The evolution of consciousness and complexity have for a long time been seen as inherently linked. In thinking about the possibility of sentience in machines and nonhuman organisms, our collective willingness to attribute phenomenological experience to them seems to be primarily driven by a measure of their complexity. For organisms such as jellyfish that are too simple in terms of their behavioral repertoire and nervous system organization, it seems all but impossible to attribute them the rich kind of mindedness that we associate with human consciousness. Furthermore, it's precisely the biological complexity of cephalopods that has led to calls for a recognition of their sentience and hence for them to be included in animal welfare science and legislation (Browning, Reference Browning2019; New England Anti-Vivisection Society et al., 2020). It's in this context that Tononi's (Reference Tononi2004, Reference Tononi2005, Reference Tononi2008, Reference Tononi2012) integrated information theory (IIT) has become extremely popular in the public, despite its comparative unpopularity within the larger scientific community.

The identification of consciousness with integrated information itself, once a certain complexity of measured by “phi” Φ is reached, has long been viewed as problematic – indeed, untestable. Merker, Williford, and Rudrauf provide us with a barrage of well-worked out arguments against IIT, showing that it fails for a variety of formal, phenomenological, and neuroscientific reasons. Nevertheless, the popularity of IIT is not entirely ill-motivated and my goal here is to recover some of its merits while offering an additional criticism of the project. IIT has obvious virtues such as the in-principle applicability to systems very different from ourselves, allowing them to be placed along a continuum from more to less conscious (Tononi & Koch, Reference Tononi and Koch2015). Furthermore, the idea of beginning with minimal theoretical commitments and a very simple model is familiar from other sciences studying complex phenomena (Veit, Reference Veit2019a, Reference Veit, Magnani, Nepomuceno, Salguero, Barés and Fontane2019b). Aided by the perceived link between complexity and consciousness, one may be forgiven for thinking that IIT provides us with a good starting point. Yet, instead of providing us with a simple and general framework that can be tested and improved, Merker, Williford, and Rudrauf convincingly argue that the entire framework is resistant to change and empirical progress, thus holding us back, rather than enabling us to move forward.

An additional problem of the theory is its failure to take evolutionary considerations on board; and this may be its greatest problem yet. Now, there is admittedly something deeply right about IIT. Of course, complexity matters! No one can deny that and the explicit acceptance of a gradualist model of consciousness seems to lend itself very well to evolutionary considerations (Veit & Huebner, Reference Veit and Huebner2020). But it cannot just be a mere one-dimensional scale of information integration that matters for subjective experience. The biological world not only contains gradations, but also varieties, and this should be part and parcel of a biological account of consciousness. The IIT simply asserts that it's their chosen measure of complexity that matters – more so that it's all that matters – but doesn't offer any compelling reasons why this should be so (Browning & Veit, Reference Browning and Veit2021). As Dan Dennett once said at the 2017 NYU Animal Consciousness conference: “Complexity matters, but which complexity?” In order to determine which complexity matters for consciousness, we cannot avoid the teleonomic question of what consciousness is for. But this question has deliberately received very little attention within both the IIT framework and the science of (human) consciousness, despite the Darwinian insight that it's only once we address the function of a complex biological phenomenon that we can truly begin to understand it.

In an effort to begin with a minimal and theoretically neutral model, IIT deliberately avoids commitments to the evolutionary rationale of consciousness. But this silence on biological matters is ultimately the primary reason why the theory must be abandoned. From an evolutionary perspective, we must focus on the complexity of a new biological mode of being (Browning & Veit, Reference Browning and Veit2021). This is why an investigation of consciousness as a response to, rather than a mere product of complexity ought to be how we can begin a true biological science of consciousness; one that emphasizes varieties and gradations instead of an all-or-nothing quality, as was advocated early on by Griffin (Reference Griffin1976) and more recently by Godfrey-Smith (Reference Godfrey-Smith2019) in their emphasis on the different lifestyles of animals. Instead of asking for a neutral measure of complexity that constitutes consciousness, we deliberately ask for an evolutionary loaded sense of living complexity that makes consciousness worth having – this I shall refer to as “pathological complexity”: the complexity of an organism's life history (Veit, Reference Veit2021). Does that just replace one theoretically intractable notion of complexity with another? Not quite, because there already exists a science that has done the work for us. Pathological complexity can be operationalized as the computational complexity of the optimization problem studied by state-dependent or state-based behavioral and life history theory. It is a biological measure of complexity that scales up alongside organism's degrees of freedom and that exploded during the Cambrian (Veit, Reference Veit2022). Based on this notion of evolutionary complexity, we can build an alternative framework for the study of consciousness on what I call the pathological complexity thesis:

Pathological complexity thesis: The function of consciousness is to enable the agent to respond to pathological complexity.

Similar to the IIT, the pathological complexity thesis offers us a general theoretical framework and model for thinking about consciousness. As such, it will inevitably share some of the problems all theories at this level of generality are faced with. But it overcomes one central problem of the IIT: This is a sense of complexity that makes evolutionary sense. And it's this important feature that can serve as a scaffold for future bottom-up approaches that take Darwinian thinking seriously, emphasizing both gradations and varieties of subjective experience in animal life, and allowing us to make predictions about the phenomenological complexity of other animals – which can then feed back into our understanding of the pathological complexity they evolved to deal with.