Machine Ethics and Reasoning in AI

About the Workshop

The Machine Ethics and Reasoning in AI (MERe AI) Workshop is a two-day academic workshop bringing together researchers in philosophy, machine ethics, AI safety, and evaluations to strengthen the scientific basis for normative competence in frontier AI systems: the ability to recognize morally relevant features of a situation, apply principles in unfamiliar contexts, and reason through conflicting commitments and uncertainty.

As AI systems encounter situations not spelled out during training, how can we tell whether their moral reasoning is sound? The workshop connects philosophical foundations with empirical evaluation to identify what we should measure, how to measure it, and where current approaches fall short. Shared evaluation targets can support safer systems, more effective human oversight, and responsible use across domains where AI decisions affect people.

Day 1 will feature invited talks, contributed papers, and posters. Day 2 will focus on a shared task: developing a taxonomy of moral reasoning competencies and drafting an open problems paper to guide future research.

Our aim is to build agreement on what competent moral reasoning requires while preserving reasonable disagreement about moral conclusions. This is the second edition of the workshop, following the first MERe AI workshop held online in July 2025.

Call for Papers

We welcome contributions on machine ethics and moral reasoning in AI, spanning philosophical, theoretical, and empirical work. Papers on work in progress are welcome, as are submissions of authors' recently published work on topics highly relevant to the workshop. The workshop is non-archival.

We accept submissions in three length categories: up to 2 pages, up to 4 pages, or up to 8 pages. Bibliographies do not count toward these page limits. Submissions must be anonymized for review.

Submissions must be uploaded as PDFs in the single-column ACM manuscript format. Templates are available for Microsoft Word, LaTeX, and Overleaf. LaTeX users should use \documentclass[manuscript,screen,review,anonymous,nonacm]{acmart}.

Submit your work through the OpenReview submission site by Friday, October 30, 2026, 23:59 Anywhere on Earth (UTC−12). Acceptance notifications will be sent on Wednesday, November 11, 2026. Presenters must confirm their participation by November 20, 2026.

Topics of Interest

We welcome submissions on the following topics, grouped by theme:

Important Dates

Submission deadline

Friday, October 30, 2026, 23:59 Anywhere on Earth (UTC−12)

Acceptance notifications

Wednesday, November 11, 2026

Presenter confirmation

Friday, November 20, 2026

Workshop

December 1–2, 2026

Program

The detailed schedule and invited speakers will be announced soon.

Day 1: Presentations and Posters · December 1

Invited talks, contributed research presentations, and a poster session. Day 1 will be streamed for the public and for participants who cannot attend in person.

Day 2: Shared Task and Collaborative Writing · December 2

A working day of facilitated discussions to develop a taxonomy of moral reasoning competencies, identify evaluation gaps, and begin a co-authored open problems paper. Participants will shape a research agenda and form collaborations around the gaps they identify.

Shared Task: Moral Reasoning Competencies

There is growing interest in measuring and improving moral reasoning in AI, but limited agreement on the broadly applicable competencies that constitute successful reasoning. Participants will work together to identify these competencies and organize them into a shared taxonomy. The scope includes AI agents acting in the world, multimodal understanding and interaction, everyday morally relevant situations, and deliberation with users or other models, extending beyond conversation, single-turn dilemma answers, and chain-of-thought traces.

Possible starting points include:

• Understanding moral arguments and the reasons supporting their conclusions
• Recognizing morally relevant features in everyday situations
• Applying principles in unfamiliar contexts and resolving competing commitments
• Reasoning across multi-turn interactions with people and other models
• Handling uncertainty, missing information, and reasonable disagreement
• Maintaining sound reasoning across changes in context

The intended outputs are a draft taxonomy and an open problems paper that help researchers and funders prioritize concrete research directions. Participants will explore candidate evaluation tasks to clarify what a model should do, how its reasoning could be assessed, and where expert judgment or reasonable disagreement remains. The workshop aims to seed ongoing collaborations; implemented benchmarks are possible follow-on work.

Attendance and Registration

We welcome interest from researchers in philosophy, AI, cognitive science, and related disciplines. The workshop will take place at the Vector Institute in Toronto. Day 1 will be streamed; Day 2 will focus on in-person collaboration.

Virtual attendance is open to everyone. In-person attendance is reserved for participants with accepted research and/or an invitation from the organizers.

Register for virtual attendance

Travel Support

We have a very limited travel-support budget for in-person participants. Eligible in-person attendees can request consideration through the registration form shared by the organizers. Support is subject to availability and cannot be guaranteed.

Organizers

• 
  Aidan Kierans

  University of Connecticut; MIT Algorithmic Alignment Group

• 
  Lorena Almaraz De La Garza

  Vector Institute

• 
  Elizaveta Tennant

  Google DeepMind

Contact

For workshop inquiries, email contact@mere-ai-workshop.com.