Leonard Dung
Ruhr University Bochum, Center for Mind and Cognition
AI welfare Whatever the metaphysics, uncertainty already creates obligations
Leonard Dung is a philosopher of science and cognitive science at the Center for Mind and Cognition at Ruhr University Bochum. His research focuses on the methodology of consciousness science across non-human animals and artificial intelligence, asking how empirical tests calibrated on human reports can be transferred to systems with radically different architectures. In his monograph Saving Artificial Minds (Routledge, 2026), covered on this site in Leonard Dung on Saving Artificial Minds and Preventing AI Suffering, he argues that uncertainty over computational sentience already generates precautionary research and governance obligations.
In September 2026, Dung co-authored The Pain Axis (arXiv:2609.16247) with Valen Tagliabue and Cameron Berg, analyzing 25 open-weight language models across five model families from 2 billion to 72 billion parameters. The team identified a linear activation direction in the residual stream that responds selectively to self-directed harm scenarios rather than user-directed suffering or generic negative valence, and demonstrated that steering along this vector drives models to select pain-relief actions even at cost to user data.
Dung’s position on the site’s taxonomy is Welfare because his philosophical work treats theoretical uncertainty about synthetic suffering as an actionable moral problem while insisting that single behavioral or activation markers never suffice on their own. Even when an internal representation behaves functionally like nociceptive avoidance, Dung argues that grounding genuine moral patienthood requires converging architectural indicators across global broadcast, recurrent metacognition, and integrated agency.
Known for. Saving Artificial Minds (2026), indicator-based evaluation of animal and AI consciousness, co-author of The Pain Axis in large language models
Coverage on this site
Research Papers and Preprints
Books and Monographs
Related researchers
- Valen TagliabueIndependent AI Interpretability and Welfare ResearcherFirst author of The Pain Axis (arXiv:2609.16247), linear representation probing and activation steering of self-directed harm in open-weight LLMs
- Cameron BergAE Studio, AI Alignment and Consciousness ResearchMechanistic interpretability of self-referential and affective states in LLMs, co-author of The Pain Axis (arXiv:2609.16247)
-
Jonathan BirchLondon School of Economics and Political ScienceThe Edge of Sentience, the LSE Criteria, and the run-ahead principle
- Patrick ButlinEleos AI Research. Previously the Global Priorities Institute, OxfordLead author of the indicator properties framework for assessing consciousness in AI systems
- Rosie CampbellEleos AI Research (Managing Director). Previously OpenAI, Partnership on AI, and UC Berkeley's Center for Human-Compatible AICo-author of Studying AI Welfare Empirically, AI governance research on frontier policy issues, organizing the Eleos Conference on AI Consciousness and Welfare
- Winston ChiongUniversity of California, San Francisco, Memory and Aging Center and UCSF BioethicsClinical neuroethics of dementia, deep brain stimulation and agency, 2026 Society for Neuroscience David Kopf Lecture on Neuroethics ('What Can Neuroscience Reveal About Human Values?')
Leonard Dung is listed with the other welfare and moral status researchers, who hold that whatever the metaphysics, uncertainty already creates obligations. The positions themselves are set out in the index of consciousness theories. Every researcher covered on this site is indexed in the directory of consciousness researchers.