Alien life as information structures

Strong Group Minds, Information Theory, and the Search for Extraterrestrial Intelligence

In my book (“From groups to gods: Evolutionary and neuroscience studies of the collective mind” to be published by Cambridge University Press next month in June 2026) I argue that there are entities I call “strong group minds”. By group mind I mean (roughly speaking) the personality and cognitive capacities that are associated with a collection of individuals assembled for some purpose in a group. The concept of a “group mind” is often used by biologists to describe the collective behavior of hyper-social or eusocial organisms, such as some species of ants and bees, or by social scientists to describe a sense of collective consciousness or a collective personality among human groups. It has become increasingly clear that “group mind” is not merely a metaphor for group identity and related effects on individual behavior. Rather, although we cannot “see” them with our eyes, some special types of group minds that I call “strong group minds as opposed to weak group minds are real entities that possess standard criteria for life and for mind such as agency, intentionality, and rationality. Weak group minds are simply the standard distributed forms of cognition that cognitive scientists have been working with for decades-they emerge from the bottom up and simply aggregate the attitudes of each member of the collective. Strong group minds on the other hand exhibit higher order cognitive states that structures the attitudes of individual members of the group (i.e. downward causation). We can measure dimensions, cognitive capacities, cognitive content, operations and effects of strong group minds in the world. They are real and I argue that they are likely to be significant for understanding many political, religious and social phenomena.

I lay out the evidence for strong group minds in that book. Here I want to purse one interesting application of the idea—namely whether such entities exist elsewhere than on Earth. In short, can we construe aliens, if they exist, in terms of the strong group minds hypothesis. I think it may be worth considering. First I need to show you that some group minds satisfy information theoretic criteria for life.

I. What Life Is: From Chaitin to Walker-Davies

The question of what makes a system alive, as opposed to merely complex or merely organized, has attracted serious mathematical attention since at least Gregory Chaitin’s 1979 paper “Toward a Mathematical Definition of Life.” Chaitin’s starting point was a dissatisfaction with purely physical or chemical definitions. A crystal is highly organized (an in some sense it reproduces itself) but trivially describable. A gas contains enormous information in its microstate but is entirely unorganized. Neither is alive. Chaitin proposed that what distinguishes living systems is a specific kind of structured mutual information, the degree to which parts of a system are interdependent in ways that cannot be accounted for by describing each part independently.

His formal tool was the d-diameter complexity Hd(X): the minimum number of bits needed to describe a system X as the sum of independent parts each of diameter no greater than d. For an organized system, Hd(X) grows rapidly as d decreases below the scale of significant patterns: you are forced to describe more and more independently when you cannot exploit the interdependencies that make the system what it is. Chaitin proved formally that neither gas nor crystal exhibits this behavior, while hierarchically organized structures do. The rate of increase of Hd(X) as d decreases is a “Fourier spectrum” of organization. The mutual information of parts, H(X:Y) = H(X) + H(Y) − H(X,Y), is the quantity that captures whether parts are genuinely interdependent or merely colocated.

Chaitin’s framework was explicitly substrate-independent. You could get life in Chaitin’s sense without the typical biology that instantiates it on Earth. There is a very low probability of creating organization by chance without a long evolutionary process. Random strings (his “gas”) and constant strings of reproducible items (his “crystal”) both fail exhibit organized interdependence that characterizes life. From this algorithmic point of view life is characterized not by complexity per se, not by entropy per se, but by organized interdependence, the mutual information between parts that reflects a higher-level pattern.

Walker and Davies (2013) advanced this program decisively in their paper “The Algorithmic Origins of Life,” shifting from static organization to dynamic causal architecture. Their contribution was to identify the mechanism by which Chaitin’s organized mutual information is actively maintained rather than merely present. In non-living systems, information is passive: a crystal encodes information about the forces that shaped it, but that information sits inert. It does not reach back and instruct the crystal’s future growth in light of higher-level goals. Causation flows exclusively upward, from physics to pattern. In living systems, something qualitatively different occurs. The genome does not merely record the organism’s current state. It actively constrains what the organism will do, reaching across time to shape future physical states in ways that serve the system’s own persistence. This is top-down information control: higher-level informational states exert genuine downward causal efficacy on lower-level physical dynamics. The whole governs the components. Life, on the Walker-Davies account, is what happens when a system crosses the phase transition from passive information, Chaitin’s organized mutual information merely present in the system, to active information: informational states that govern the system’s future dynamics. A living system achieves causal closure at the informational level: the system-level information state constrains component-level dynamics in ways that cannot be explained by component-level information alone. The system knows something, as a system, that no component knows individually, and it deploys that knowledge to organize its components in service of its own persistence.

