Doctor Versus Artificial Intelligence: Patient and Physician Evaluation of Large Language Model Responses to Rheumatology Patient Questions in a <scp>Cross‐Sectional</scp> Study
Bibliographic record
Abstract
OBJECTIVE: The objective of the current study was to assess the quality of large language model (LLM) chatbot versus physician-generated responses to patient-generated rheumatology questions. METHODS: We conducted a single-center cross-sectional survey of rheumatology patients (n = 17) in Edmonton, Alberta, Canada. Patients evaluated LLM chatbot versus physician-generated responses for comprehensiveness and readability, with four rheumatologists also evaluating accuracy by using a Likert scale from 1 to 10 (1 being poor, 10 being excellent). RESULTS: Patients rated no significant difference between artificial intelligence (AI) and physician-generated responses in comprehensiveness (mean 7.12 ± SD 0.99 vs 7.52 ± 1.16; P = 0.1962) or readability (7.90 ± 0.90 vs 7.80 ± 0.75; P = 0.5905). Rheumatologists rated AI responses significantly poorer than physician responses on comprehensiveness (AI 5.52 ± 2.13 vs physician 8.76 ± 1.07; P < 0.0001), readability (AI 7.85 ± 0.92 vs physician 8.75 ± 0.57; P = 0.0003), and accuracy (AI 6.48 ± 2.07 vs physician 9.08 ± 0.64; P < 0.0001). The proportion of preference to AI- versus physician-generated responses by patients and physicians was 0.45 ± 0.18 and 0.15 ± 0.08, respectively (P = 0.0106). After learning that one answer for each question was AI generated, patients were able to correctly identify AI-generated answers at a lower proportion compared to physicians (0.49 ± 0.26 vs 0.97 ± 0.04; P = 0.0183). The average word count of AI answers was 69.10 ± 25.35 words, as compared to 98.83 ± 34.58 words for physician-generated responses (P = 0.0008). CONCLUSION: Rheumatology patients rated AI-generated responses to patient questions similarly to physician-generated responses in terms of comprehensiveness, readability, and overall preference. However, rheumatologists rated AI responses significantly poorer than physician-generated responses, suggesting that LLM chatbot responses are inferior to physician responses, a difference that patients may not be aware of.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 teacher head, 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".