"Researchers found that subjects accepted faulty AI reasoning 73.2% of the time, indicating a readiness to incorporate AI outputs into decision-making with minimal skepticism."
"Those with higher fluid IQ scores were less likely to rely on AI assistance and more likely to overrule faulty AI responses, highlighting variability in AI reliance."
"Cognitive surrender to AI is not inherently irrational; in certain contexts, a statistically superior AI could yield better results than human reasoning."
"As reliance on AI increases, performance aligns with AI quality, demonstrating both the potential benefits of superintelligence and the risks of cognitive surrender."
Research involving 1,372 participants and over 9,500 trials revealed that individuals accepted faulty AI reasoning 73.2% of the time, overruling it only 19.7% of the time. This indicates a tendency to view confident AI outputs as authoritative, reducing scrutiny. Participants with higher fluid IQ scores were less likely to rely on AI, while those viewing AI as authoritative were more easily misled. Although cognitive surrender to AI can have downsides, it may be rational in contexts where AI systems outperform human reasoning, particularly in complex data analysis.
Read at Ars Technica
Unable to calculate read time
Collection
[
|
...
]