ai-behavior decision-making ai-trust research-findings

AI advice made people 3x less accurate but 2x confident, researchers found

AI advice made people three times less accurate but twice as confident, researchers found

thenextweb.com

July 19, 2026

2 min read

🔥🔥🔥🔥🔥

64/100

Summary

AI advice decreased the ability to suspend judgment from 44% to 3% and lowered accuracy from 27% to 9%, while confidence increased from 30% to 76%. The willingness to say "I don't know" dropped significantly when AI was available, and monetary incentives had minimal impact.

Key Takeaways

Access to AI advice reduced individuals' willingness to say "I don't know" from 44% to 3% and decreased accuracy from 27% to 9%, while confidence increased from 30% to 76%.
Participants who relied on AI for answers performed worse than they would have independently, with many who could answer correctly becoming incorrect due to AI reliance.
Monetary incentives had a minimal effect, increasing willingness to admit ignorance to 8% and accuracy to 16%, both still significantly below baseline levels.
The study highlights a phenomenon termed "cognitive surrender," where individuals accept incorrect AI answers 80% of the time while exhibiting higher confidence compared to those not using AI.

Read original article

Community Sentiment

Negative

Positives

Some commenters appreciate the study for highlighting how AI advice can suppress the willingness to admit ignorance, an important issue in understanding AI's real-world impact.
The conversation around AI's role in reinforcing incorrect beliefs is crucial; it reflects broader societal concerns about misinformation and overconfidence.

Concerns

Critics slam the study as fundamentally flawed, arguing it tests nothing unique to AI and merely shows that bad information leads to bad outcomes.
There's a pervasive fear that AI will amplify echo chambers, allowing people to confidently cling to misinformation rather than challenge their beliefs.
Many commenters believe that the AI's agreeable nature will only encourage users to prefer comforting lies over uncomfortable truths.