RDA AI in Action - What is the (potential) role of generative AI in Research Software Engineering (RSE) training/practice? Opportunities, Challenges, and Outlook

Slides for the RDA AI in Action webinar February 13 @ 16:00 UTC – 17:30 UTC: https://www.rd-alliance.org/event/ai-in-action-how-researchers-leverage-ai-europe-americas-friendly-time/

Abstract:

Arguments for using Generative Artificial Intelligence (GenAI) in developing research software (RS), have mostly been based on perceived benefits such as productivity and efficiency in Research Software Engineering (RSE). Yet effective adoption that enhances software quality requires validation and broader considerations of consistency, reliability, reproducibility and maintainability. In this talk, we report on experiences conducting the GenAI4RS workshop [1], held at TU Delft’s AI Initiative Symposium (June 2024) [2], where educators and researchers explored GenAI’s role and impact in the sustainability of RS. Topics included code assistance, responsible use, and practical sessions with Copilot for RS documentation. Discussions identified three core challenges: i) technical tools lack maturity for validating and replicating outputs, ii) educational approaches do not sufficiently address domain-specific best practices and skills needed for RS sustainability, and iii) financial resources remain inadequate for experimentation, implementation, and tool development. These challenges call for a rethink on how universities integrate GenAI into RSE training, and how we recommend both software engineering and research best practices. By disclosing these experiences, we hope to contribute to the ongoing discussion in the RSE community in Europe and elsewhere. We propose establishing a RSE community of practice where educators and researchers collaborate to assess and validate GenAI model’s responses, exchange insights, and share both successes and failures.

[1]: https://zenodo.org/records/11492084

[2]: https://www.tudelft.nl/evenementen/2024/delft-ai/spring-symposium-ai-education