The Research Software Engineering in the Age of Generative AI: Building a Community Vision workshop, held 25-27 March in Edinburgh, UK, worked towards developing a roadmap for how research software will be produced in the age of Generative AI – independent of today’s specific roles or job titles.

The workshop built on the Alliance for Data Science and AI (ADSA) and United States Research Software Engineering Association (US-RSE)-led Position Statement on Generative AI in the RSE Workplace on how GenAI is impacting RSEs, and the vision RSEs have for their profession in this new technological landscape.

Through lightning talks, deep critique of a draft vision, and multiple rounds of focused working groups, participants explored future practices, identified opportunities and risks, and identified a range of high-impact pilots to advance safe, reproducible, and effective use of AI in research software and workflows. The areas focused on by working groups included:

• Suggesting policies and narratives for research performing institutions
• Developing a framework to document costs, benefits, and risks
• Understanding future incentives around publishing, preserving and crediting software
• Verifying and validating research software
• Defining the RSEs of the future
• Training
• Developing a playbook for RSE managers and open-source software project leaders
• Making GenAI accessible to all
• Collaborating together across people, community, and disciplines, not just with AI

Across these areas, participants identified and prioritized over 50 different activities, ranging from writing sprints and community-of-practice activities that could begin soon, to longer-term research studies to investigate how verification practices, collaboration patterns, and training needs are changing as AI tools become embedded in research workflows.

The workshop participants also developed an informal list of resources relevant to these topics, both before and during the workshop.

The 37 workshop participants were selected to represent a cross-section of early adopters and supporters of innovative AI tooling in research organisations. This representation was not exhaustive, and this was acknowledged and explored during the workshop. ReSA was supported to undertake this work as part of Schmidt Sciences grant G-25-69965, with local support from the Software Sustainability Institute.

Participants

• Samantha Ahern – UCL / SocRSE / The Carpentries
• Michelle Barker – Research Software Alliance
• Neil Chue Hong – University of Edinburgh / Software Sustainability Institute
• Ian Cosden – Princeton University
• Tina Dang – Schmidt Sciences
• Ryan Daniels – Accelerate Programme for Scientific Discovery
• Stephan Druskat – German Aerospace Center (DLR)
• Anshu Dubey – Argonne National Laboratory
• Yael Elmatad – Open Athena
• Cunliang Geng – Netherlands eScience Center
• Sandra Gesing – US Research Software Engineer Association
• Josh Greenberg – Alfred P. Sloan Foundation
• Robert Haines – The University of Manchester
• Kim Hartley – Research Software Alliance
• James Hetherington – UCL, ARC
• Toby Hodges – The Carpentries
• James Howison – University of Texas at Austin
• Daniel S. Katz – University of Illinois Urbana-Champaign
• Kamilla Kopec-Harding – University of Birmingham
• Richard Littauer – CURIOSS
• Christina Maimone – Northwestern University
• Vani Mandava – University of Washington, Scientific Software Engineering Center (SSEC)
• Rohan Marwaha – National Center for Supercomputing Applications (NCSA)
• Elle O’Brien – University of Michigan
• Adrian Price-Whelan – Simons Foundation
• Karthik Ram – UC Berkeley
• Joseph Shingleton – University of Glasgow
• Arfon Smith – Schmidt Sciences
• Sue Smith – Independent
• Nathan TeBlunthuis – University of Texas at Austin
• Anelda Van der Walt – Talarify
• Steve Van Tuyl – Alliance for Data Science and AI
• Ben van Werkhoven – Leiden University
• Kirstie Whitaker – Berkeley Institute for Data Science
• Greg Wilson – Independent
• Sherry Wu – Carnegie Mellon University
• Yo Yehudi – OLS

To stay up to date on the more detailed workshop outputs still to come, subscribe to the ReSA newsletter.