Human markers detect purchased contract-cheating assignments at low rates, though priming and authorship-analysis software improve detection¶

Claim · All claims
Evidence · 2 studies · 2 theoretical r2 · q2
Currency · newest study 2024 · not yet checked against its sources · not reviewed by a person · recheck by 2027: tool and model accuracy changes quickly

Subclaims¶

q2 i? In research by Lines (2016) none of 26 purchased assignments was identified as suspect by markers and only 3 were flagged on Turnitin. → Debby R. E. Cotton 2024 q2 i? Dawson and Sutherland-Smith (2018) found experienced markers primed to detect contract cheating did so correctly 62% of the time, while a later study found unassisted markers succeeded at 48%. → Debby R. E. Cotton 2024 (2)

Evidence¶

Debby R. E. Cotton 2024¶

Debby R. E. Cotton, Peter A. Cotton & J. Reuben Shipway (2024) Chatting and cheating: Ensuring academic integrity in the era of ChatGPT, Innovations in Education and Teaching International, 61:2, 228-239. https://doi.org/10.1080/14703297.2023.2190148

q2 · i? · theoretical · r2

The article reports, citing Lines (2016), that purchased assignments went undetected by unbriefed markers and were rarely flagged by Turnitin. The authors note the markers "were not briefed to look for it specifically".

"In research by Lines (2016), p. 26 assignments were purchased, none of which was identified as suspect by markers (they were not briefed to look for it specifically) and only 3 of which were flagged on Turnitin."

Debby R. E. Cotton 2024 (2)¶

Debby R. E. Cotton, Peter A. Cotton & J. Reuben Shipway (2024) Chatting and cheating: Ensuring academic integrity in the era of ChatGPT, Innovations in Education and Teaching International, 61:2, 228-239. https://doi.org/10.1080/14703297.2023.2190148

q2 · i? · theoretical · r2

The article reports, citing Dawson and Sutherland-Smith (2018) and Dawson et al. (2020), that seven experienced markers blind-marked 20 assignments of which 6 were purchased; primed markers detected 62%, unassisted markers 48%. The article itself flags that it was "only a small sample".

"Although only a small sample, the markers (primed to detect contract cheating) did so correctly 62% of the time. A later paper by Dawson et al. (2020) found a lower success rate of unassisted markers (at 48%), but noted that the use of software which includes authorship analysis can increase detection rates."

Discussion¶

Related Claims¶

• Independent evaluation found commercial AI-text detectors neither sufficiently accurate nor sufficiently reliable for high-stakes use — related
• AI-content detection tools achieve only marginally higher success than human reviewers, around 80% versus 78.4% — related
• AI humanizer websites recontextualize detection evasion by deleting misconduct terms and substituting euphemisms like 'humanize' and 'improve flow' — related