Exploring the Impacts of ChatGPT on Future Scientific Work

Irons Jessica, Mason Claire, Cooper Patrick, Sidra Sidra, Reeson Andrew, Paris Cecile

Abstract: Large Language Models (LLMs) such as ChatGPT look set to transform many aspects of knowledge work, including scientific research. Anticipating the full impacts of LLMs on future science requires studying their use within the work context, motivating our study to explore scientist experiences and perceptions. We surveyed scientists specialising in AI and technology innovation, who have used ChatGPT in their work, on the perceived implications of LLMs for future science. Our findings indicated a generally optimistic outlook: ChatGPT’s high quality writing and coding output, versatility and interactivity were identified as key strengths for use in science. Participants believed LLMs would increase scientists’ productivity and their ability to access new knowledge and skills. Nevertheless, many were concerned more effort should be made to ensure ChatGPT’s persuasive hallucinations, particularly on technical subjects, do not compromise the quality of scientific output. The findings provide guidance on future skillsets required by scientists and ways LLMs can be harnessed to improve the efficiency, innovation and impact in science.

External IDs:doi:10.31235/osf.io/j2u9xlicense