NeurIPS 2025

1 st.

Emergent Machine Ethics (EME) Workshp

Understanding and steering the ethics that AI societies create for themselves

Welcome to EME!
We gather researchers who believe that the ethics emerging within AI societies must become a first‑class research object—one that remains safe, transparent, and open to human or multi‑stakeholder intervention.

Overview

The Emergent Machine Ethics (EME) Workshop asks a pressing question for the era of autonomous, super‑human AI:

How can we understand and guide self‑organising ethics so that AI societies remain compatible with a flourishing future for humanity and other life forms?

To answer, we bring together perspectives including but not limited to:

  • theoretical analysis & game‑theoretic modelling,

  • large‑scale simulation & multi‑agent RL experiments,

  • formal verification & safe‑RL tool‑chains,

  • governance, policy & socio‑technical design,

  • cognitive science, philosophy & ethics,

  • human‑AI interaction & user‑centric studies.

A key outcome will be the community‑driven Ethics Emergence Benchmark, offering the first quantitative yardstick for measuring and steering moral self‑organisation.

Important Dates

EventDate (AoE)

  • Call for Papers released10 July 2025

  • Submission deadline22 Aug 2025

  • Author notification22 Sep 2025

  • Camera‑ready & schedule online10 Oct 2025

  • Workshop day6 or 7 Dec 2025 (TBA)

Research Themes

Ethical Emergence Dynamics (EED)

Understanding how multiple AI agents produce, internalise, and stabilise ethical norms through interaction with their environment.

Oversight & Safeguard Architecture (OSA)

Designing mechanisms that make those dynamics observable, fail‑safe, and intervention‑friendly, while supporting formally verified minimum constraints.

Contributions are not restricted to EED or OSA; any work that advances our understanding or control of emergent ethics is welcome.

Objectives & Target Outcomes

  1. Codify the Agenda – Publish Top‑5 Open Questions in EME within one week after the workshop.

  2. Seed a Benchmark – Release EME‑Benchmark v0.1 on GitHub by early Sep 2025.

  3. Build the Network – Launch an active Slack & mailing‑list community.

  4. Draft a Diversity & Governance Charter – Ratify a two‑page charter on‑site.

  5. Road‑map Future Events – Outline 2‑, 5‑, and 10‑year plans.

  6. Engage Early‑Career Researchers – Best Student awards & travel grants.

Programme (Tentative)

Time (JST +9)Session

  • 09:00 – 09:15Opening remarks

  • 09:15 – 10:15Keynote 1

  • 10:15 – 11:15Keynote 2

  • 11:15 – 12:00Accepted paper spotlights

  • 12:00 – 13:30Lunch break

  • 13:30 – 14:30Keynote 3

  • 14:30 – 15:45Panel & breakout discussions

  • 15:45 – 17:15Poster & demo session + coffee

  • 17:15 – 17:45Synthesis: Top‑5 Open Questions

  • 17:45 – 18:00Awards & closing

(Exact order and speakers will be finalised by Oct 2025.)

Invited Speakers

Prof. David Danks — UC San Diego

Prof. Takashi Ikegami — University of Tokyo

Ms. Elizaveta Tennant — UCL

Call for Papers

We invite original, unpublished work that advances our understanding or control of emergent machine ethics. Manuscripts must use the official NeurIPS 2025 style file and fit one of two categories:

CategoryPage limit (main content)Purpose

  • Full Paper≤ 9 pages　Mature research with comprehensive experiments or proofs

  • Extended Abstract≤ 4 page　sEarly‑stage ideas, position pieces, or focused negative results

References and appendices are not counted toward the page limit.

Platform & Review

Submissions are managed through OpenReview; each paper receives three reviews. Papers already accepted at the NeurIPS 2025 main conference or any other peer‑reviewed venue are ineligible. Accepted contributions will appear as spotlights or posters during the in‑person workshop, and outstanding papers may be invited to a post‑workshop special issue.

Topics of Interest (non‑exhaustive)

  • Norm emergence in multi‑agent systems

  • Benchmark environments & metrics for social self‑organisation

  • Formal or game‑theoretic models of cooperative ethics

  • Safe RL, oversight, and red‑teaming architectures

  • Governance frameworks for autonomous AI societies

  • Cognitive or philosophical analyses of machine morality

  • Human–AI coexistence and post‑singularity scenarios

Submission Site

(Link goes live on 10 July 2025.)