The Walker-Davies criterion, like Chaitin’s, is substrate-independent and scale-free. It makes no reference to carbon, water, DNA, or neurons. It specifies a relationship between information levels, specifically between higher-order system states and lower-order component dynamics, that can in principle be instantiated in any substrate capable of storing and processing information. This is not a coincidence of formulation. Both frameworks are working within the same mathematical tradition: information theory as a language for describing the relationship between wholes and parts, between the irreducible and the decomposable, wherever that relationship occurs.

II. The Full Signature of a Strong Group Mind: Information Properties and Network Architecture

The Walker-Davies criterion remained largely formal until the last decade, when information theory developed the tools to operationalize it with precision. But causal synergy, the Walker-Davies core, is necessary and not sufficient. In my book From Groups to Gods (McNamara, 2026) I try to identify a richer set of information-theoretic and network-architectural properties that together constitute the full signature of a living strong group mind. These properties are mutually reinforcing, each following from the others when a collective system has genuinely crossed the Walker-Davies phase transition, and each produces a distinct measurable signature. Together they constitute the strong group mind or SGM information signature: the detection target for the index proposed in Section V.

Property 1: Synergistic causal information. The foundational measure, introduced by Tononi and colleagues and substantially extended by Varley (2024a, 2024b), Mediano and colleagues (2022), and Luppi and colleagues (2024). Integrated information (Phi) quantifies the degree to which a system generates information as a whole irreducible to its parts, precisely the quantity that Chaitin’s d-diameter complexity approximates from the algorithmic direction. The new math framework known as Partial information decomposition (PID) decomposes total information into redundant, unique, and synergistic components, where synergistic information is present only in the joint state of multiple sources and cannot be extracted from any proper subset. Causal emergence in the framework, is the Walker-Davies top-down causal closure expressed as a time-series measure: the macro-level variable predicts future micro-level states better than any individual micro-level variable does. Varley’s (2024b) synergy-first backbone decomposition makes this measurable at any system scale by sweeping the spectrum from the most fragile synergy, destroyed by the failure of any single component, to the most robust. Kemp, Kline, and Bettencourt (2024) show that this synergy-maximization is not merely a descriptive property of existing SGMs but a growth-rate principle derivable from the formal properties of information alone: groups that pool synergistic rather than redundant information outgrow all competitors over time. That is one major reason why natural selection facilitates the emergence of strong group minds regardless of substrate. Synergistic causal information is thus the selection pressure that drives major evolutionary transitions (MET) wherever information-processing collectives (from cells to animals) exist.

Property 2: Quasi-rational collective decision-making with temporal depth. An SGM is not merely a system that generates synergistic information in a static sense. It processes that information temporally, indicating that is moving toward some goal whether it is dissipation of energy, maintaining Markov blanket boundaries or maximizing synergistic information content. Its exploration of the problem spaces around achieving its goals pushes the system toward outputs that reflect something analogous to rational deliberation at the collective level. This produces a specific temporal signature in the system’s state variables: evidence accumulation toward decision thresholds at the macro level. You can see this in the drift-diffusion dynamics documented in colony-level decision-making by Navas-Zuloaga and colleagues (2022). A system doing quasi-rational collective deliberation exhibits a characteristic 1/f power spectrum in its state variable dynamics, associated with optimal information integration across timescales, distinct from white noise (pure randomness), red noise (purely physical processes like stellar variability), and the simple harmonic signatures of individual periodic processes. This temporal depth, the system’s capacity to accumulate evidence across time and integrate it toward collective decisions, is a property that Walter-Davies causal synergy alone does not capture. I suggest that the collective analog of logical depth is measurable in principle from the autocorrelation structure of multi-epoch observational time series—at least that is what cognitive scientists look for when they are working with this type of data.

Property 3: Transactive memory architecture with complementary signal heterogeneity. In my book I argue that SGMs are constituted by a specific architecture of knowledge distribution: different members hold complementary, specialized informational pools that no single member possesses in full, and the collective integrates these pools to generate outputs no subset could produce. Cognitive scientists have shown that when information is deliberately distributed across group members so that no individual has the full picture, the group can still solve the experimental problem presented to it by accessing the collective integration across individuals to arrive at the correct solution. Informationally this is the maximum synergy principle in structural form. It produces a specific PID signature in the unique atoms of the decomposition: each component contributes unique information that no other component contributes, organized around a collective whole that synergistically exceeds all of them. A system of individually-acting agents coordinating by communication produces predominantly redundant information at the collective level, since each agent’s contribution is partially recoverable from the others. A transactive memory architecture, on the other hand, produces high unique information per component combined with high synergy at the collective level, a signature that is formally distinguishable from both pure redundancy and undifferentiated synergy.

