Why Is Anything Conscious? (2409.14545v5)

Abstract: We tackle the hard problem of consciousness taking the naturally selected, embodied organism as our starting point. We provide a formalism describing how biological systems self-organise to hierarchically interpret unlabelled sensory information according to valence. Such interpretations imply behavioural policies which are differentiated from each other only by the qualitative aspect of information processing. Natural selection favours systems that intervene in the world to achieve homeostatic and reproductive goals. Quality is a property arising in such systems to link cause to affect to motivate interventions. This produces interoceptive and exteroceptive classifiers and determines priorities. In formalising the seminal distinction between access and phenomenal consciousness, we claim that access consciousness at the human level requires the ability to hierarchically model i) the self, ii) the world/others and iii) the self as modelled by others, and that this requires phenomenal consciousness. Phenomenal without access consciousness is likely common, but the reverse is implausible. To put it provocatively: death grounds meaning, and Nature does not like zombies. We then describe the multilayered architecture of self-organisation from rocks to Einstein, illustrating how our argument applies. Our proposal lays the foundation of a formal science of consciousness, closer to human fact than zombie fiction.

• The paper introduces a pancomputational enactivist framework that integrates lower and higher-order theories to explain the emergence of conscious selves.
• It employs natural selection and mechanistic computation to classify stages of consciousness from basic adaptations to fully developed meta self-awareness.
• The work challenges traditional views by asserting that phenomenal consciousness is more pervasive than access consciousness, questioning the zombie concept.

Hierarchical Construction of Conscious Selves in Biological Systems

This paper aims to address the enduring "hard problem" of consciousness by proposing a mathematical framework that integrates lower and higher-order theories of consciousness. The authors, Bennett, Welsh, and Ciaunica, develop a pancomputational enactivist framework to elucidate how biological systems self-organize and construct conscious selves. Their framework is grounded in natural selection and mechanistic computation, departing from abstract computational dualism. This essay provides an expert-level overview of their argument and formalism.

Self-Organizing Biological Systems and Conscious Selves

The authors begin by establishing that any formalization of consciousness must apply to all conceivable environments characterized by a set of states. They introduce axioms which mandate that an environment must have distinguishable states, and they describe any aspect of the environment through a set of declarative programs. Notably, the authors adopt a representationless form of pancomputationalism, which contends that phenomenality and computation are intertwined in a mechanistic sense.

A central thesis in the document is the distinction between software and hardware is a simplification—software is nothing more than the state of hardware. Therefore, abstract "representations" are merely higher-order interpretations of lower-level qualitative experiences. This forms the basis for dissolving the hard problem of consciousness. Quality precedes quantity, and consciousness is a continuum beginning at the fundamental level of biological self-organization.

Relevance Realization and Causal Learning

The authors argue that "relevance realization" in self-organizing systems is tantamount to the construction of weak, adaptable policies that are optimized through natural selection. These policies serve as qualitative classifiers for the organism's interactions with its environment. The mathematical framework accommodates this by associating each policy with a set of tasks and preferences, allowing for a lattice structure of progressively weaker (and thus more general) policies.

In an interactive environment, the distinction between intervention and observation is paramount. A causal identity is constructed to differentiate an organism's interventions from passive observations. This identity allows the organism to predict the effects of its actions efficiently. The framework formalizes causal learning, enabling an organism to infer not just immediate effects but also complex social dynamics.

Hierarchical Levels of Conscious Selves

By scaling natural selection pressures and the capacity for weak policy optimization (WPO), the authors construct a hierarchy of conscious selves:

1. Stage 0: Unconscious Systems - Characterized by entities like rocks, which have no self-organizing behavior or consciousness.
2. Stage 1: Hard-Coded Adaptations - Embodied by organisms with hard-wired responses, such as single-celled protozoa, which adapt to their environment through preset behaviors.
3. Stage 2: Learning Systems - Exemplified by organisms like jellyfish and nematodes, which exhibit learning but lack a centralized "self" for causal reasoning.
4. Stage 3: 1ST Order Self - Marked by organisms that possess a first-order self, facilitating reafference and allowing for the prediction of the consequences of their actions. Houseflies exemplify this stage.
5. Stage 4: 2ND Order Selves - Including organisms that exhibit access consciousness and theory of mind. Ravens, which can anticipate and manipulate the behavior of others, reflect this stage.
6. Stage 5: 3RD Order Selves - Characterized by higher-order self-awareness, marking the stage of fully developed human consciousness, capable of meta self-reflection.

The primary radical claim here is that phenomenal consciousness is likely more common than access consciousness, invalidating the concept of a "zombie," i.e., an entity behaving indistinguishably from a conscious being but lacking any qualitative experience.

Implications and Future Directions

The proposed formalism offers a unifying framework spanning lower and higher-order consciousness theories. It situates conscious experience within a continuum that begins with basic biological self-organization motivated by valence. This perspective aligns with the view that subjective experience and qualitative aspects form the foundation upon which abstract representations and higher-order consciousness are built.

The authors suggest that future research should aim to identify first, second, and third-order selves in biological organisms and explore the similarities and differences between biological and artificial systems in this context. Understanding how artificial systems might construct conscious selves could provide valuable insights into the nature of consciousness itself.

The formalism proposed lays a robust groundwork for a formal science of consciousness that is deeply rooted in natural selection mechanisms, offering a promising avenue for both theoretical and empirical investigations into the nature of conscious experience.