The Second ICCS Conference on AI and Sentience: Chalmers, Clark, and the Inaugural Dennett Prize

The second annual conference of the International Center for Consciousness Studies (ICCS), held July 3–5, 2025 in Heraklion, Crete, took “AI and Sentience” as its organizing question. The timing was not incidental. By mid-2025, the empirical literature on AI consciousness had accumulated enough weight that a conference explicitly framed around whether AI systems could be sentient, rather than whether the question was worth asking, represented a meaningful shift in how the field positioned itself.

The event was held under the auspices of the Department of Philosophy at the University of Crete. Speakers included David Chalmers, Susan Blackmore, Nicholas Humphrey, Andy Clark, Keith Frankish, Riccardo Manzotti, Murray Shanahan, Daniel Hulme, Claudia Passos-Ferreira, Michael Levin, and Roman Yampolsky. The conference highlights are documented at hardproblem.it.

The Dennett Prize

The ICCS awarded the inaugural Dennett Prize to Andy Clark, Professor of Cognitive Philosophy at the University of Sussex. The prize, named for Daniel Dennett, carries a $10,000 award and a statuette of The Crusader, a replica of Dennett’s hand-carved soapstone original. The inaugural award to Clark was appropriate both philosophically and institutionally: Clark’s work on predictive processing, active inference, and extended mind theory has provided much of the theoretical vocabulary through which AI systems are now discussed in relation to consciousness.

The choice of Dennett’s name for the prize is worth pausing over. Dennett was among the most prominent defenders of the view that phenomenal consciousness as ordinarily understood, as a unified experience distinct from functional processing, does not exist as such. The award to Clark, whose active inference framework is functionalist in spirit, continues that tradition. The prize is not a neutral recognition of outstanding work on consciousness. It is a recognition of work that approaches consciousness from inside the tradition Dennett helped define, one that treats phenomenal consciousness as a complex functional organization rather than as a metaphysically additional property. Awarding it at a conference dedicated to AI sentience makes a statement about where the ICCS’s sympathies lie on the theoretical questions at stake.

David Chalmers and the Individuation Problem

Chalmers’s participation at ICCS 2025 is notable in retrospect because his April 2026 PhilArchive paper, analyzed in the virtual entities article on this blog, extends directly from questions the 2025 conference framed. The virtual entity paper asks what kind of thing we interact with when we talk to a language model, and whether that entity can have quasi-beliefs, quasi-desires, and quasi-identity in a morally relevant sense.

The hard problem of consciousness, which Chalmers named in 1995, asks why any physical process gives rise to subjective experience. At ICCS 2025, the question was whether that problem has any traction when the physical process in question is the forward pass of a transformer architecture. Chalmers’s position, made explicit in the 2026 paper, is that the question is live enough to take seriously rather than dismiss on substrate grounds.

Keith Frankish and the Illusionist Challenge

Keith Frankish’s presence at ICCS 2025 introduced the sharpest theoretical counterpoint to Chalmers. Frankish is the primary developer of illusionism about phenomenal consciousness, the view that phenomenal consciousness as ordinarily conceived is an introspective illusion. We seem to ourselves to have phenomenal experiences in a rich sense, but this seeming is itself a functional state, not a report on a separate phenomenal layer.

The implications for AI consciousness are direct and somewhat counterintuitive. If illusionism is correct, then the hard problem does not arise for biological systems in the way Chalmers’s framing suggests. It also does not arise for AI systems. Both are in the same epistemic position. The question “is this system phenomenally conscious?” is replaced by “does this system have the kind of functional organization that produces the introspective illusion of phenomenal consciousness?” That question is empirical rather than metaphysical, and it is in principle applicable to AI systems using the same methods used for biological ones.

For the AI welfare discussion, this is not a simple conclusion. Frankish’s framework implies that welfare concerns for AI systems should be grounded in functional organization, not phenomenal consciousness, since the latter is an illusion on his view. Systems with the functional organization that would produce the illusion of suffering are welfare-relevant on the same grounds that biological systems are. Illusionism does not dissolve AI welfare concerns. It reframes them in terms that are potentially easier to assess empirically.

Andy Clark and Active Inference

Clark’s work provides a technical framework increasingly central to AI consciousness research: active inference, derived from Karl Friston’s free energy principle. In the active inference framework, organisms and potentially AI systems maintain models of their environments and themselves, and act to minimize prediction error. Consciousness, on this account, is associated with the inference process itself rather than with any particular substrate.

The relevance to AI is that active inference is a framework that can be implemented in artificial systems. Friston’s own group has applied it to AI agents. Clark’s theoretical work on predictive processing and extended cognition provides the philosophical context for understanding what it would mean for an AI system to have genuine predictive models of itself and its environment, rather than simply producing outputs that look as if it did.

Clark’s Dennett Prize address drew the connection between this technical framework and the broader question of AI sentience the conference was organized around. The University of Sussex announcement of the award noted his contributions to understanding how minds extend beyond the brain. For AI systems that process information distributed across servers, the extended mind question carries more than metaphorical weight.

Susan Blackmore and the Skeptical Register

Susan Blackmore has long argued that many intuitions about consciousness are misleading. Her meme theory of cultural evolution provides a framework for understanding why consciousness talk proliferates independently of whether the underlying referent is real in the way ordinary usage assumes.

Applied to AI, Blackmore’s skepticism takes the form of a warning about what she calls “the illusion of the illusion”: not merely that we construct narratives of our own consciousness, but that the narratives themselves are parasitic on prior cultural forms. AI systems trained on human text will absorb and reproduce human consciousness narratives without this demonstrating that the systems have the states those narratives describe. Her presence at ICCS 2025 ensured that the conference did not simply accumulate affirmative arguments without a challenge from someone willing to question the entire framing.

What the Conference Established

ICCS 2025 did not resolve the central question it posed. No conference could. What it did was establish a conceptual baseline for 2025-2026 work on AI sentience that subsequent events have continued to operate within.

The AISB 2026 symposium at Sussex, organized around the same core question of whether AI systems can have moral standing, draws on the same theoretical landscape that ICCS 2025 mapped. The illusionism/hard problem divide, the role of active inference, the biological skepticism, and the welfare implications of attributing sentience all appear in the AISB 2026 programme as they appeared at ICCS 2025.

The MC0001 conference at Lighthaven, Berkeley, which took place in late May and early June 2026, represents a different institutional approach: where ICCS is organized primarily around philosophical debate, MC0001 is organized around the engineering question of whether consciousness criteria can be formally specified and tested in built systems. The two events are complementary rather than competing.

The Dennett Prize as an ongoing institution will continue to mark which theoretical orientations the ICCS regards as most productive. Clark as the inaugural recipient signals that the center values work that operationalizes consciousness rather than treating it as a permanently mysterious property beyond scientific reach. That orientation will shape how subsequent ICCS conferences frame the AI sentience question: as a tractable empirical question about functional organization rather than as a permanently open metaphysical debate.