The formal modeling of memes as decision-making agents with preferences, utility functions, or goal-directedness is largely absent from academic literature. Despite extensive use of agential language in memetics and cultural evolution, no established research program treats memes as literal agents in the decision-theoretic sense. What exists instead is a spectrum of positions—from purely metaphorical usage to methodological heuristics to near-literal agency claims—alongside sharp critiques questioning whether such attribution is coherent at all.
The foundational theorists themselves disagree fundamentally: Richard Dawkins treats meme agency as pedagogical metaphor, Daniel Dennett defends it as a useful explanatory stance, and Susan Blackmore approaches treating it literally. Meanwhile, mainstream cultural evolution researchers (Boyd, Richerson, Henrich) explicitly reject the "meme's eye view" entirely, modeling cultural variants as passive information subject to human decision-making.
The most striking finding is what doesn't exist. Despite searching for game-theoretic frameworks treating memes as players, decision-theoretic models with meme utility functions, Bayesian frameworks for meme belief-updating, or agent-based models where cultural variants are the agents, no such formal apparatus has been developed.
The closest approximation is philosopher Samir Okasha's Agents and Goals in Evolution (2018), which rigorously examines whether evolved entities can legitimately be treated as utility-maximizing agents. Okasha asks whether the fitness-maximizing paradigm of evolutionary biology can be mapped to rational choice theory's utility-maximizing paradigm, distinguishing "Type 1" agential thinking (about products of selection—justified) from "Type 2" (about the process—problematic). However, Okasha focuses entirely on biological evolution and does not extend his framework to cultural replicators.
Jonathan Birch's The Philosophy of Social Evolution (2017) develops a "cultural inclusive fitness" theory using a meme's-eye perspective, deriving a cultural analogue of Hamilton's rule (rb > c, where r = cultural relatedness). Birch explicitly employs the framing "Suppose you are a meme... you have two ways to achieve your goal." Yet he includes a critical caveat: "Memes are a ladder we can kick away." The meme-as-agent perspective serves as a heuristic tool to derive mathematical results, not as a model of literal meme decision-making.
In standard evolutionary game theory, memes appear only as strategies being selected, never as players making strategic decisions. The replicator dynamics equation treats strategy frequencies as evolving based on payoffs, but the "agency" is implicit selection pressure, not meme-level decision-making.
The three foundational figures in memetics hold strikingly different positions on this question:
Richard Dawkins (1976–present) introduced memes in The Selfish Gene as cultural replicators analogous to genes, using agential language throughout. However, in the 1989 second edition, he "states unequivocally that he does not intend to imply that they are driven by any motives or will, but merely that their effects can be metaphorically and pedagogically described as if they were." His famous claim that "we alone on earth can rebel against the tyranny of the selfish replicators" preserves human agency while treating meme selfishness as purely metaphorical. Dawkins explicitly calls memes "unconscious, blind replicators."
Daniel Dennett (1991–2017) develops the most sophisticated philosophical treatment through his "intentional stance"—the strategy of explaining behavior by attributing beliefs, desires, and intentions to a system. In Darwin's Dangerous Idea and From Bacteria to Bach and Back, Dennett defends the "meme's eye view" as a legitimate application of this stance. He argues that cultural entities may "evolve according to selectional regimes that make sense only when the answer to the Cui bono question is that it is the cultural items themselves that benefit." Yet Dennett characterizes this as "quasi-agency"—"all that purposive hustle and bustle, and yet there's nobody home." His position is methodological rather than metaphysical: we need not claim memes literally have beliefs/desires, only that treating them as if they did yields predictive and explanatory power.
Susan Blackmore (1999–present) takes the most radical position in The Meme Machine. She explicitly argues that memes are "the beneficiaries, not the genes, and certainly not us—their creatures." She directly challenges Dawkins: "Dawkins claims that we alone can rebel against the tyranny of the selfish replicators but really... there is no one to rebel." On her view, the self is "a selfplex"—just another memeplex creating the illusion of a unified agent. Blackmore approaches treating meme agency literally, arguing that "we are meme machines through and through" with no genuine free will.
| Theorist | Position on agency | Metaphorical vs. literal | Human free will |
|---|---|---|---|
| Dawkins | Pedagogical metaphor only | Explicitly metaphorical | Preserved |
| Dennett | Quasi-agency via intentional stance | Methodological stance | Complicated |
| Blackmore | Genuine autonomous replicators | Approaches literal | Denied |
The dual inheritance theory tradition—represented by Robert Boyd, Peter Richerson, Joseph Henrich, and Luigi Luca Cavalli-Sforza—explicitly rejects treating cultural variants as agents. This is not a matter of emphasis but of theoretical commitment.