Property 4: Intentional directedness toward environmental goals. I also argued in my book that SGM is characterized by full-blown intentional states, the aboutness or directedness of mental states toward objects and goals in the world. A turbulent stellar atmosphere may exhibit synergistic spectral correlations but is not directed toward anything. On the other hand an SGM’s state variables can carry information about environmental states organized around the pursuit of collective goals. This means the joint mutual information between the collective’s internal state variables and external environmental state variables should exhibit a specific info structure: synergistic information that is about something…strongly linked to something like environmental states, carried jointly by the collective’s components in ways that cannot be recovered from any subset, and organized around the system’s operating directives. One way to look at this in terms of the alien life problem is when so called transfer entropy from collective state variables to future environmental states, exceed the reverse. That sort of situation would at least suggest intentional coupling: the collective is not merely tracking its environment but predicting and acting upon it.

Property 5: Collective predictive processing and active inference. My book grounds SGM information processing in Friston’s free-energy framework. A system doing collective active inference generates models that minimize surprise, actively predicting its environment and updating its predictions in light of new evidence. This is the group-level analog of individual predictive processing. Informationally it means the collective’s state variables exhibit forward-looking temporal correlations, where current collective states carry information about future environmental states that exceeds what current environmental states alone predict. In other words in addition to property 4 above property 5 implies that the Kullback-Leibler divergence between the collective’s predicted environmental states and the actual environmental states should decrease over time as the collective learns, a signature measurable from multi-epoch observations as systematic reduction in prediction error across the collective’s observable outputs.

Property 6: Stable collective identity, personality, and emotional attractor states. My book identifies collective personality and collective emotions as persistent dispositional properties that characterize the SGM’s orientation. Informationally these correspond to stable attractor states in the collective’s dynamical information landscape. The group returns to characteristic configurations under perturbation; its state variables exhibit characteristic persistence, autocorrelation, and return times. This is Chaitin’s hierarchical organization applied dynamically: the SGM’s temporal trajectory through state space is organized around stable attractors rather than diffusing randomly or converging to a fixed point. I am not mathematician but Claude AI tells me that this kind of attractor stability is measurable from the Lyapunov spectrum of the collective’s state variable time series: a system organized around stable collective identity exhibits bounded, recurrent dynamics with characteristic return times, distinct from both the divergent dynamics of chaotic systems and the monotonic convergence of equilibrium systems. I have seen measurements like this in behavioral time series where we can rule out both randomness and spurious associations when collective systems evolve over time so it sounds reasonable to me.

The network architecture of biotic SGMs. Beyond these six information-theoretic properties, biotic SGMs exhibit a characteristic network architecture that is itself a signature of their cognitive organization. In my book I draw upon recent work in social network analyses to identify the structural properties that distinguish SGM networks from both random networks and scale-free networks of individually-acting agents. As is well-known now, biotic agents (like SGMs) tend to exhibit small-world topology: high local clustering, meaning densely interconnected subgroups, combined with short average path lengths across the whole network, meaning any two nodes can reach each other through a small number of steps. This combination, first characterized by Watts and Strogatz (1998), maximizes information integration efficiency. It allows specialized subgroups to develop deep local expertise while maintaining rapid global information integration across the whole collective. A random network achieves short path lengths but poor local clustering. A regular lattice achieves high clustering but long path lengths. The small-world architecture achieves both, and it is the architecture that both individual brains and collective cognitive systems converge on under the pressure of efficient information integration.

In addition, I suggest that biotic SGMs exhibit functional role differentiation with gatekeeping nodes. SGMs develop an internal division of labor, with specialized nodes controlling information flow between the system interior and the external environment. Gatekeeping nodes, positioned at the boundaries of the collective, selectively admit and filter information transactions with the environment. Hub nodes, with high betweenness centrality, broker information flows among specialized subgroups. This produces a characteristic degree distribution in the network: not necessarily scale-free (which would be the signature of preferential attachment in a growing network of independent agents) but a distribution reflecting functional role specialization, with a small number of high-centrality hubs serving integrative and gatekeeping functions and a larger number of specialized peripheral nodes. Biotic SGMs exhibit high network efficiency combined with low divisibility. A network with high efficiency and low divisibility is one in which information can flow rapidly across the whole system and in which the system cannot easily be decomposed into independent subnetworks. This is the network analog of high Phi: the system is informationally unified in a way that resists decomposition. Defection, by contrast, produces a network that is highly modular and divisible, decomposable into the individual components it is composed of.

