Can Residency Programs Detect Artificial Intelligence Use in Personal Statements?
Bibliographic record
Abstract
OBJECTIVE: To evaluate widely used artificial intelligence (AI) detectors' ability to identify ChatGPT's (OpenAI, San Francisco, CA, USA) use in personal statements submitted as part of the residency program application. MATERIALS AND METHODS: This qualitative analysis was performed to evaluate the ability of three different AI detectors to detect the use of AI in personal statements submitted as part of residency applications for obstetrics and gynecology. A total of 25 writings were selected and analyzed by GPTZero (Princeton, NJ, USA), Undetectable AI (Sheridan, WY, USA), and Winston AI (Montreal, Quebec, Canada). RESULTS: In total, 25 separate writing samples of approximately 700 words were entered into three different AI detectors. AI-generated works had high rates of AI-detection, while classic literature samples had low rates of detection. Human-written personal statements before and after the availability of ChatGPT technology results were mixed, with results ranging from 64-100% and 3-100% of content appearing to be AI, respectively. DISCUSSION: AI-chatbots have been shown to produce writing that may be indistinguishable from human work and may already be commonly used to create personal statements. It is unclear who is utilizing ChatGPT in their writing, and residency programs everywhere will seek a reliable way to detect unethical usage. This study shows that available AI detectors may be able to detect AI use in applicants' personal statements, but the use of invalidated tools may harm honest applicants. CONCLUSION: Residency programs may be able to detect AI use in personal statements by utilizing AI-detection tools. Clear guidelines regarding the appropriate use of AI and authorship must be developed in order to maintain the integrity of student submissions.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.023 | 0.150 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.003 |
| Scholarly communication | 0.003 | 0.003 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".