MétaCan
Menu
Back to cohort

Performance of an Artificial Intelligence Chatbot in Ophthalmic Knowledge Assessment

2023· article· en· W4367175039 on OpenAlex

Why this work is in the frame

A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.

affAt least one author lists a Canadian institution in the pinned OpenAlex snapshot.

Bibliographic record

VenueJAMA Ophthalmology · 2023
Typearticle
Languageen
FieldMedicine
TopicArtificial Intelligence in Healthcare and Education
Canadian institutionsSt. Michael's HospitalUniversity of TorontoWestern University
Fundersnot available
KeywordsMedicineChatbotCertificationMultiple choiceBoard certificationMedical educationCurriculumFamily medicineSignificant differenceArtificial intelligencePsychologyInternal medicineComputer scienceContinuing medical educationPedagogy

Abstract

fetched live from OpenAlex

Importance: ChatGPT is an artificial intelligence (AI) chatbot that has significant societal implications. Training curricula using AI are being developed in medicine, and the performance of chatbots in ophthalmology has not been characterized. Objective: To assess the performance of ChatGPT in answering practice questions for board certification in ophthalmology. Design, Setting, and Participants: This cross-sectional study used a consecutive sample of text-based multiple-choice questions provided by the OphthoQuestions practice question bank for board certification examination preparation. Of 166 available multiple-choice questions, 125 (75%) were text-based. Exposures: ChatGPT answered questions from January 9 to 16, 2023, and on February 17, 2023. Main Outcomes and Measures: Our primary outcome was the number of board certification examination practice questions that ChatGPT answered correctly. Our secondary outcomes were the proportion of questions for which ChatGPT provided additional explanations, the mean length of questions and responses provided by ChatGPT, the performance of ChatGPT in answering questions without multiple-choice options, and changes in performance over time. Results: In January 2023, ChatGPT correctly answered 58 of 125 questions (46%). ChatGPT's performance was the best in the category general medicine (11/14; 79%) and poorest in retina and vitreous (0%). The proportion of questions for which ChatGPT provided additional explanations was similar between questions answered correctly and incorrectly (difference, 5.82%; 95% CI, -11.0% to 22.0%; χ21 = 0.45; P = .51). The mean length of questions was similar between questions answered correctly and incorrectly (difference, 21.4 characters; SE, 36.8; 95% CI, -51.4 to 94.3; t = 0.58; df = 123; P = .22). The mean length of responses was similar between questions answered correctly and incorrectly (difference, -80.0 characters; SE, 65.4; 95% CI, -209.5 to 49.5; t = -1.22; df = 123; P = .22). ChatGPT selected the same multiple-choice response as the most common answer provided by ophthalmology trainees on OphthoQuestions 44% of the time. In February 2023, ChatGPT provided a correct response to 73 of 125 multiple-choice questions (58%) and 42 of 78 stand-alone questions (54%) without multiple-choice options. Conclusions and Relevance: ChatGPT answered approximately half of questions correctly in the OphthoQuestions free trial for ophthalmic board certification preparation. Medical professionals and trainees should appreciate the advances of AI in medicine while acknowledging that ChatGPT as used in this investigation did not answer sufficient multiple-choice questions correctly for it to provide substantial assistance in preparing for board certification at this time.

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.

Full frame distilled prediction

Teacher imitation

Not 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.

metaresearch head score (Codex)0.001
metaresearch head score (Gemma)0.000
Version: codex-gemma-dda1882f352aValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Observational · Consensus signal: none
GenreCandidate signal: Empirical · Consensus signal: Empirical
Teacher disagreement score0.532
Threshold uncertainty score0.610

Codex and Gemma teacher scores by category

CategoryCodexGemma
Metaresearch0.0010.000
Meta-epidemiology (narrow)0.0000.000
Meta-epidemiology (broad)0.0000.000
Bibliometrics0.0010.001
Science and technology studies0.0000.000
Scholarly communication0.0000.000
Open science0.0000.000
Research integrity0.0000.000
Insufficient payload (model declined to judge)0.0000.000

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.

Opus teacher head0.240
GPT teacher head0.491
Teacher spread0.252 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it