Their 2008 paper "Five Misunderstandings About Cultural Evolution" disputes that discrete replicators are necessary for cumulative evolution, that cultural variants must be gene-like particles, or that "cultural fitness" can be inferred from transmission success. Richerson and Boyd (2005) state plainly: "Cultural variants are not replicators."
In a 2023 paper, "Agentic processes in cultural evolution," Richerson, Boyd, and Efferson carefully distinguish agentic from non-agentic forces in cultural change. Their list of agentic forces—individual learning, selective social learning, partner selection, collective decision-making—are all exercised by humans, not by cultural variants. Cultural variants themselves are modeled as passive information subject to transmission biases (conformist bias, prestige bias, content bias) that are properties of human learners.
In agent-based computational models throughout this tradition, agents are always humans or simulated individuals, never cultural variants themselves. The Axelrod model of cultural dissemination, Acerbi and Mesoudi's tutorial frameworks, and cumulative culture simulations all treat cultural traits as what agents carry, not as agents themselves.
Susan Blackmore, reviewing Boyd and Richerson's Not By Genes Alone, noted their explicit rejection of memetics because their population approach "does not imply that cultural evolution is analogous to genetic evolution" and does not depend on "discrete, faithfully replicating, genelike bits of information."
The philosophical literature reveals no consensus, with arguments ranging from pragmatic defense to rejection of the entire framework.
For meme agency: Dennett's pragmatic defense represents the strongest philosophical case—the intentional stance applied to memes is a useful explanatory tool regardless of metaphysical status. He argues we should ask "Cui bono?" and recognize that cultural items themselves may benefit from their adaptations. Dawkins endorsed philosopher N.K. Humphrey's suggestion that "memes should be considered as living structures, not just metaphorically."
Against meme agency: Dan Sperber's critique in "An Objection to the Memetic Approach to Culture" (2000) challenges the replicator status of memes entirely. He argues cultural items are "re-produced" (produced many times) but not truly "reproduced" (copied from one another). Cultural recurrence is explained by cognitive "attractors"—shared biases that reconstruct similar representations—not copying. If memes aren't genuine replicators, attributing agency to them is doubly problematic.
Kim Sterelny's "Memes Revisited" (2006, British Journal for Philosophy of Science) offers a nuanced middle ground: some cultural traits may be meme-like, but many are not. He questions whether memes form genuine lineages, since ideas often have multiple "parents," and argues cumulative cultural evolution doesn't require cultural replicators. Stephen Jay Gould called memetics a "meaningless metaphor." Ernst Mayr declared the meme an "unnecessary synonym" for 'concept.'
The category error argument runs through multiple critiques. William Wimsatt notes cultural elements have "no stable analogue to the genome"—different elements temporarily acquire "replicator status" depending on human attention. C.R. Hallpike argues that for memes to be "selfish," they must have a "self" with "defined existence in the world"—a criterion difficult to satisfy. One critic (Benzon) characterized Dennett's active meme as "physically impossible and conceptually empty... cultural preformationism."
Several factors explain why formal meme-as-agent models haven't been developed:
Academic literature contains extensive discussion of whether memes should be treated as having agency, but essentially no formal models that actually do so. The "meme's eye view" exists as a rhetorical device (Dawkins), a methodological stance (Dennett), and an approach toward literal ontological commitment (Blackmore)—but never as a developed mathematical framework with meme utility functions, game-theoretic analysis of meme strategies, or decision-theoretic models of meme choice.
This represents either a genuine gap awaiting development or—as mainstream cultural evolutionists and philosophical critics would argue—a framework correctly recognized as incoherent or unnecessary. The strongest theoretical foundation for potential development is Okasha's Agents and Goals in Evolution, which provides rigorous criteria for when agential language about evolved entities is justified. However, no one has systematically extended this framework to cultural replicators. The meme-as-agent model remains philosophically contested, formally undeveloped, and explicitly rejected by the dominant tradition in cultural evolution research.