Assessment & Integrity
3 gains2 Traces7 problems
Every measured gain and problem on the record for this topic, newest first.Too few to weigh yet. Updated .
One square per record, darker for stronger evidence:Evidence: 0 high7 moderate5 emerging
12 records so far, too few to weigh yet. Is your study missing? Add it
AI, assessment and academic integrity
A Trace is one record with both a gain and a problem, so it counts on both sides: 5 records show a gain and 9 show a problem. 11 context records are in the record list but not counted, because they report no measured gain or problem.
| Evidence | Gain side | Problem side |
|---|---|---|
| HighStrong study design, backed by at least two independent sources | 0 | 0 |
| ModerateA measured study or an official source | 4 | 5 |
| EmergingEarly, small, unreviewed or reported second-hand | 1 | 4 |
By type of evidence: 2 interview studies; 2 benchmark or lab tests; 2 commentaries or narrative reviews; 2 controlled comparisons; 2 surveys; 2 news reports. How evidence is scored
12 records so far, too few to weigh yet. A written summary appears once a topic has at least 20 counted records, with at least 5 on each side. Is your study missing? Add it
In a study comparing AI and expert scoring, agreement on one checklist item rose from 0.56 to 0.89 after prompt tuning, and 7 of 11 items reached good or very good agreement (AC1 of 0.71 or higher).
In a study of 35 pre-university students, participants reported significantly lower anxiety in AIvaluate-mediated assessments than in face-to-face ones, and rated the system's usability as good.
On 902 medical student answers, Claude Sonnet 4 matched the human majority score closely (QWK 0.88-0.90), against 0.90-0.92 agreement among the human graders themselves.
Students surveyed in 109 countries said ChatGPT promotes cheating and plagiarism, is less reliable for classroom learning and is less useful for building critical thinking and decision-making skills.
A review of 213 journal articles from 2015 to 2025 found African contributions on AI, ethics and integrity in universities remained limited, while the United States, China and the United Kingdom led output.
In a survey of 781 Taiwanese university students, those in the Narc-Psych-Sad personality profile reported cheating with generative AI more often than those in the Machiavellian profile.
3 records
Researchers compared six large language models with three human graders on 902 student answers to 12 short-answer questions from hepatology and gastroenterology exams at Sorbonne University in Paris, collected from 2023 to 2025.
JMIR Medical Education
Researchers at Bursa Uludag University Faculty of Medicine gave 20 clinically oriented anatomy multiple-choice questions to five LLMs: ChatGPT-5, Gemini 2.5 Pro, Claude, Microsoft Copilot and DeepSeek R1.
Researchers compared traditional face-to-face performance-based assessments with sessions mediated by AIvaluate, an LLM-augmented emotionally intelligent conversational agent, using 35 pre-university students in a…
ace