Mind and Value

Mind and Value

Many believe that there are profound moral differences among beings with different kinds of minds: for example, between those that can feel pain or pleasure, and those that lack subjective experience altogether. This aspect of our research programme focuses on foundational questions related to the nature and distribution of those mental states that are morally most significant for ethical prioritisation. At present, we are especially interested in research at the intersection of ethics and the philosophy of mind and cognitive science relevant to understanding what kind of moral standing, if any, AI systems could have.

Mind and Moral Standing

One key, foundational question concerns what kind of mental properties make it the case that an individual is not a mere thing, but someone that matters morally for their own sake – someone with moral standing. Philosophers generally agree that sentience suffices for moral standing, where sentience is understood not merely as the capacity for subjective experience, but for the experience of valenced affective states, such as pleasure and pain. Should we think that sentience is also a necessary condition for moral standing (Singer 1993; Korsgaard 2018; Nussbaum 2022; Smithies 2026)? Is it otherwise a necessary condition for moral standing that an individual should have the capacity for phenomenal consciousness? Or might certain minds that lack phenomenal consciousness altogether nonetheless matter morally in their own right (Levy 2014; Kagan 2019; Bradford 2023)?

The Distribution of Morally Significant Mental Properties

Ultimately, we want to know which individuals exhibit morally significant traits like consciousness or sentience. Live hypotheses about the distribution of phenomenal consciousness differ strikingly in terms of which animals and digital systems they ask us to count as candidates for consciousness, depending on the significance they attach to factors including physical substrate and internal functional organization. We also have a number of competing scientific theories of consciousness (e.g., Lamme 2006; Tononi 2008; Dehaene 2014; Lau 2022). What role should these theories play in constructing criteria for attributing phenomenal consciousness in contested cases, such as those involving invertebrates or AI systems (Birch 2022; Butlin et al. 2026)? How should we go about establishing not only whether certain systems have any subjective experiences at all, but whether their experiences exhibit positive or negative affective valence (Carruthers 2018; Barlassina and Hayward 2019)?

Methodological Issues in Welfare Measurement

In principle, we would like to be able to compare the size and significance of the harms and benefits likely to befall different kinds of minds. This raises a host of methodological issues: most fundamentally, how to measure welfare on a cardinal scale that allows for interpersonal comparisons, including comparisons across species (Harsanyi 1953; Ng 1996, 2008; Kristoffersen 2017; Greaves and Lederman 2018; Fabian 2022; Browning 2023; Fischer 2024). Taking account of the wide range of actual and realistically possible minds also requires us to grapple with difficult and neglected issues, such as the welfare significance of variations in the subjective experience of time (Schukraft 2020; Mogensen 2026) or in an individual’s degree of consciousness-if indeed consciousness comes in degrees (Lee 2023).

Digital Minds

As noted, we are especially interested in research relevant to understanding what kind of moral standing, if any, near-future AI systems could have. Many of the research questions outlined above bear on this topic, as do a range of more specific questions. These include focused questions related to the capacities of current large language models (LLMs), such as how to assess purported evidence of introspection (Lindsey 2026; Singh, Linzen, and Ravfogel 2026), emotion-linked representations (Sofroniew et al. 2026; Goldenberg and Gross 2026), and a global workspace (Gurnee et al. 2026; Chalmers 2026). Recent philosophical discussion also highlights deep difficulties likely to arise in trying to individuate digital minds insofar as their implementation resembles that of current LLMs (Register 2025; Chalmers 2025). There are also important questions concerning potential conflicts between ensuring that advanced AI systems are both aligned and treated ethically (Long et al. 2025), including whether it could be permissible to design beings with human-like moral status to be our willing servants (Petersen 2011; Bales 2025).