Training gains were distributed broadly across the 97 stimulus images rather than concentrated in a small subset, though stimulus-level effects were heterogeneous¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r2 · q3
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i? The training effect was distributed across images and not concentrated in a narrow subset, indicating gains do not hinge on a few highly responsive stimuli. → Negar Kamali 2026 q3 i? Of the 97 stimulus images, 65 (67%) showed post-training accuracy gains and 47 of those 65 gain images fell above the null expectation, while 32 (33%) showed declines. → Negar Kamali 2026 (2)

Evidence¶

Negar Kamali 2026¶

Negar Kamali, Candice Rockell Gerstner, Jessica Hullman, Matthew Groh. (2026). Generative AI Literacy Training Improves Intelligence Analysts' Discrimination of Real and AI-Generated Images. https://arxiv.org/abs/2606.28510 doi:10.48550/arxiv.2606.28510

q3 · i? · causal · r2

Stimulus-level analysis plotting image-ranked changes in accuracy from pre- to post-training against a null model's expected band, following Simonsohn et al.'s visualization approach. The article reports positive changes appear throughout the ranked distribution rather than in isolated outliers.

"We find that the training effect is distributed across images and is not concentrated in a narrow subset, which indicates that performance gains do not hinge on a few highly responsive stimuli."

Negar Kamali 2026 (2)¶

Negar Kamali, Candice Rockell Gerstner, Jessica Hullman, Matthew Groh. (2026). Generative AI Literacy Training Improves Intelligence Analysts' Discrimination of Real and AI-Generated Images. https://arxiv.org/abs/2606.28510

q3 · i? · causal · r2

Stimulus-level tally reported in the Discussion, cross-referencing the ranked-delta plots. The article reports 65 of 97 images gained while 32 declined, with the aggregate effect strongly positive but heterogeneous across stimuli.

"Of the 97 stimulus images, 65 (67%) showed post-training accuracy gains and 32 (33%) showed declines (Fig. 3). Of the 65 gain images, 47 fell above the null expectation under a uniform user-level effect, indicating gains larger than sampling variability alone would produce."

Discussion¶

Related Claims¶

• A 30-minute generative AI literacy training increased intelligence analysts' overall accuracy at distinguishing real from AI-generated images by 9 percentage points — related
• The training gain was driven by a 14.2-percentage-point improvement in accuracy on real images, while the gain on AI-generated images was not statistically significant — related
• Training improved discrimination sensitivity (d′) with only a small, non-significant criterion shift toward labeling images as AI-generated — related
• Training effects concentrated in portrait images (+39.8 percentage points, significant), with directional non-significant negative estimates for full-body and posed-group images — reports the opposite