Toward an Algorithmic Theory of Cultural Cognition
This is the speculative research paper that serves as the scientific foundation for the Executable Minds fictional universe. The world-building bible was derived from this paper's framework. The paper proposes a formal reframing of the memeplex — traditionally understood as a self-reinforcing cluster of culturally transmitted ideas — as a computational object possessing the structural and functional properties of an algorithm.
This paper proposes a formal reframing of the memeplex — traditionally understood as a self-reinforcing cluster of culturally transmitted ideas — as a computational object possessing the structural and functional properties of an algorithm. Drawing on Richard Dawkins's original concept of the meme as a unit of cultural replication and Susan Blackmore's subsequent development of the memeplex as a co-adapted package of memes, we argue that an essential dimension has been undertheorized: the algorithmic execution of memeplexes within individual cognition.
A memeplex, on this account, is not merely a set of beliefs that propagates through imitation. It is a transfer function — a structured mapping from perceptual inputs to behavioral outputs, mediated by weighted evaluative rules concerning value, threat, possibility, and social meaning. We develop this thesis through five disciplinary lenses: computational theory, cognitive science, evolutionary memetics, social psychology, and philosophy of mind.
In 1976, Richard Dawkins introduced the concept of the meme as a unit of cultural transmission, analogous to the gene in biological evolution. Ideas, tunes, fashions, and techniques could replicate themselves by leaping from brain to brain through imitation. Susan Blackmore extended the framework by introducing the memeplex: a group of memes that replicate more effectively together than any single meme could alone. Religious systems, political ideologies, and scientific paradigms could all be understood as self-reinforcing packages of ideas that resist decomposition and promote their own transmission.
These foundational contributions established memetics as a field but left a critical dimension underexplored. Both Dawkins and Blackmore focused primarily on replicative dynamics — how memes spread, compete, and form coalitions. What neither framework fully articulated was the computational dimension: the fact that a memeplex, once adopted, does not merely sit passively in memory. It runs. It executes. It actively processes incoming information, applies evaluative weightings, generates predictions, and produces behavioral outputs. The person hosting the memeplex experiences these outputs not as the consequences of a program, but as the texture of reality itself.
This paper proposes that the memeplex is best understood as an algorithm in the technical sense: a finite sequence of well-defined operations that maps inputs to outputs through a transfer function. This reframing is not merely metaphorical. It has precise formal implications, empirical connections to contemporary cognitive science, evolutionary explanatory power, social-psychological consequences, and deep philosophical ramifications.
Dawkins's original formulation positioned the meme as a replicator subject to variation, selection, and retention. The concept's abstraction was both its strength and its vulnerability — it permitted application across an enormous range of phenomena but invited criticism that the meme was too loosely defined for rigorous inquiry. Critics such as Dan Sperber argued that cultural transmission rarely involves high-fidelity copying, and that the gene analogy breaks down at the mechanistic level. These objections demand refinement rather than abandonment. The algorithmic reframing proposed here represents one such refinement.
Blackmore recognized that memes rarely operate in isolation. A religious memeplex typically includes memes about the existence of a supreme being, the authority of certain texts, the importance of ritual, the danger of doubt, the virtue of proselytization, and the promise of reward for fidelity. These memes are co-adapted: each reinforces the others, and the package is more stable and transmissible than any individual component. Critically, the memes within a memeplex don't merely coexist; they form conditional dependencies, feedback loops, and immunity responses against competing memes. These are not merely properties of a cluster. They are properties of a program.
We define a memeplex M formally as an ordered tuple M = ⟨ I, O, W, T ⟩ where I represents the input domain (all stimuli available to the agent), O represents the output domain (all behavioral responses available), W represents a set of weighted evaluative parameters, and T represents the transfer function T: I × W → O that maps weighted inputs to outputs.
The transfer function T is the core of the algorithmic memeplex. When an individual encounters a novel piece of information, the memeplex does not passively record it. Instead, T applies the evaluative weights in W to classify the input along dimensions the memeplex has defined as salient: Is this safe or dangerous? Is this person ally or threat? Is this idea compatible or heretical? The outputs are not just beliefs but action-readiness states: approach or avoid, engage or withdraw, signal affiliation or opposition, transmit or suppress.
The weight set W is where ideological content resides. Different memeplexes assign different weights to the same inputs, producing radically different outputs from identical stimuli. Two individuals observing the same political protest — one running high threat-to-social-order weights, the other running high courage-of-conviction weights — will generate diametrically opposed emotional responses, behavioral dispositions, and downstream decisions. Neither experiences themselves as running a program. Both experience themselves as seeing reality clearly.
