Beyond the <i>AJR</i>: Unpredictably Unequal Effects of Artificial Intelligence Augmentation
Bibliographic record
Abstract
of the InvestigationThe study by Yu et al. [1] revealed that artificial intelligence (AI) has a heterogeneous effect on individual radiologist performance improvement (treatment effect) in various tasks (aggregated and individual), being sometimes helpful and sometimes a hindrance.The study featured performance results of 140 radiologists without and with AI assistance across 15 chest radiograph diagnostic tasks.Analysis focused on the influence of experience-based predictors, direct measures of diagnostic skill, and AI error on radiologist performance improvement.Experience-based factors included years as a radiologist, subspecialization in thoracic radiology, and experience with AI tools.Performance improvement was measured as the difference without and with AI assistance in radiologists' predictions relative to the ground truth.The researchers found that individual performance improvement was not reliably predicted by experience-based factors and radiologist performance without AI assistance [1].The improvement in absolute error with AI assistance ranged from -1.295 to 1.440 across all pathologies.Heterogeneity in treatment effect increased for high-prevalence pathology (> 10% in dataset); the largest effect extended from -8.914 to 5.563 (IQR, 3.245).No singular trend was seen across individual pathologies.Radiologists with inferior AI-unassisted performance did not benefit more from AI assistance than those with superior performance.One factor that reliably predicted radiologist performance improvement was accuracy of AI predictions.Additionally, AI underestimation resulted in improved radiologist performance compared with overestimation, at the same margin of error.
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 | 0.006 | 0.044 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.004 | 0.007 |
| Scholarly communication | 0.006 | 0.006 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.044 | 0.039 |
| Insufficient payload (model declined to judge) | 0.007 | 0.003 |
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".