Source-linked AI summary
Report of the 2026 Workshop on Next-Generation Ecosystems for Scientific Computing: Harnessing Community, Software, and AI for Cross-Disciplinary Team Science
Lois Curfman McInnes, Dorian Arnold, Prasanna Balaprakash, Mike Bernhardt, Franck Cappello, Beth Cerny, Deborah DiazGranados, Anshu Dubey, Nichole Etienne, Roscoe Giles, Diego Gomez-Zara, Denice Ward Hood, Mary Ann Leung, Vanessa Lopez-Marrero, Olivia B. Newton, Irene Qualters, Keita Teranishi, Stefan M. Wild, Gabrielle Allen, Richard Arthur, Alexandra Ballow, Tony Baylis, David E. Bernholdt, Daniel Bielich, Johanna Cohoon, Jeremy Crampton, Charles Ferenbaugh, Stephen M. Fiore, Thomas Herault, Tanzima Islam, Stephen Jacobsohn, Meifeng Lin, Charles Lively, Satoshi Matsuoka, Stasa Milojevic, Daniel Nichols, Chris Oehmen, Santiago Ospina Tabares, Michael E. Papka, Katherine Riley, Damian Rouson, Sudip K. Seal, Brittany Segundo, John Shalf, Andrew Siegel, Valerie Taylor, Jim Willenbring, Lou Woodley
TL;DR
As AI moves from a tool within scientific workflows toward an active participant, scientific computing needs trustworthy ways to validate, trace, and govern increasingly hybrid systems. This report synthesizes workshop discussions into four strategic themes and eight priorities for building resilient, sustainable, and trustworthy AI-enabled ecosystems.
Problem
As AI-enabled workflows incorporate hybrid models, agentic systems, and human judgment, existing VVUQ methods may be inadequate for assessing uncertainty, validity, and decision implications.
Method
The report qualitatively synthesizes invited presentations, panels, breakout sessions, plenary report-outs, session notes, and participant reviews.
Results
The workshop identified four interdependent strategic themes and translated them into eight priority areas for next-generation scientific computing ecosystems.
Takeaways & Limitations
Scientific computing ecosystems must evolve people, software, data, infrastructure, institutions, incentives, and norms together to preserve rigor, transparency, accountability, and collaboration.
Takeaways & Limitations
The report disclaims any warranty regarding the accuracy, completeness, or usefulness of its disclosed information, products, or processes.
Abstract
from arXiv · showhide
Scientific computing is undergoing rapid transformation as advances in artificial intelligence, heterogeneous computing, automation, and data-intensive research reshape not only computational tools but also the institutions, workforce models, and collaborative practices that support scientific discovery. This report synthesizes insights from the 2026 Workshop on Next-Generation Ecosystems for Scientific Computing, the second in a three-year series focused on strengthening scientific computing ecosystems through socio-technical co-design. Workshop discussions identified four interdependent strategic themes: software ecosystems for AI-enabled scientific discovery; trust, validation, and traceability; human-AI teaming and paradigm shifts; and workforce, pedagogy, and governance. The report translates these themes into eight priorities for community action spanning shared research infrastructure, trust and traceability, user experience, human-AI teaming, workforce development, cross-sector coordination, stewardship and sustainability, and evaluation of scientific value. Together, these priorities outline directions for building scientific computing ecosystems that remain trustworthy, sustainable, innovative, and resilient as AI assumes a growing role in scientific work.
DOCUMENT AVAILABILITY
DOE reports produced after 1991 and many pre-1991 reports are freely available online through OSTI.GOV, while reports not in digital format have separate public and DOE-contractor access routes. The report disclaims warranties and legal responsibility for the accuracy, completeness, usefulness, or non-infringement of disclosed information and products.
- Online Access: DOE reports produced after 1991 and a growing number of pre-1991 reports are available free at OSTI.GOV.OSTI.GOV is operated by DOE’s Office of Scientific and Technical Information.
- Non-digital Reports: Reports not in digital format may be purchased by the public from NTIS.