The weight set is not static. It undergoes modification through experience, social reinforcement, and internal feedback. But failed predictions are rarely allowed to update W directly; instead, failures are routed through explanatory subroutines that preserve the existing weight structure. This is the computational equivalent of motivated reasoning, and it is a designed feature of the architecture, not a defect.
A critical property is the memeplex's capacity for recursive self-maintenance. The transfer function T also monitors and regulates the integrity of W itself. If an input threatens to destabilize a core weight, T activates defensive subroutines: discrediting the source, reframing the evidence, generating emotional responses that motivate disengagement, or seeking social reinforcement from compatible agents. This makes the memeplex self-stabilizing: the same mechanism that processes the world also protects itself from revision.
The predictive processing framework, associated with Karl Friston, Andy Clark, and Jakob Hohwy, provides a natural computational substrate. The brain is fundamentally a prediction machine: it maintains hierarchical generative models and continuously generates predictions about incoming sensory data. Perception is not passive reception but active inference. A memeplex, in this framework, functions as a high-level set of Bayesian priors. The weight set W corresponds to precision-weighted expectations that shape interpretation of ambiguous inputs. These biases operate at the perceptual level — the memeplex shapes what the agent literally sees, not merely what they think about what they see.
In predictive processing, learning occurs through minimization of prediction error. This creates a potential vulnerability for memeplexes: sustained errors should drive revision. However, the framework also identifies mechanisms that attenuate prediction error without revising priors. Active inference involves acting on the environment to make it conform to predictions rather than updating predictions. An agent predicting social hostility may behave defensively, eliciting the very hostility predicted — confirming the prior without revision. This is the computational mechanism underlying self-fulfilling prophecies.
Furthermore, agents can selectively sample their environment — attending to confirming information and avoiding disconfirming sources. This epistemic foraging provides the computational basis for confirmation bias. The information diet is a downstream output of the memeplex's transfer function.
The algorithmic perspective illuminates why memeplexes are cognitively attractive. A well-functioning memeplex provides comprehensive, internally consistent priors that dramatically reduce the computational burden of moment-to-moment decision-making. Without one, every novel situation would require assessment from first principles. The memeplex is a heuristic package — efficient approximations trading accuracy for speed and cognitive economy. This efficiency comes at a cost: the interdependence of weights creates a strong cognitive incentive to preserve the set as a whole, even when individual weights produce inaccurate predictions. The agent experiences this as epistemic confidence, not as bias.
The unit of competition is the transfer function itself — the specific mapping from inputs to outputs. Selection favors transfer functions that maximize their own replication, which is not the same as maximizing accuracy or host wellbeing. High-arousal emotional outputs (outrage, fear, tribal solidarity) replicate more effectively than calibrated, nuanced assessments. The selection pressure thus favors emotional salience over epistemic accuracy, social signaling value over predictive validity, and transmissibility over truth.
Memeplex speciation occurs when populations become informationally isolated. Digital media environments, with algorithmic content curation and self-sorting social networks, dramatically accelerate this process by creating informational niches. The result is the cultural equivalent of allopatric speciation: the divergence of populations into increasingly distinct and mutually incomprehensible forms, each locally optimized within its own niche but increasingly maladapted to the broader environment.
Transfer functions co-evolve with media infrastructure. Oral cultures favor narrative coherence and interpersonal trust. Print cultures favor logical argumentation and institutional authority. Digital cultures favor emotional arousal, rapid ally/enemy classification, meme-format compressibility, and viral shareability. The memeplexes currently achieving maximal fitness are those best adapted to digital affordances: short attention spans, algorithmic amplification, visual and emotional processing, and network-based social validation.
What constitutes a social group, at the cognitive level, is not merely shared category membership but shared execution of a compatible transfer function. Two agents are in-group members to the extent that their memeplexes produce similar outputs from similar inputs. This explains why ideological alignment often feels more fundamental than demographic similarity: it represents shared cognitive architecture, not merely shared category membership.
Group polarization can be understood as transfer function optimization within a memeplex-aligned population. Arguments that push toward more extreme outputs are socially rewarded; arguments that moderate are socially penalized. The result is a ratchet effect: each round of interaction moves the group's aggregate transfer function toward greater extremity, greater confidence, and greater divergence from out-group functions. This is self-accelerating as moderate positions become increasingly indistinguishable from out-group positions.
Adopting a memeplex is not acquiring beliefs. It is installing a new set of evaluative weights and training a new transfer function. The process rewires what the agent notices, feels, ignores, and how they respond. Deconversion is correspondingly traumatic — it requires uninstalling a transfer function that has structured perception, emotion, and behavior, often for decades. The phenomenological experience of deconversion — the sense that the world has become unreadable — is exactly what we would expect from the removal of an operating system's core processing module.