These network-architectural properties are not merely descriptive additions to the information-theoretic criteria. They are, in an important sense, the implementation of those criteria in biotic systems. High Phi requires both differentiation and integration, and the small-world topology with functional role differentiation is how biotic systems achieve this balance. Causal emergence requires macro-level states that govern micro-level dynamics, and the gatekeeping hub architecture is how biotic systems implement this downward causation. Synergistic information requires that the joint state exceed all subsets, and the transactive memory architecture with complementary signal heterogeneity is how biotic systems generate and maintain that excess. The information-theoretic properties and the network-architectural properties are two descriptions of the same underlying phenomenon: a collective system that has crossed the Walker-Davies phase transition and is in a very real sense alive at the collective scale.

III. Strong Group Minds: Scale and the Major Evolutionary Transition

Before turning to the cosmic implications of this framework, a clarification about scale is essential, one that significantly affects how we think about detection of alien life.

A strong group mind, as defined by the information-theoretic criteria above, is not necessarily civilizational in scale. Small groups can satisfy the criteria. A four-person work group exhibiting high Phi, as Engel and Malone (2018) demonstrated empirically, is a weak example of collective informational integration. Eusocial insect colonies of tens of thousands of individuals exhibit robust causal emergence at the colony level, as Navas-Zuloaga, Pavlic, and Smith (2022) document. Human religious communities, enterprise associations, scientific collaborations, and military units, any group that generates substantial synergistic information and exhibits top-down causal closure on its members satisfies the formal criteria to some degree. The signature is not binary but graded: groups vary in how much synergistic information they generate, how robustly their collective-level states predict individual-level dynamics, how far their collective Phi exceeds the sum of individual Phis. Scale matters for detection, not for definition. A small strong group mind, whether a hunting band, a colony, or a ritual community, may satisfy the formal criteria but produce physical signatures too faint to detect across interstellar distances. A civilizational-scale strong group mind, one in which the information-theoretic signatures of collective organization extend across an entire planetary system’s worth of coordinated activity, would produce signatures orders of magnitude larger. The detection argument in what follows is therefore calibrated to civilizational-scale SGMs, not because smaller SGMs are theoretically uninteresting, but because they are practically undetectable from astronomical distances. The theoretical framework applies at all scales; the detection proposal applies to the largest and therefore most observable instantiations.

What biological processes drive groups across the SGM threshold, and why do they scale to civilizational scope? In my book I argue that the answer lies in what biologists following Maynard Smith and Szathmáry (1995) call major evolutionary transitions (METs), the great integrative events in the history of life. The defining biological signature of a MET is fitness decoupling and transfer: natural selection shifts from operating primarily at the lower-level unit to operating primarily at the higher-level collective. Individual fitness becomes contingent on collective fitness. In information-theoretic terms, this is the Walker-Davies phase transition expressed evolutionarily: the collective entity’s informational structure stops being a passive record of individual states and starts governing them, because individual reproductive success now runs through the collective.

The eusocial insect colonies documented by Navas-Zuloaga and colleagues are the clearest existence proof. Colony-level cognitive properties, including decision accuracy, attention allocation, and exploration-exploitation balance, emerge from mixtures of individuals with heterogeneous behavioral phenotypes and cannot be predicted from the average individual phenotype alone. The colony implements information processing algorithms just like brains do. These info processing properties emerge from the interaction structure of the whole. This is synergistic information in the formal PID sense: content unrecoverable from any proper subset of sources that has downward causation effects on individual members of the colony. And crucially, none of this depends on the colony’s substrate being biological. The information-theoretic criteria apply to whatever substrate implements the relevant interaction structure.

IV. The Universality Argument: METs as Cosmic Attractors

It is reasonable to expect that alien life, if it exists, will obey principles of natural selection that are operative on Earth. The problems that drive METs are not Earth-specific. They are consequences of physics and mathematics that apply wherever replicating, competing, cooperating entities exist.

The free-rider problem is a mathematical property of replicator dynamics under any physics that permits heritable variation and collective action. Recall that groups that pool synergistic rather than redundant information maximize their growth rate, and this principle, derivable from the formal properties of information alone, applies across scales from individuals forming groups to specialized groups organizing into complex collectives. The pressure toward synergy maximization is not a biological contingency. It is an informational law.