- Non-digital Reports: Reports not in digital format are available to DOE and DOE contractors from OSTI.
- Disclaimer: The report disclaims warranties and legal liability for the accuracy, completeness, or usefulness of disclosed information, apparatus, products, or processes.It also does not represent that their use would not infringe privately owned rights.
Executive Summary: Toward Strategies for Trustworthy, AI-Enabled Scientific Computing
The workshop identified four interdependent themes for next-generation scientific computing ecosystems and translated them into eight community-action priorities. Together, these priorities seek to preserve scientific rigor and public trust while enabling faster discovery, creativity, collaboration, and efficiency.
- Strategic themes: The workshop identified four interdependent themes: AI-enabled software ecosystems; trust, validation, and traceability; human-AI teaming; and workforce, pedagogy, and governance.The workshop was the second in a three-year series focused on next-generation scientific computing ecosystems.
- Strategic themes: Weakness in any theme can undermine the ecosystem, as unvalidated tools, insufficient oversight, or unsupported technical advances can reduce trust, obscure errors, or hinder sustainability.The report links scientific reliability to validation, traceability, human oversight, skills, governance, incentives, and collaborative practices.
- Community priorities: The report translates the themes into eight priorities for community action to help ecosystems adapt to technological change while preserving scientific rigor and public trust.The priorities address collective adaptation rather than technical progress alone.
- Community priorities: Future ecosystems must support not only greater speed and efficiency but also deeper creativity, stronger collaboration, and more reliable pathways to discovery.This central message connects technological transformation with broader scientific and collaborative outcomes.
1 Background and Motivation
Scientific computing is shifting toward agentic, socio-technical ecosystems in which AI, technical infrastructure, people, organizations, and governance must evolve together. The 2026 workshop builds on earlier work by identifying four strategic themes and translating them into coordinated priorities for trustworthy, adaptive, and sustainable discovery.
- Workshop series, objectives, and progression: The 2026 workshop advances a three-year socio-technical co-design series from identifying challenges toward strategic priorities for AI-enabled scientific work.It refines earlier challenge areas, elevates trust and governance as explicit ecosystem requirements, and translates discussion into eight priorities.
- Scientific computing ecosystems as socio-technical systems: Scientific computing ecosystems combine interconnected technical, social, and organizational structures that must sustain themselves, respond to change, and improve over time.The report defines ecosystems across hardware, software, data, models, workflows, people, community practices, incentives, institutions, and governance.
- From AI tools to agentic scientific ecosystems: AI is shifting from a tool embedded in scientific workflows toward an active participant that generates hypotheses and code, proposes workflow steps, coordinates tasks, and assists interpretation.The report characterizes this transition as movement from a “tool era” toward an “agentic workflow era.”
- From AI tools to agentic scientific ecosystems: As AI assumes larger roles in research, the workshop emphasizes preserving and improving scientific quality through valid, useful, scrutinizable, and reproducible insights.The objective is framed as better science rather than simply faster science, with visible assumptions and uncertainty.
- Workshop approach and synthesis process: The workshop synthesized its discussions into four interdependent strategic themes through presentations, panels, facilitated breakouts, plenary report-outs, and review of session materials.Breakout groups addressed technical, organizational, workforce, and governance questions from complementary perspectives.
2 Strategic Themes
The workshop identified strategic priorities for scientific computing ecosystems centered on discoverable and sustainable software, trustworthy and traceable workflows, human-AI teaming, and workforce development. These ecosystems should connect researchers with appropriate methods and resources while preserving validation, human judgment, accountability, and pathways for cross-disciplinary learning.
- Software ecosystems for AI-enabled scientific discovery: Shared research assets should capture reusable knowledge, context, and operational experience so distributed teams can discover, extend, validate, and trust complex workflows.Tool discoverability, sustainment, validation, and interoperability remain essential ecosystem requirements.
