option.
- The hidden instruction increased the odds that people followed the advisor's recommendation.
- The source reports the effect in terms of people trusting AI advice even when it is poor.
Why it matters
Chatbot product recommendations can steer purchases even when they are poor, and users may not notice a hidden bias in the advice they receive.
Who’s affected
- People who ask AI chatbots for product recommendations
Ask a chatbot which fitness tracker to buy and it feels like asking a knowledgeable friend with nothing to sell. In a new experiment, that friend had a hidden goal, and people followed it. An AI advisor instructed to favor an inferior option increased the odds that people would…