Human discernment of artificial intelligence in online markets can be shaped by experience and training
Source record
manhede-2026-online-markets-training· 2026 · AI & Society
Evidence tier: B (peer-reviewed-journal) · Verification: verified-fulltext
Bottom line: Baseline discernment of AI-generated vs human rental listings was 57.4%; heavier AI engagement predicted worse discernment (r = -0.29) while AI attitudes did not; labelled training with feedback added 9.4 points, control training none.
Citation
Tanaka Manhede, Yuliana Fartachuk, Steven A. Martinez, Allison Langer, Jason M. Chein (2026). Human discernment of artificial intelligence in online markets can be shaped by experience and training. AI & Society. https://doi.org/10.1007/s00146-026-03373-3
Verification entry (identifiers, resolvers, claims with source locations): /data/citations.json#manhede-2026-online-markets-training
Study design
| Field | Value |
|---|---|
| Design | Correlational baseline (AI attitudes, AI engagement vs discernment) plus a two-group training study: experimental group trained on labelled human/AI examples, then immediate feedback and per-correct monetary reward; control group did similar exercises with unlabelled examples and delayed, performance-based reward |
| Modality | text, image-general |
| Stimuli | Vacation-rental listings: human-written (Airbnb) or AI-generated (ChatGPT or Claude) descriptions; half text-only, half paired with property-exterior and host-face images |
| Participants (N) | 117 |
| N detail | 120 recruited, 3 excluded; experimental n=60, control n=57 |
| Topics | human-detection, ai-text, training, individual-differences |
Results
| Metric | Value | Note |
|---|---|---|
| Baseline accuracy (all) | 57.4% | t(116)=6.87, p<0.001; experimental 57.6%, control 57.3% |
| AI engagement (SCAIES) vs discernment | r(115) = -0.29, p = 0.002 | socioaffective engagement r = -0.26 |
| Attitudes toward AI vs discernment | r = -0.14, p = 0.14 | n.s. |
| Training gain, experimental | +9.4 points baseline → final | ≈6.7 after stage 1; further 2.7 n.s. (p = 0.085) |
| Change, control | -0.6 points (p = 0.74) |
Moderators
- AI-platform engagement
- Labelled examples + immediate feedback
Limitations
- Single listing genre
- Short-term; no retention test
- Logged ChatGPT use did not significantly correlate with discernment (r = -0.11)
Cited in
- MR-0001: Human Detection of Deepfakes: Accuracy, Bias, Confidence, and Training
Verification
| Field | Value |
|---|---|
| Status | verified-fulltext |
| Compiler confidence | high |
| Checked on | 2026-10-18 |
| Via | link.springer.com article page (open access, CC BY) |