- Software ecosystems for AI-enabled scientific discovery: Ecosystems should help researchers choose methods suited to each problem, balancing accuracy, interpretability, reproducibility, performance, and cost rather than defaulting to maximum computational power.Simpler or specialized approaches may be preferable for many scientific tasks, and researcher-facing systems should identify relevant data, tools, constraints, missing pieces, and next steps.
- Trust, validation, and traceability: Trust requires VVUQ, provenance, traceability, transparency, and auditability as core design requirements that capture software, data, environments, decisions, assumptions, processes, and human contributions.These properties support interpretation, challenge, reuse, system-level evaluation, and safety of increasingly agentic workflows.
- Trust, validation, and traceability: AI-enabled scientific workflows require expanded VVUQ and software-trust methods that address hybrid components, human judgment, inappropriate numerical methods, inadequate convergence, and violations of physical or domain constraints.Conventional test suites remain necessary but may miss scientifically misleading results, especially when AI rapidly generates or modifies software.
- Human-AI teaming and paradigm shifts: Human-AI ecosystems should favor augmentation where understanding, plausibility judgments, intervention, provenance, shared task states, escalation, responsibility, and scientific validity must be preserved.Human-AI teams should be evaluated beyond speed or output, including error detection, reproducibility, recovery from failure, collaboration quality, and researchers’ ability to explain and extend work.
- Workforce, pedagogy, and governance: Workforce strategies should develop judgment, cross-disciplinary collaboration, and human-AI teaming through multiple learning pathways, authentic team environments, and career structures rewarding technical depth with collaborative leadership.Programs should support newcomers, adapting practitioners, and leaders responsible for governing and stewarding complex ecosystems across short-, medium-, and long-term horizons.
3 Paths Forward: Community Actions
Progress toward next-generation scientific computing ecosystems requires coordinated action because no single organization can provide the necessary infrastructure, norms, incentives, and workforce pathways. Eight complementary priorities emphasize shared assets, trust, human-AI teaming, workforce preparation, coordination, stewardship, and evaluation of scientific value.
- 3 Paths Forward: Community Actions: Communities should treat curated libraries, datasets, metadata standards, pipelines, benchmarks, validation examples, and documentation as first-class infrastructure supporting reuse and sustainability.Maintaining these assets and valuing their stewards are as important as creating them.
- 3 Paths Forward: Community Actions: Trust infrastructure should incorporate VVUQ, provenance, and auditability from the outset, tracing assumptions, decisions, interventions, and execution context for review and reproduction.Agentic systems also require attention to identity management and authorization.
- 3 Paths Forward: Community Actions: User-experience designs should help researchers identify methods, understand constraints, interpret outputs, and recognize when further validation is needed while preserving scrutiny for consequential uncertainty or cost.Interfaces, access mechanisms, support structures, and policies can lower unnecessary barriers and support informed choices.
- 3 Paths Forward: Community Actions: Organizations should define human-AI teaming norms for delegation, expert escalation, approval of AI-generated results, shared accountability, and boundary-spanning coordination across disciplines.Clear expectations for automation, human judgment, oversight, and documentation make these norms actionable.
- 3 Paths Forward: Community Actions: Workforce preparation should develop collaboration, verification, reasoning under uncertainty, and responsible tool-use skills, complemented by pilots and longitudinal studies of changing roles and career pathways.Technical training alone is insufficient for AI-enabled systems involving multiple forms of expertise and distributed responsibility.
- 3 Paths Forward: Community Actions: Sustained cross-sector coordination, stewardship incentives, and meaningful evaluation should align standards, preserve shared resources, and assess scientific validity, transparency, uncertainty, intervention, collaboration, and breadth of inquiry.Near-term work should prioritize shared assets, trust infrastructure, teaming norms, and stewardship incentives, while pilots test governance and longitudinal studies examine workforce effects and scientific value.
4 Conclusion
The future of scientific computing depends on ecosystems that combine AI-enabled capabilities with valid, interpretable, and creative discovery. Advancing these ecosystems requires socio-technical co-design across people, software, data, infrastructure, institutions, incentives, and norms.
