People significantly underestimate AI-assisted completion times on simple cognitive tasks, even though AI assistance saves little or no actual time (the speedup illusion)¶

Claim · All claims
Evidence · 2 studies · 2 causal r3 · q3
Currency · newest study 2026 · checked against its sources 2026-10-20 (automated, 2 of 2 entries) · not reviewed by a person

Subclaims¶

q3 i? Actual AI-assisted completion time was almost one minute greater than predicted time, while actual and predicted independent times did not differ significantly. → Sunny Yu 2026

q3 i? Predicted time savings from AI (55.7 seconds) far exceed actual savings (7.5 seconds), and actual AI-assisted completion time (86.2 s) is nearly double the predicted time (43.3 s). → Sunny Yu 2026

Evidence¶

Sunny Yu 2026¶

Sunny Yu, Myra Cheng, Ahmad Jabbar, Ilia Sucholutsky, Katherine M. Collins, Dan Jurafsky, Robert D. Hawkins. (2026). Cognitive offloading and the speedup illusion in human-AI interaction. arXiv. https://arxiv.org/abs/2605.23177 doi:10.48550/arxiv.2605.23177

q3 · i? · causal · r3

Preregistered between-subjects behavioral study (prediction and completion samples, N=1237) fitting a linear mixed-effects model of predicted vs. actual times. The article reports "the actual completion time was almost one minute greater than the predicted time" for AI assistance, holding across all four task categories and both difficulty levels. Overall, actual AI-assisted and independent completion times did not differ, but AI assistance did make difficult tasks faster, by 26.14 seconds (SE = 11.9, p < .05); no difference was found for easy tasks.

"the actual completion time was almost one minute greater than the predicted time (β=57.8,SE=8.79,z=6.57,p<0.001). The speedup illusion holds across all four task categories and both difficulty levels."

Sunny Yu 2026 (2)¶

Sunny Yu, Myra Cheng, Ahmad Jabbar, Ilia Sucholutsky, Katherine M. Collins, Dan Jurafsky, Robert D. Hawkins. (2026). The efficiency-gain illusion: People underestimate the rate of AI use and overestimate its benefits on simple tasks. arXiv. https://arxiv.org/abs/2605.22687 doi:10.48550/arxiv.2605.22687

q3 · i? · causal · r3

Pre-registered Study 2 (prediction sample N=600, completion sample N=1001 on Prolific) compared predicted versus actual times for independent and AI-assisted completion. The speedup illusion held "across all four task categories and both difficulty levels" per the article.

"Specifically, participants expected AI to save more time (55.7 seconds) than it actually did (7.5 seconds) (Equation (1);β=−48.33,p<0.001). This difference primarily arises from miscalibration about AI-assisted performance, i.e., ˆtA(τ) versust A(τ): the actual completion time with AI (86.2 seconds) is nearly double the predicted time with AI (43.3 seconds) (β= 42.9,p<0.001)."

Discussion¶

Learner Variables¶

• Time and Continuity — outcome: instruction changes it

Related Claims¶

• Access to AI advice roughly doubled reported confidence in answers in one experiment — the general claim this finding bears on, with the studies behind it
• People are well-calibrated about their own independent completion times — related
• People expect AI assistance to save far more time than assistance from another highly intelligent human participant — related
• AI assistance sped up only difficult tasks and only a few individual tasks, not easy ones — related
• Participants more averse to thinking are more susceptible to the speedup illusion; AI familiarity measures do not predict calibration error — related
• Prompting is a major source of AI-assisted cognitive effort: copying prompts reduces effort but not time, and verbose model responses can make AI-assisted completion slower than predicted — related
• AI assistance reduces subjective mental effort across all tasks even when it does not reduce completion time, dissociating time and effort — related
• AI assistance provides no significant overall time savings and can slow completion on easy task variants, with chat-interface friction (prompting time) dominating on trivial tasks — related
• People overestimate how much mental effort AI assistance alleviates (offloading illusion), driven by overestimating the effort of independent completion while AI-assisted effort is accurately estimated — related
• Practicing with AI did not create an illusion of mastery: participants reported learning and skill levels similar to other conditions — related
• Educators benefit from AI chatbots through time-saving administrative assistance and improved pedagogy — reports the opposite
• Students' actual task selections deviate about half a difficulty level and half a support level from SRLTS model-based selections — related
• Prior AI use carries over: brief AI exposure increases subsequent AI adoption and exacerbates the speedup illusion, while independent exposure does not reduce AI use — related
• People use AI on simple tasks at a rate significantly higher than they predict for themselves (self-estimate miscalibration), especially on easy task variants — related
• AI-related placebo effects lead users to overestimate their own abilities while or after being assisted by an AI — related
• Registered report authors report more constructive review, greater candor about findings, and time savings — related