Cognition spaces: natural, artificial, and hybrid

Cognition spaces: natural, artificial, and hybrid

Cognitive processes are realized across an extraordinary range of natural, artificial, and hybrid systems, yet there is no unified framework for comparing their forms, limits, and unrealized possibilities. Here, we propose a cognition space approach that replaces narrow, substrate-dependent definitions with a comparative representation based on organizational and informational dimensions. Within this framework, cognition is treated as a graded capacity to sense, process, and act upon information, allowing systems as diverse as cells, brains, artificial agents, and human-AI collectives to be analyzed within a common conceptual landscape. We introduce and examine three cognition spaces- basal aneural, neural, and human-AI hybrid-and show that their occupation is highly uneven, with clusters of realized systems separated by large unoccupied regions. We argue that these voids are not accidental but reflect evolutionary contingencies, physical constraints, and design limitations. By focusing on the structure of cognition spaces rather than on categorical definitions, this approach clarifies the diversity of existing cognitive systems and highlights hybrid cognition as a promising frontier for exploring novel forms of complexity beyond those produced by biological evolution.

I. INTRODUCTION

What kinds of cognitions exist? What kinds of minds could exist but have never evolved? What kinds of constraints might limit the possible when dealing with minds? What forms of cognitive complexity might be engineered beyond those realized by living systems? Cognition, when broadly defined as the capacity of a system to sense, process, and respond to information, is not restricted to humans or even to neural substrates. Evidence across biology, physics, and the synthetic sciences indicates that cognitive processes can emerge within a wide range of material and organizational substrates, from single cells and multicellular organisms, to swarms, artificial neural networks and distributed sociotechnical systems. Despite this diversity, however, contemporary accounts of cognition remain shaped by disciplinary boundaries and, in some cases, anthropocentric assumptions about what qualifies as a "thinking" or "learning" system. A recurring challenge in the study of cognition and related notions such as intelligence or consciousness is the lack of consensus on a definition that is sufficiently general to encompass the full spectrum of possible cognitive processes and systems.

Debates about the nature of cognition have long revolved around whether it should be understood as a graded, continuous property of natural systems or as a discrete kind that appears only once certain organizational or functional thresholds are crossed. Continuum-based views emphasize that many of the mechanisms underlying known examples of cognition, such as bioelectrical networks and algorithms, such as information processing, feedback control, learning, adaptive regulation, and active inference are already present, in rudimentary forms, across a wide range of biological and even non-biological systems at many scales. In this sense, it has been speculated that differences between cellular chemotaxis, animal perception, and reasoning arise primarily from changes in degree and organization, rather than from categorical differences in kind.

By contrast, a discontinuous view of cognition holds that cognitive systems belong to a finite number of discrete classes. In this sense, cognition requires specific architectures or representational capacities. These are often tied to qualitative features such as movement, nervous systems, symbolic manipulation, or intentional states. Systems that lack these features may still display complex or adaptive behavior. However, they are not genuinely cognitive. Proponents of this view warn crete classes. In this sense, cognition requires specific architectures or representational capacities. These are often tied to qualitative features such as movement, nervous systems, symbolic manipulation, or intentional states. Systems that lack these features may still dis- play complex or adaptive behavior. However, they are not genuinely cognitive. Proponents of this view warn that overly permissive definitions risk trivializing cognition. Such definitions may extend cognition to any self-organizing or responsive system. The emerging field of diverse intelligence treats this as an experimental question. It is not primarily a linguistic or philosophical one. The goal is not to preserve the inherited conceptual categories. Instead, tools from cybernetics and the cognitive and behavioral sciences are tested across novel substrates. The aim is to empirically determine when these tools offer practical advantages. It is also to determine when they do not apply in a useful way. Nevertheless, both continuum and discontinuous perspectives rely on non-trivial theoretical assumptions. These assumptions concern which organizational features are cognitively relevant. The tension between these views reflects a deeper methodological choice. Cognition may be defined by minimal functional criteria. Such criteria allow smooth transitions across systems. Alternatively, cognition may be defined by stronger conditions. These seek to classify nature into distinct kinds.

Is there a way to accommodate the diversity of cognitive systems while still providing principled answers to the questions posed above? Within biology, closely related challenges arise across systems and scales. Consider, for example, the long-standing problem of defining life. Despite sustained efforts across disciplines, no single definition has achieved broad consensus, suggesting that an alternative conceptual strategy may be required. An influential approach is to take diversity itself as the starting point: rather than seeking necessary and sufficient conditions, one examines a wide range of case studies and situates them within a space defined by a small set of relevant dimensions, such as information processing, structural organization, and metabolic autonomy. When living (and in some cases non-living) systems are mapped into such a space, a striking pattern often emerges, characterized by highly uneven occupation and extensive voids. Both features are highly informative. Regions of dense occupation reveal recurrent combinations of traits and provide clues about how different forms of complexity have co-evolved, while unoccupied regions may signal effectively inaccessible domains, i.e. configurations that are evolutionarily inaccessible, physically or chemically constrained, or dynamically unstable. The resulting picture is both rich and interpretable: a morphospace that does not merely catalog diversity, but helps explain it.

In this paper, we adopt this conceptual framework to introduce and analyse three spaces of cognition spanning natural, artificial, and hybrid systems. Figure one surveys a set of case studies encompassing natural, artificial, and hybrid systems that combine living and non-living components. Our aims are threefold. First, we address the problem of cognitive complexity by situating diverse case studies within a set of relevant spaces, thereby shifting attention away from potentially ambiguous definitions of cognition. On this view, the structure of the space itself and the patterns of its occupation constitute an appropriate response to key questions about the nature and diversity of cognition. Second, we interpret the voids separating clusters within these spaces in terms of design principles, engineering challenges faced by artificial cognitive systems, and evolutionary or physical constraints.

Finally, across all three spaces, we undertake a comparative analysis of natural and artificial systems while explicitly examining hybrid forms of cognition that integrate both. Specifically, we define and analyse three cognition spaces, namely:

· Basal, aneural cognition space: single cells, simple multicellular organisms, xenobots, and organoids can exhibit minimal yet rich forms of memory, learning, and perception-action loops, offering minimal models of living information processing.

· Neural cognition space: multicellular systems equipped with neurons and brains, along with artificial agents. Synthetic biology and microbiome engineering reveal how designed genetic circuits can reshape behavioral attractors, introducing new modes of learning, coordination, and decision-making.

Human-AI hybrid cognition space: interactions between AI systems (e.g., large language models) and human users exemplify reciprocal learning processes. Several levels of pairwise complexities arise, including emergent classes of hybrid cognitive ecosystems-what we may call the humanbot, reflecting tightly-coupled but still distinctive agencies.

As we shall see, defining hybrid systems is far from trivial. Strictly speaking, hybridity no longer resides solely in physical attachment or material fusion. Systems such as Xenobots are not merely assemblies of living cells, but extended networks that include human experimenters, AI design tools, and robotic platforms that actively explore biological possibility spaces. These components interact through functional and informational interfaces, forming an integrated causal system. As a result, the boundary of the "system" becomes ambiguous: it is unclear where the organism ends, and its cognitive technological scaffold begins. Hybrid systems are therefore increasingly characterized by distributed function and collective agency rather than by clear material or ontological criteria. By constructing cognition spaces, we aim to sharpen the distinction between qualitative classes of cognition, compare their complexities, and identify potential analogies and qualitative discontinuities among them.