Comparison of ChatGPT and DeepSeek on a Standardized Audiologist Qualification Examination in Chinese: Observational Study
Bibliographic record
Abstract
BACKGROUND: Generative artificial intelligence (GenAI), exemplified by ChatGPT and DeepSeek, is rapidly advancing and reshaping human-computer interaction with its growing reasoning capabilities and broad applications across fields such as medicine and education. OBJECTIVE: This study aimed to evaluate the performance of 2 GenAI models (ie, GPT-4-turbo and DeepSeek-R1) on a standardized audiologist qualification examination in Chinese and to explore their potential applicability in audiology education and clinical training. METHODS: The 2024 Taiwan Audiologist Qualification Examination, comprising 300 multiple-choice questions across 6 subject areas (ie, basic hearing science, behavioral audiology, electrophysiological audiology, principles and practice of hearing devices, health and rehabilitation of the auditory and balance systems, and hearing and speech communication disorders [including professional ethics]), was used to assess the performance of the 2 GenAI models. The complete answering process and reasoning paths of the models were recorded, and performance was analyzed by overall accuracy, subject-specific scores, and question-type scores. Statistical comparisons were performed at the item level using the McNemar test. RESULTS: ChatGPT and DeepSeek achieved overall accuracies of 80.3% (241/300) and 79.3% (238/300), respectively, which are higher than the passing criterion of the Taiwan Audiologist Qualification Examination (ie, 60% correct answers). The accuracies for the 6 subject areas were 88% (44/50), 70% (35/50), 86% (43/50), 76% (38/50), 82% (41/50), and 80% (40/50) for ChatGPT and 82% (41/50), 72% (36/50), 78% (39/50), 80% (40/50), 80% (40/50), and 84% (41/50) for DeepSeek. No significant differences were found between the two models at the item level (overall P=.79), with a small effect size (accuracy difference=+1%, Cohen h=0.02, odds ratio 0.90, 95% CI 0.53-1.52) and substantial agreement (κ=0.71). ChatGPT scored highest in basic hearing science (88%), whereas DeepSeek performed the best in hearing and speech communication disorders (84%). Both models scored lowest in behavioral audiology (ChatGPT: 70% and DeepSeek: 72%). Question-type analysis revealed that both models performed well on reverse logic questions (ChatGPT: 79/95, 83%; DeepSeek: 80/95, 84%) but performed moderately on complex multiple-choice questions (ChatGPT: 9/17, 53%; DeepSeek: 11/17, 65%). However, both models performed poorly on graph-based questions (ChatGPT: 2/11, 18%; DeepSeek: 4/11, 36%). CONCLUSIONS: Both GenAI models demonstrated strong professional knowledge and stable reasoning ability, meeting the basic requirements of clinical audiologists and suggesting their potential as supportive tools in audiology education. However, the presence of errors underscores the need for cautious use under educator supervision. Future research should explore their performance in open-ended, real-world clinical scenarios to assess practical applicability and limitations.
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 | 0.004 | 0.017 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 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 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".