AI is increasingly being adopted across scientific fields, yet significant gaps remain in the evidence base about its drivers, barriers, and impacts. This session will bring together researchers, practitioners, and policymakers to explore the state of knowledge, and opportunities to study AI adoption and its impact on science.
Long Abstract
AI is increasingly being adopted across scientific fields, yet significant gaps remain in the evidence base about its drivers, barriers, and impacts. This session will bring together leading researchers, practitioners, and policymakers for a presentation and panel discussion to explore the state of knowledge, evidence gaps, and opportunities to study AI adoption and its impact on science, and how AI itself can be leveraged to enhance our understanding of these dynamics. Participants include:
> • Juan Mateos-Garcia (Google DeepMind), who will present evidence of AI as a general purpose technology for science and discuss implications for measurement
> • Iulia Georgescu (Institute of Physics), will present lessons learned from research analysing adoption of AI in the UK physics community
> • Dashun Wang (Northwestern University) will talk about SciSciGPT, a tool to advance human-AI collaboration in the science of science
> • Carla Carbonell (La Caixa Foundation), will present her work evaluating the impact of AI tools science funding effectiveness and outcomes
> • Matt Clancy (Open Philanthropy), will be the discussant
> • Ben Steyn (UK Metascience Unit), who will chair a panel where we will discuss policy implications
The Institute of Physics carried out a study to understand how AI is used in physics, the physicists’ attitudes towards AI and what are the challenges and opportunities for AI to contribute to physics research and innovation. The study uncovered discipline-specific perspectives.
Long abstract
In some areas of research, physicists have been using AI for decades, but these methods have only become pervasive throughout physics in the past five years. Physics both contributes to the development of and benefits from AI technologies, which leads to a special relationship between the two fields.
Based on a survey on 700 physicists and community workshops, the study found that many of the views and attitudes of the physics community align with those of the broader scientific community identified in previous work. However, our study found that the respondents were concerned by the environmental impact of AI and the loss of understanding that could result from the use of AI. The community was keen to develop codes of best practice, guidance for data sharing and sustainable software development, as well as collaborate with industry.
The “la Caixa” Foundation has been exploring AI-assisted decision-making to alleviate the administrative burden within its funding programmes evaluation pipeline. These implementations highlight AI’s potential to enhance efficiency while upholding ethical standards and transparency.
Long abstract
The “la Caixa” Foundation has been exploring AI-assisted decision-making to alleviate the administrative burden within its funding programmes evaluation pipeline. Currently, we employ an algorithm to detect immature project proposals before they reach the peer-review stage (content eligibility), recommending their exclusion to a human overseer. Additional AI tools include a reviewer-matching system and an assistant for the revision of the evaluation report and their compilation in case of Selection Committees.
These implementations highlight AI’s potential to enhance efficiency while upholding ethical standards and transparency. Key considerations include the critical role of human oversight and the need to address bias, accountability, and fairness in research funding.
Looking ahead to 2025, planned developments include a Chatbot for Query Resolution, a Proposal Writing Assistant, a Proposal Review Assistant, an Explainability project for the content eligibility algorithm, and public access to this algorithm.
Dashun Wang (Northwestern University Kellogg School of Management)
Matt Clancy (Open Philanthropy)
Short Abstract
AI is increasingly being adopted across scientific fields, yet significant gaps remain in the evidence base about its drivers, barriers, and impacts. This session will bring together researchers, practitioners, and policymakers to explore the state of knowledge, and opportunities to study AI adoption and its impact on science.
Long Abstract
AI is increasingly being adopted across scientific fields, yet significant gaps remain in the evidence base about its drivers, barriers, and impacts. This session will bring together leading researchers, practitioners, and policymakers for a presentation and panel discussion to explore the state of knowledge, evidence gaps, and opportunities to study AI adoption and its impact on science, and how AI itself can be leveraged to enhance our understanding of these dynamics. Participants include:
> • Juan Mateos-Garcia (Google DeepMind), who will present evidence of AI as a general purpose technology for science and discuss implications for measurement
> • Iulia Georgescu (Institute of Physics), will present lessons learned from research analysing adoption of AI in the UK physics community
> • Dashun Wang (Northwestern University) will talk about SciSciGPT, a tool to advance human-AI collaboration in the science of science
> • Carla Carbonell (La Caixa Foundation), will present her work evaluating the impact of AI tools science funding effectiveness and outcomes
> • Matt Clancy (Open Philanthropy), will be the discussant
> • Ben Steyn (UK Metascience Unit), who will chair a panel where we will discuss policy implications
Accepted papers
Session 1 Tuesday 1 July, 2025, -