The promises and pitfalls of the use of artificial intelligence and large language model tools in qualitative surgical research.

Abstract

To examine how artificial intelligence and large language model tools may support qualitative surgical research, where they may threaten rigor and trustworthiness, and what principles should guide their responsible use. This perspective reviews potential applications of artificial intelligence and large language model tools across the qualitative research process and considers their implications for analytic rigor, trustworthiness, and ethical practice. Artificial intelligence and large language model tools may be useful for bounded tasks, such as transcript summarization, data organization, preliminary code suggestion, and excerpt retrieval. However, these tools may also flatten nuance, generate hallucinated or misleading outputs, obscure interpretive processes through black-box functioning, reproduce training-data biases, and reduce researcher immersion in the data. Their responsible use depends on human oversight, verification against source data, and transparent reporting. Artificial intelligence and large language model tools may be most useful when applied to bounded supportive tasks rather than as substitutes for researcher interpretation, reflexivity, or analytic judgment.