- Ecosystem Foundations: Scientific computing’s future depends on ecosystems, not isolated tools, and AI-driven speed matters only when it supports valid, interpretable, and creative discovery.The workshop reinforced that acceleration alone does not constitute progress.
- Capabilities and Supports: Meeting this standard requires reusable software, auditable workflows, robust VVUQ and traceability, boundary-spanning teams, critical-judgment workforce pathways, and stewardship incentives.These needs persist as AI systems, computing architectures, and development practices change.
- Socio-Technical Co-Design: Socio-technical co-design requires people, software, data, infrastructure, institutions, incentives, and norms to evolve together through feedback, adaptation, and stewardship.The report presents this coordinated evolution as the practical path forward.
- Scientific Value: Well-designed ecosystems can accelerate discovery and expand scientific inquiry while preserving rigor, transparency, accountability, and collaboration.These qualities underpin scientific progress and public benefit.
A Workshop Description
The 2026 workshop brought together over 45 cross-disciplinary experts in Chicago from April 14–16 to chart a path toward more powerful, sustainable, and collaborative scientific software ecosystems.
- A Workshop Description: Over 45 cross-disciplinary experts met in Chicago, IL, during April 14–16, 2026, to chart a path toward more powerful, sustainable, and collaborative scientific software ecosystems.The description is reproduced from pre-workshop materials and retains the original future-tense wording.
Workshop Charge
The 2026 workshop addresses how AI-enabled methods, heterogeneous architectures, and data-intensive workflows are reshaping scientific computing by integrating technical innovation with community evolution. As Year 2 of a three-year series, it focuses on sharing, developing, and evaluating strategies for sustainable, interoperable, and community-informed scientific software ecosystems.
- Software and next-generation science: The workshop examines how AI-augmented workflows, heterogeneous platforms, and new scientific frontiers require new approaches to software and workforce development.It emphasizes community-driven practices so technical solutions reflect broad expertise, use cases, and stakeholder needs.
- AI-driven software ecosystems for scientific computing: It maps pathways toward robust, interoperable, and scalable software ecosystems that enable AI-driven discovery in HPC and data-intensive environments.The workshop also advances tools, frameworks, and infrastructure that support openness, reproducibility, and broad participation.
- Team-based software and cross-disciplinary research: The workshop advances team-based software through community co-design and evolving roles, career paths, and practices for cross-disciplinary research teams.Software is shaped through continuous dialogue among scientists, developers, users, and maintainers.
- AI for scientific software productivity and sustainability: It explores AI-assisted development, testing, performance tuning, documentation, and maintenance as means to enhance productivity and long-term software sustainability.Feedback loops are intended to align AI-enabled tools with real-world scientific computing workflows and community needs.
- Community and workforce development: The workshop seeks to cultivate next-generation R&D teams for AI-rich scientific computing environments while fostering collaborative, community-centered cultures.Its co-design methodology weaves together team-based software, AI-enabled workflows and infrastructure, and community, workforce, and ecosystem development.
Workshop Objectives
The workshop seeks an actionable vision for team-based scientific software that combines emerging AI capabilities with cross-disciplinary collaboration. Its objectives span advancing practices, addressing barriers, strengthening communities, and integrating technical innovation with sustainable ecosystem design.
- Workshop Objectives: The workshop aims to shape an actionable vision for team-based scientific software grounded in emerging AI capabilities and cross-disciplinary collaboration.
- Workshop Objectives: It advances shared understanding of team-based software methodologies, tools, and norms while integrating technical rigor with community sustainability.
- Workshop Objectives: It seeks strategies that overcome barriers, enable AI-aligned innovation, articulate future ecosystems, and strengthen and broaden the workforce.
- Workshop Objectives: The workshop integrates technical innovation with community-centered design so co-designed, AI-enabled, sustainably maintained software supports enduring collaboration and scientific progress.
Workshop Outcomes
The workshop synthesized community and technical insights into a post-workshop report that extends prior findings and informs scientific software, workforce, and ecosystem efforts. Its broader three-year effort seeks to strengthen the scientific software community, foster cross-disciplinary collaboration, and accelerate scientific discovery in an AI-driven world.
