The RSS Discussion Meetings Committee, along with the RSS Emerging Applications Section, invites submissions of discussion papers on the roles of statistics and machine learning in large language models and artificial intelligence.
Papers selected for publication will be presented at a multi-paper discussion meeting held during the RSS International Conference in September 2027 in Newport, and subsequently published in one of the Journals of the Royal Statistical Society series together with formal discussion contributions from the meeting and written contributions submitted afterwards.
Proposal
Artificial intelligence (AI) has advanced rapidly through developments in machine learning, particularly large language models and related generative methods. Yet many of the central challenges, including uncertainty quantification, generalisation, robustness, interpretability and causal reasoning, remain fundamentally statistical in nature.
This RSS discussion meeting will explore how statistical thinking both complements and critically informs modern AI and machine learning practice, and how these fields interact and converge in contemporary research and applications. Topics may include Bayesian and causal machine learning, trustworthy and explainable AI, statistical perspectives on large language models, and principled approaches to AI for scientific discovery and other high-stakes decision-making.
The aim is to clarify and advance the distinctive contribution of statistical methodology and reasoning to the future development of AI. Papers may address methodological, theoretical, computational or applied aspects of the topic.
Topics of interest include (but are not limited to):
Statistical foundations and theory
- Bayesian machine learning and simulation-based inference
- Causal machine learning and causal discovery
- Statistical theory for neural networks, including central limit and large deviation results
- Stochastic processes for AI and machine learning
Trustworthy and interpretable AI
- Uncertainty quantification
- Trustworthy, interpretable, and explainable AI
- Generalisation, robustness, and reproducibility
- Statistical perspectives on evaluation and benchmarking
Large language and multimodal models
- Large language models (LLMs)
- Large multimodal models
- Statistical perspectives on generative AI
Scientific and applied AI
- Statistically principled AI for scientific discovery
- Innovative applications in health, environmental science, biology, chemistry, education, economics, robotics, sociology, physics, and related fields
- Open data practices and reproducible AI research
Submitted papers will be reviewed both for scientific quality and for their potential to stimulate broad discussion across statistics and AI. Papers that satisfy the former criterion but are considered less suitable for a discussion meeting may, with the agreement of the authors, be referred for consideration as regular journal papers.
Submission Details
Submitted papers should be substantially shorter than is typical for a single‐paper discussion meeting (16 pages max excluding supplementary material, following standard JRSS formatting instructions).
To streamline the review process, submissions will follow a two-stage procedure consisting of:
1. a short abstract submission for initial consideration;
2. invited full-paper submission followed by standard peer review.
Full details are as follows:
1. Abstract submission. Authors are invited to send a one‐page abstract (400 words max) outlining the proposed discussion paper to journal@rss.org.uk by 8th July 2026.
2. Invitation for full papers. Authors of selected abstracts will be notified by 29th July 2026 and invited to submit a full paper for review.
3. Full paper submission. Full papers (maximum 16 pages excluding supplementary material, prepared using standard JRSS formatting guidelines) should be submitted via Manuscript Central to the most appropriate JRSS series (A, B or C) selecting the “Discussion Paper” option.
The deadline for full paper submission is 20th November 2026.
4. Refereeing. All submitted papers will undergo peer review according to the Society’s standard criteria for discussion meeting papers, considering both scientific quality and potential for discussion.
5. Final accepted papers are expected to be ready for pre-printing by mid-2027 ahead of the RSS International Conference in September 2027.
6. Authors of accepted papers will be expected to present their work at the discussion meeting during the RSS International Conference in September 2027. At least one author should therefore be available to attend and participate in the meeting.
Informal enquiries about the call are welcome and may be directed to the Discussion Papers Editor, Ben Swallow, bts3@st-andrews.ac.uk.