The thermodynamic ceiling on individual cognition is equally universal. Any substrate that processes information at civilization-scale complexity operates under energetic constraints. Distributing cognition across multiple coordinated units, the collective intelligence strategy, is the generic solution to individual cognitive ceilings under energetic constraint, because it accesses synergistic information unavailable to any single unit.

The solution space for these joint constraints is not unlimited. Earth biology has found MET solutions independently at least twelve times in the evolution of eusociality across radically different phylogenetic lineages. This convergence is the signal: MET-like transitions are attractor states in the space of solutions to universal problems. Wherever life achieves civilization-scale complexity, selection pressure has been operating on the same constraints, using the same limited solution space. The information-theoretic signatures of SGMs, high Phi, high synergy, and causal emergence, are convergent outcomes of this attractor dynamics, not parochial accidents.

If MET-like transitions are cosmic attractors, then civilizations that have survived long enough to be detectable have almost certainly undergone them. Their civilization-level information states will exhibit the formal signatures of strong group minds. And this is the central claim: the mathematics that identifies those signatures is the same everywhere in the universe, because the mathematics of information is the same everywhere.

V. The SGM Detection Index: Spectral and Temporal Signatures

How does all this matter for the search for extra-terrestrial life and intelligence (ETI)? If my arguments so far have been sound then it is reasonable to suggest that the information-theoretic analysis of observable state variables I have mentioned above could be added to the SETI methodological toolkit, not as a replacement for electromagnetic signal detection, spectroscopic biosignature analysis, or any other existing approach, but as a complementary analytical tool targeting a different kind of signature. The argument here is not that we have been looking in the wrong places or the wrong things/variables but that we can add other tools to identify informational structure in the data we are already collecting. Existing instruments already gather the multivariate time-series data that what we might call the SGM detection index requires. I am asking whether we could analyze that data for the information-theoretic and network-architectural signature of collective causal organization I described above.

In addition to asking is this molecule present at detectable abundance in this wavelength band, we might also ask does the joint information structure of the full multivariate spectral time-series carry the signature of a system organized by collective cognitive states rather than by abiotic physics or by individually-acting biological agents? This is a question about the mereological architecture of the observed system, how information is distributed across its components and across time.

Several active research programs in astrobiology and technosignature science are probably already computing measures that partially overlap with components of the SGM detection index. Basically they apply information-theoretic measures to multi-channel spectral time series without presupposing specific chemistry, but the question is whether they also compute the joint temporal structure of the full spectral signal rather than individual molecular features. I am not familiar enough with the literature to say for sure but my impression is that they do not do so. The mereological structure of that temporal signal, specifically the PID decomposition into redundant, unique, and synergistic components across spectral channels might yield new information about possible life. I understand that teams looking for alien life are now systematically exploring ensembles of hundreds of molecular species simultaneously rather than targeting single molecules. That kind of work produces precisely the kind of high-dimensional joint spectral data that PID analysis requires. I know that some astrobiologists have argued explicitly for temporal biosignatures as a complement to static chemical ones, using seasonal multi-gas oscillations as the detection target. The temporal structure they are already analyzing is the same temporal structure the SGM index’s decision-making and predictive-processing signatures would be computed from. In each case the SGM index proposes adding a layer of information-theoretic analysis to data pipelines and retrieval frameworks that are already generating the relevant data. The additional computational cost is low, the required data are already being collected or planned for collection.

If advanced ETI is constituted by strong group minds, the absence of narrow-band radio transmissions is not evidence of the absence of intelligence. It is evidence only of the absence of civilizations organized around individual agents who have decided to broadcast. A civilization constituted by a strong group mind does not have individuals who decide to broadcast. Its cognitive activity is organized by collective information states, by synergistic content belonging to the whole and to no individual member. The functional analog of communication for a civilizational SGM may not only be the outward transmission of information packets toward unknown receivers. It might also include the internal generation and maintenance of the integrated information that constitutes the collective entity’s cognitive life. More precisely: a strong group mind’s outputs are organized around maximizing collective synergy, and detecting that kind of activity would be a clue that we are dealing with ETI.

Patrick McNamara is Professor of Psychology at National University and Associate Professor of Neurology at Boston University School of Medicine. He is the author of The Cognitive Neuroscience of Religious Experience (Cambridge, 2022), From Groups to Gods (Cambridge, 2026), and The Anarch as Katechon.

References for this post available upon request