Fooled twice: People cannot detect deepfakes but think they can
Source record
kobis-2021-fooled-twice· 2021 · iScience
Evidence tier: B (peer-reviewed-journal) · Verification: verified-abstract (compiler confidence: high)
Bottom line: People could not reliably detect deepfake videos, were biased toward judging them authentic, overestimated their ability, and neither incentives nor awareness raising helped.
Citation
Nils C. Köbis, Barbora Doležalová, Ivan Soraperra (2021). Fooled twice: People cannot detect deepfakes but think they can. iScience, 24(11), 103364. https://doi.org/10.1016/j.isci.2021.103364
Verification entry (identifiers, resolvers, claims with source locations): /data/citations.json#kobis-2021-fooled-twice
Study design
| Field | Value |
|---|---|
| Design | Pre-registered online behavioural experiment; real vs deepfake video classification; manipulations of financial incentives and awareness; confidence ratings |
| Modality | video |
| Stimuli | Face-swap deepfake videos and authentic videos |
| Participants (N) | 210 |
| Topics | human-detection, overconfidence, response-bias, incentives |
Results
| Metric | Value | Note |
|---|---|---|
| Detection | not reliable | exact accuracy not yet checked |
| Response bias | deepfakes mistaken for authentic more than the reverse | 'seeing-is-believing' heuristic |
| Financial incentives | no improvement | |
| Awareness raising | no improvement | |
| Calibration | overconfident |
Limitations
- Single generator family (2020-era face swaps)
Cited in
- MR-0001: Human Detection of Deepfakes: Accuracy, Bias, Confidence, and Training
Verification
| Field | Value |
|---|---|
| Status | verified-abstract |
| Compiler confidence | high |
| Checked on | 2026-10-07 |
| Via | Semantic Scholar API abstract |
Values on this page have not all been checked against the full text. Confirm them against the primary source before high-stakes use.