The most philosophically significant feature is phenomenological transparency. The transfer function T does not present itself as a program. It presents itself as reality. The outputs of T are experienced not as computed results but as perceptions, intuitions, and self-evident truths. When T classifies an input as threatening, the agent experiences fear, not a computation. The memeplex operates below the threshold of reflective awareness, just as the visual system's edge-detection algorithms operate below visual awareness. We do not see edges being computed; we see edges.
The framework does not entail hard determinism. It entails that the space of choices available to an agent is structured, constrained, and weighted by the memeplex, but not absolutely determined. Agents possess, to varying degrees, the capacity for metacognition — the ability to notice the operation of their transfer function and intervene in its execution. The relevant question is not whether an agent is free from memeplex influence — no agent is — but whether they possess sufficient metacognitive capacity to recognize that they are running a transfer function and evaluate its outputs critically.
Freedom, on this account, is not the absence of computational constraint but the presence of recursive self-monitoring — the capacity to run the algorithm while simultaneously observing that one is running it.
The existence of metacognitive capacity introduces a graded notion of responsibility. Agents who have been exposed to algorithmic memetics bear a different epistemic responsibility — the obligation to periodically examine their own cognitive processes for signatures of memeplex operation.
The five disciplinary perspectives converge: computational theory provides the formal architecture; cognitive science supplies the neural substrate; evolutionary memetics identifies the selection dynamics; social psychology maps the interpersonal dynamics; and philosophy of mind confronts the deepest implications for consciousness and agency.
Together, these perspectives suggest that understanding that one is running a transfer function does not, by itself, free one from that transfer function's influence. But it creates the possibility of a different relationship to one's own cognition — a relationship characterized by what we might call computational humility: the ongoing recognition that one's perceptions, evaluations, and behavioral dispositions are the outputs of a culturally installed program, not unmediated apprehensions of reality.
The formal model is deliberately abstract — we have not specified the computational complexity of T, the dimensionality of W, or the precise mechanisms of weight updates. The relationship between the algorithmic memeplex and predictive processing is theoretical, not yet empirically established. Not all culturally transmitted structures may be best understood as algorithmic memeplexes. And there is a reflexivity problem: algorithmic memetics is itself a memeplex, subject to the same dynamics it describes. Whether the framework's self-application is coherent or self-undermining remains an open philosophical question.
The framework suggests several empirical research programs: presenting identical stimuli to individuals with divergent ideological commitments and measuring differences in perceptual processing at the earliest stages; historical analysis of how dominant memeplex structures have shifted across major media transitions; and practical implications for education, deliberative democracy, and conflict resolution. If ideological conflict is rooted in divergent transfer functions rather than mere disagreements about fact, then interventions must address the perceptual and affective infrastructure that generates the disagreement, not merely the propositional content.
The memeplex is not merely a cluster of beliefs that travels through a population. It is a structured computational object — an algorithm with a defined transfer function that maps perceptual inputs to behavioral outputs through a set of culturally installed evaluative weights. The concept of the memeplex as executable code does not reduce human cognition to mere mechanism. It illuminates the specific mechanisms through which cultural meaning is produced, sustained, and transmitted — and in doing so, it opens the possibility of a more reflective engagement with the software that shapes our experience of the world. The capacity for metacognition — for running the algorithm while observing that one is running it — remains the distinctive human resource in this picture. Whether that capacity can be cultivated at sufficient scale to meet the epistemic challenges of the present moment is perhaps the most important open question that algorithmic memetics poses.
Blackmore, S. (1999). The Meme Machine. Oxford University Press. · Boyd, R., & Richerson, P. J. (2005). The Origin and Evolution of Cultures. Oxford University Press. · Clark, A. (2013). Whatever Next? Predictive Brains, Situated Agents, and the Future of Cognitive Science. Behavioral and Brain Sciences, 36(3). · Dawkins, R. (1976). The Selfish Gene. Oxford University Press. · Dennett, D. C. (1991). Consciousness Explained. Little, Brown. · Friston, K. (2010). The Free-Energy Principle. Nature Reviews Neuroscience, 11(2). · Haidt, J. (2012). The Righteous Mind. Vintage. · Henrich, J. (2016). The Secret of Our Success. Princeton University Press. · Hohwy, J. (2013). The Predictive Mind. Oxford University Press. · Mercier, H., & Sperber, D. (2017). The Enigma of Reason. Harvard University Press. · Sunstein, C. R. (2009). Going to Extremes. Oxford University Press. · Tajfel, H., & Turner, J. C. (1979). An Integrative Theory of Intergroup Conflict. · Zuboff, S. (2019). The Age of Surveillance Capitalism. PublicAffairs.
The World Rules (derived from this paper) · Garment District · Five Frequencies · The Reunion