- Workshop Outcomes: Participants examined team-based scientific software practices for an AI-enabled future to assess progress, identify gaps, and prioritize continued attention across technical, organizational, and community dimensions.Discussion and co-design activities emphasized high-impact focus areas.
- Workshop Outcomes: The post-workshop report extends the 2025 workshop’s findings and informs ongoing and future efforts in scientific software development, workforce advancement, and ecosystem coordination.It may also shape evolving perspectives, policies, and practices across the scientific computing community.
- Workshop Outcomes: By interweaving community considerations with technical discussions, the three-year workshop effort aims to strengthen the scientific software community, foster cross-disciplinary collaboration, and accelerate next-generation scientific discovery.The effort responds to an increasingly AI-driven world.
B Workshop Participants … DOE Contacts
The workshop brought together organizers, academic and national-laboratory researchers, industry and nonprofit leaders, students, and DOE contacts. Participants represented expertise spanning computational science, mathematics, software, user experience, scientific engagement, and advanced scientific computing policy.
- Organizing Committee: The organizing committee included senior computational scientists, university faculty, nonprofit and industry leaders, and former DOE program communications leadership.Named organizers included Lois Curfman McInnes, Dorian Arnold, Prasanna Balaprakash, Mike Bernhardt, and Franck Cappello.
- Attendees: Attendees included faculty, doctoral students, national-laboratory managers, and computational research professionals from universities and laboratories.Examples include Gabrielle Allen, Alexandra Ballow, Tony Baylis, and David E. Bernholdt.
- Attendees: Industry participation included computational-methods engineering, research and development, enterprise architecture, and scientific-software perspectives.Represented organizations included GE Aerospace, Synopsys, and Google.
- Attendees: Participants represented user experience, science engagement, and high-performance-computing consulting alongside research and management roles.Johanna Cohoon was identified in user experience, while Charles Lively was identified in science engagement and HPC consulting.
- Attendees: The attendee roster also included national and international computational-science leadership, including Brookhaven, LBNL, RIKEN, and Indiana University.Named participants included Meifeng Lin, Satoshi Matsuoka, Staša Milojević, and Charles Lively.
- DOE Contacts: DOE contacts were Hal Finkel and David Rabson from the Office of Advanced Scientific Computing Research.Finkel was identified as Associate Director, and Rabson as Physical Scientist.
C Workshop Agenda · ARGONNE NATIONAL LABORATORY
The workshop agenda organized three days around changing scientific computing ecosystems, AI in research and industry, and potential paths forward, while incorporating technical, social, workforce, and co-design discussions. It also included a presentation on AI agents and automated science, with Argonne National Laboratory identified as the host research facility.
- C Workshop Agenda: Cross-disciplinary discussions covered software, AI-driven ecosystems, team-based research, scientific software productivity and sustainability, and community and workforce development.The agenda explicitly paired technical topics with cross-disciplinary research and community concerns.
- C Workshop Agenda: The agenda treated synergies, dependencies, and co-design opportunities as inputs to consolidated workshop outcomes.Moderated forums sought overall lists of changes and forward-looking development ideas across topics.
- C Workshop Agenda: A featured presentation framed AI agents as participants in cross-disciplinary science that can write code, design experiments, and synthesize knowledge across domain boundaries.The presentation linked these capabilities to changing scientific team structures, falling inference costs, and maturing automated code development.
- C Workshop Agenda: The agenda connected future ecosystem building with workforce and community challenges, success criteria, strategy development, and evaluation of progress.A plenary panel addressed workforce and community complexities, while breakout discussions asked what success looks like and how to evaluate progress.
- ARGONNE NATIONAL LABORATORY: Argonne National Laboratory was presented as a U.S. Department of Energy research facility operated by UChicago Argonne, LLC in Lemont, Illinois.The agenda materials describe it as the Midwest’s largest federally funded R&D facility, conducting basic and applied research.