How People Mix Their Own Thoughts With Advice

When Humans Trust Their Own Brains Over Helpful Machines

People often try to make good choices, but they don’t always use outside help the best way. In two studies, researchers asked 89 adults to guess which of two images was brighter while getting tips from seven computer helpers. Each helper behaved differently, showing unique ways of thinking and judging how sure they were.

The scientists used a statistical model that tells what the best possible mix of self‑judgment and advice would be. When they compared this ideal to what people actually did, they found that people usually rely more on their own guesses than the helpers’ suggestions. This tendency is called an egocentric bias and it makes decisions less accurate.

Another problem is that people struggle to adjust their trust level from one helper to the next. Even when they know a helper is usually very accurate, they still treat every new question the same way. This failure to change on a moment‑by‑moment basis is known as a global‑to‑local deficit.

These mistakes show up even when participants are told exactly how good each helper is and how confident they themselves feel. That means the issue isn’t a lack of information but rather how the brain processes it.

The bad mix hurts most when people talk to helpers who are actually very good. Also, how much a person’s own confidence matches reality (their metacognitive sensitivity) predicts how badly they perform. Some people are better at blending their own thoughts with advice, while others lose a lot of potential accuracy.

The research points to limits in how humans combine information from themselves and others. The findings matter for everyday choices, team work, and future systems where people rely on artificial assistants.