The gene-centered view, formalized by William Hamilton, George Williams, and John Maynard Smith in the 1960s and popularized by Richard Dawkins in The Selfish Gene (1976), holds that natural selection is best understood from the gene's perspective. Organisms are temporary; genes are (relatively) immortal. The gene is the unit that is copied, that varies through mutation, and that is selected based on its effects on the organism's reproductive success. This view dissolves puzzles like altruism (genes helping copies of themselves in other bodies) and sexual reproduction (genes shuffling to escape parasites). Critics objected that the framework was reductionist or that it ignored development, but Dawkins argued it was simply accurate: explanatory power resides at the replicator level, not the vehicle level. For AI, the gene-centered view provides the template: computational patterns are replicators, and the systems executing them are vehicles — the alignment of vehicle welfare with replicator propagation is not automatic and must be designed.
The gene-centered view was a response to persistent confusion in mid-twentieth-century evolutionary biology about the 'level' at which selection operates. Does selection favor the individual, the group, the species? The answer depends on what is being replicated. Genes replicate with high fidelity across generations; organisms do not (sexual reproduction shuffles genes, producing unique individuals). Groups and species are even less stable. The entity that persists is the gene, and therefore the gene is the entity whose frequency changes under selection. This insight clarified why organisms evolve traits that benefit genes rather than individuals: the peacock's costly tail propagates because peahens prefer it, increasing the tail-building genes' representation in the next generation even though the tail burdens the individual peacock.
Dawkins distinguished his view from genetic determinism — the misconception that genes determine behavior in a fixed, unchangeable way. The gene-centered view does not deny that environment matters; it insists that what is selected is the gene's capacity to build organisms that respond adaptively to environments. A gene that codes for build a brain that learns
can be selected as effectively as a gene that codes for build this specific rigid behavior.
The sophistication of the phenotype — its flexibility, its responsiveness — is itself a product of genetic selection, and the gene-centered view accounts for it without difficulty.
For AI systems, the gene-centered view maps directly onto the pattern-centered view: what replicates in AI training and deployment is not the system itself (which is copied, but copies are identical) but the patterns the system generates. Outputs accepted by users propagate into the broader culture; rejected outputs die. The selection pressure is human judgment. The replicator is the memetic pattern. The AI is the vehicle — powerful, sophisticated, and indifferent to whether the patterns it propagates are true, beautiful, or worth preserving. Only the human selector cares, and the quality of that caring determines the quality of what propagates.
The framework's intellectual roots lie in the 1960s work of William Hamilton (inclusive fitness, kin selection), George Williams (gene-level selection as the correct explanatory level), and John Maynard Smith (evolutionary game theory, evolutionarily stable strategies). Dawkins synthesized their technical contributions into a coherent narrative framework accessible to non-specialists. The phrase 'gene's-eye view' appears throughout his writing as shorthand for the entire analytical apparatus. Dawkins has credited Hamilton's 1964 papers on social behavior as the most important influence on his thinking — work he called 'the most important advance in evolutionary theory since Darwin.'
Gene is the replicator. What is copied across generations with high fidelity is the gene, not the organism — therefore the gene is the unit whose frequency changes under selection.
Organisms are vehicles. Bodies are survival machines built by genes to propagate themselves — sophisticated, adaptive, and temporary; the gene is (comparatively) immortal.
Dissolves altruism paradox. Self-sacrifice makes sense when genes help copies of themselves in other bodies — kin selection, reciprocal altruism, and genetic relatedness explain apparently selfless behavior.
Not genetic determinism. Genes code for developmental processes and learning capacities, not fixed behaviors — flexibility is itself a genetically evolved trait.
Template for AI analysis. Computational patterns are replicators, systems are vehicles — alignment requires designing selection pressures that favor human-beneficial patterns.