Conference Agenda

Theories of consciousness are increasingly incorporating the phenomenology of time, but they do so in different ways. Theories based on predictive processing, global workspace, and integrated information, have each begun to recognise different timescales of experience (fast and slow). By contrast, some higher-order theories propose experience as reflecting lagged averages of unconsciously integrated content over longer timescales. These apparent contrasts might be due to an impoverished view of the nature of conscious contents evolving over distinct timescales. We suggest that phenomenal content can evolve over multiple time scales simultaneously, and that these different timescales correspond to different degrees of perceptual (in)determinacy. We illustrate this view using empirical and real-world examples of sprinters, gist perception, pain reflex withdrawal, and postdiction. We further propose that the perceptual experience evoked by a stimulus becomes increasingly determinate over time. For example, (fast) conscious perception of a change can be distinct from classifying or recognizing the nature of the change. In this view, rapid perceptual experiences are possible but they may be indeterminate. More generally, depending on the target experience and experimental probes used, one could end up with conflicting notions of conscious experience, especially only if a single timescale of content is considered. This could lead to hasty claims about conscious experience being continuous or discrete, or mistakenly assigning functions relevant to an anchored duration. Finer consideration of time can lead to more productive interactions between theory and experiment in pursuing a mathematically-informed neurophenomenology.

Conscious experiences are highly structural, being related to one another in varying degrees of similarities and differences. Recently, interest towards these structural features has grown in all fields of consciousness studies, from philosophy to neurosciences. While the methodological use of structural tools such as mathematical spaces does not necessarily amount to specific metaphysical commitments, on a stronger reading, structuralism can be interpreted as favouring physicalist and reductionist positions or as aligning well with structural realism about science.

It is often argued that all that the natural sciences give access to is structure. Regardless, phenomenal experiences (or qualia) have long been considered to have non-structural, intrinsic properties. Intrinsicality lies at the core of traditional epistemic and modal arguments put forward to challenge or reject physicalist or functionalist positions.

In contrast, structuralism rejects the intrinsicality of qualia, viewing them as exhausted by their relational properties and individuating them by their position in a structure. This in turn could open the possibility of explaining qualia in purely scientific terms and of identifying them with certain neural structures.

I favourably review arguments advanced by proponents of structuralism about consciousness. In particular, I explore some empirical evidence in favour of the strict relationality of qualia as well as promising potential answers to epistemic and modal arguments on offer. I examine the behaviour and consequences of structuralist thinking when handling cases of multiple realizability. Finally, I consider structuralism's limitations when addressing the more general question of determining whether or not a particular system is conscious.

The question of which mental functions require consciousness has come to the forefront because of its relevance for investigating animal consciousness. I argue that most of the current research gets the functions associated with consciousness wrong. For two reasons: first, spontaneous intuitions about the role of consciousness are unreliable; second, the relevant empirical research is distorted by a confounding factor.

I start by flagging two fallacies that bias our spontaneous understanding of the role of consciousness in our mental lives. The ‘consciousness-only’ fallacy is to conclude that a conscious state is relevant for a mental function or behavior just because it correlates with that function or behavior. The ‘consciousness-first’ fallacy is to conclude that consciousness itself is required for a function or behavior just because conscious states are involved in the performance of that function or behavior. I illustrate these two fallacies with studies and claims from recent consciousness research.

I then argue that most empirical research interpreted as showing that some functions are associated with consciousness does not actually show this. Instead, it merely shows that the relevant functions falter when based on degraded sensory signals—which is trivial. In other words, signal strength is a significant confounding factor when investigating which functions require consciousness.

I finish by explaining how we can do better when investigating the functions that require consciousness. Until the relevant research is properly carried out, the presence of functions supposedly associated with consciousness in non-human animals should not convince the skeptics one bit.

Might two individuals, under identical external circumstances, have different experiences without noticing it? Locke’s “inverted spectrum” (IS) thereby asks how we can inter-personally individuate experience: When do you have a red experience like me? Kawakita et al. (KZJTO 2023; 2024) used Gromolov-Wasserstein alignment to approximate empirically when two individuals experience the same color. Here I explore the philosophy behind KZJTO’s approach.

The intuition behind IS motivates individuation by intrinsic qualities (qualia), which are independent of the structures they stand in (Lewis, 1929; Levine, 1995; Goff, 2017) and can switch without affecting cognitive and functional roles. Then, KZJTO fail to prove sameness. But what actually underlies KZJTO’s approach is individuation by structure (Fink et al., 2021; Fink & Kob, 2023; Prentner, 2024): Some isomorphism between the make-up of individuals A and B fixes the conditions under which A and B have the same color experience.

However, while KZJTO found an optimal transport, their data also shows slight divergences, which also need a structuralist interpretation. Some structuralists argue that IS are impossible due to the inherent asymmetries of quality spaces, which hold either necessarily (Hilbert & Kalderon, 2000) or contingently (Lyre, 2022). Then, divergences ought to be attributed to noise or measurement errors.

I explore an alternative option, based on the assumptions that identity does not allow for approximation and that structural individuation is holistic. Then, these divergences rather suggest that your red is never my red. I investigate several responses to and implications of this radical denial of inter-personal qualia individuation.

Integrated information theory (IIT) continues to garner intense interest – and controversy – thanks to its ambition to explain the fundamental physical basis of phenomenal consciousness. Unfortunately, the theory is frequently misunderstood, which holds back progress, and resolution of controversy. This presentation seeks to increase understanding of IIT and its successes, and the challenges it faces. We give a high-level overview of IIT, focusing on how the three key quantities of (i) system integrated information, (ii) cause-effect structure, and (iii) structure integrated information (Phi) purport to describe the quality and quantity of consciousness of any candidate system. We explain why, in current versions of IIT, these quantities are not well-defined for real physical systems, but only on toy model systems that lack physical interpretability. We discuss why high Phi is necessary but not sufficient for a rich inner subjective life (with reference to the so-called ‘expander grid’ thought experiment). Essentially, to have rich conscious contents a system must achieve high Phi repeatedly over time, and with distinct transitions between multiple cause-effect structures. We discuss the difference between proxy and approximation measures of Phi. We argue that proxy Phi measures, as opposed to approximations, have been successfully applied to empirical data and should encourage further development of IIT. Approximations however require the additional criterion of being well-defined. Finally, we discuss recent work explaining how experiences of space and time arise, illustrating how IIT can successfully link neural structures to conscious contents without needing to solve the problem of concocting a well-defined Phi.