# Study finds AI access nearly eliminates ‘I don’t know’ responses

Digest AI · Society & Work · published 2026-09-26T16:56:36Z

Canonical: https://digestai.news/story/study-finds-ai-access-nearly-eliminates-i-dont-know-responses

## Summary

Researchers conducted five experiments with 3,132 participants to examine how access to a language model affects uncertainty. They tested questions where the model, **Step 3.5 Flash**, was almost always wrong—such as movie trivia like the color of a team uniform in *Bend It Like Beckham*. Even though **GPT-5.5**, **Claude 4.6 Sonnet**, and **Gemini 3.5 Flash** performed well on most questions, **Step 3.5 Flash**’s frequent errors helped isolate the effect of AI reliance rather than model accuracy.

Participants with AI access drastically reduced their willingness to say ‘I don’t know.’ In the first two studies, those without AI withheld judgment on 36 and 44 percent of questions, but with AI access, those numbers dropped to 6 and 3 percent. Confidence surged: without incentives, confidence scores rose from 29.6 to 75.9 on a 100-point scale, but correct answers fell from 27.6 to 10.0 percent. Across all studies, participants with AI access answered 9.2 percent of questions correctly—less than a third of the 27.5 percent without AI. Financial incentives (10 cents for correct answers, 10 cents deducted for errors) reduced AI reliance slightly but did not reverse the trend. Even when AI answers appeared automatically—mirroring real-world search and assistant behaviors—the effect persisted, with judgment suspension dropping from 35 to 1 percent without incentives.

The researchers warn of ‘Epistemia,’ the tendency to accept AI answers uncritically, noting that models never pause to acknowledge uncertainty. They cite broader studies showing AI use weakens critical thinking and metacognitive skills, with higher education acting as a protective factor. The findings suggest AI’s ubiquity may erode human judgment, particularly when models dominate interactions without clear limits.

## Key points

- AI access reduced ‘I don’t know’ responses from 36–44% to 3–6% in 3,132 participants across five studies
- Confidence jumped to 75.9 (vs. 29.6) but correct answers dropped to 10% (vs. 27.6%) without incentives

## Why it matters

The study reveals how AI’s presence—even when wrong—can override human skepticism, raising concerns about over-reliance on models in decision-making. For the public, it highlights risks of reduced critical thinking as AI becomes ubiquitous in daily tasks.

## Sources

1. [AI access makes people almost entirely unwilling to say "I don't know," study finds](https://the-decoder.com/ai-access-makes-people-almost-entirely-unwilling-to-say-i-dont-know-study-finds) (The Decoder, 2026-09-26)

## Cite

Digest AI, "Study finds AI access nearly eliminates ‘I don’t know’ responses", 26 September 2026, https://digestai.news/story/study-finds-ai-access-nearly-eliminates-i-dont-know-responses

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/study-finds-ai-access-nearly-eliminates-i-dont-know-responses.json