Training improved accuracy in every digital-forensics experience group and most generative-AI familiarity groups, with the largest estimated gains among analysts with no forensics experience and frequent generative-AI users¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r2 · q3
Currency · newest study 2026 · checked against its sources 2026-10-20 (automated, 2 of 2 entries) · not reviewed by a person

Subclaims¶

Reading the evidence codes

q is the design tier, i the impact, n the sample, and the kind with its r the rigour judged within that kind: what each code means.

This claim: no standardized effect is recorded in observations/ for it yet, so no benchmark reading is possible.

q3 i? The BA×Post training effect was largest among participants with no prior forensics experience (9.2 percentage points, 95% CI: [3.5, 14.2]), with smaller gains of 5.4 and 2.0 points for the less-than-one-year and more-than-one-year groups. → Negar Kamali 2026 q3 i? The estimated training effect was largest among frequent or daily generative-AI users (7.2 percentage points, 95% CI: [3.9, 10.0]), with smaller estimated gains for occasional (6.0 points) and rare (2.9 points) users whose confidence intervals include zero; estimates for participants with no prior generative-AI use were not interpreted because of the small sample. → 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

Exploratory subgroup analysis of the counterbalanced experiment, splitting participants by self-reported digital-forensics experience (roughly ten per subgroup). The no-experience group also had the lowest pre-training accuracy (62.7%) and the highest false-positive rate (43.9%).

"The training effect, estimated by the BA×Post interaction, was largest among participants with no prior forensics experience, corresponding to a 9.2 percentage point increase in accuracy (95% CI: [3.5, 14.2]). Participants with less than one year and more than one year of experience showed smaller estimated gains of 5.4 percentage points (95% CI: [0.5, 10.0]) and 2.0 percentage points (95% CI: [-3.9, 8.3]), respectively."

Negar Kamali 2026 (2)¶

q3 · i? · causal · r2

Exploratory subgroup analysis splitting participants by self-reported generative AI familiarity. The article reports the training "improved participant accuracy across most generative AI familiarity groups", with the frequent-use group gaining the most.

"The estimated training effect was largest among participants reporting frequent or daily prior use of generative AI tools, corresponding to a 7.2 percentage point increase in accuracy (95% CI: [3.9, 10.0]). Participants reporting occasional and rare use exhibited smaller estimated gains of 6.0 percentage points (95% CI: [-1.3, 13.2]) and 2.9 percentage points (95% CI: [-3.3, 9.0]), respectively."

Discussion¶

Learner Variables¶

Digital Literacy — moderator: an instructional effect differs with it

AI literacy interventions show a large, statistically significant overall positive effect on student AI literacy outcomes (g = 0.837) — the general claim this finding bears on, with the studies behind it
A 30-minute generative AI literacy training increased intelligence analysts' overall accuracy at distinguishing real from AI-generated images by 9 percentage points — related
Pilot self-efficacy gains decreased with self-reported expertise: beginners gained most (+1.18), intermediates +0.94, and advanced users +0.46, all significant — a narrower finding that bears on this claim
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
Applying PET raises accuracy for every user cohort, with the largest gain for the Low Literacy cohort — 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 — related

Page metadata

Field	Value
`type`	`claim`
`id`	`training-gains-across-experience-levels-largest-for-novices`
`title`	`Training improved accuracy in every digital-forensics experience group and most generative-AI familiarity groups, with the largest estimated gains among analysts with no forensics experience and frequent generative-AI users`
`description`	`Training improved accuracy in every digital-forensics experience group and most generative-AI familiarity groups, with the largest estimated gains among analysts with no forensics experience and frequent generative-AI users`
`status`	`review`
`evidence_strength`	`mixed`
`newest_evidence`	`2026`
`source_checked`	`2026-10-20`
`generated`	by `process:wiki-ingest` on `2026-10-10`
`trust`	unverified — no one has confirmed this page's content
`sources`	2 — `negar-kamali-2026`, `negar-kamali-2026-2`

Frontmatter is the page's machine-readable record — see the Schema & Guide.