A book-club companion · science, philosophy, culture
If the book were written today, its core provocation would remain: a non-foresighted process can accumulate astonishing competence and purposive-looking organization. But its “universal acid” would need a safety label. Evolutionary history explains much; it does not automatically settle consciousness, free will, truth, moral obligation, political legitimacy, or whether an AI understands.
The 2026 thesis: keep design without a designer; replace one master algorithm with a plural picture of inheritance, development, ecology, culture, institutions, power, and historical constraint. Ask what process is actually doing the work before calling it Darwinian.
Natural selection can make biological design intelligible without a prior designer. “Cranes” are gradual, mechanistic explanations; “skyhooks” are unexplained design from above. This remains a powerful demand for continuity.
Selection is not the only process relevant to evolution. Mutation, drift, and gene flow can directly change genetic frequencies. Development and constraint shape the variation on which selection can act; niche construction and symbiosis can alter environments and inheritance; cultural transmission has additional, non-genetic mechanisms. None is a supernatural exception or makes selection unnecessary.
Whether functional organization explains experience; what makes a reason binding; whether compatibilist responsibility is enough; and whether an AI genuinely understands are persistent philosophical disputes.
Dennett’s contrast still helps. A crane explains complexity by showing the intermediate mechanisms that build it; a skyhook merely announces that intelligence, purpose, or meaning arrived from somewhere beyond explanation. The revision is not to call every higher-level cause a skyhook. Developmental systems, ecological feedback, language, schools, archives, and institutions are scaffolds: historically made structures that alter what is learned, retained, or selected.
World Machines application: do not treat “the machine” as a new hidden designer. A world-machine explanation earns its name only when it identifies material conditions, institutions, inheritance channels, actors, and evidence—rather than replacing great persons with a grander abstraction.
Biologists may say that hearts are “for” pumping blood without claiming that future blood-pumping caused hearts to appear. The phrase can abbreviate a selection-history claim. But function talk is disciplined only when alternatives have been considered: drift, by-product, developmental constraint, trade-off, and exaptation—an older feature recruited to a new role.
A technology can acquire uses beyond those for which particular communities first adopted it: writing moved from accounting into literature, law, theology, and philosophy. That can be a useful analogy to exaptation, but it is not evidence that cultural artifacts underwent biological selection. Establish the historical mechanism independently.
Terms used on this page: the intentional stance predicts a system by treating it as a rational agent with beliefs and goals; compatibilism says meaningful freedom can coexist with causal determination; basic desert is the idea that someone deserves blame in itself, not only for protection, repair, or deterrence; normativity concerns what one should believe or do, not merely how a belief arose; antimemetic here means difficult to notice or circulate publicly, not magically impossible to transmit.
Dennett’s intentional stance predicts a system by treating it as if it had beliefs, information, and goals. This is often an excellent strategy for people, animals, institutions, and machines. His anti-Cartesian point also survives: neuroscience has not found a single inner theater where a little observer watches the final show.
But four questions must stay apart: task competence, representation or understanding, phenomenal consciousness, and moral standing. A system can be predictively describable in intentional language without thereby being conscious or responsible. Contemporary consciousness science has active, testable theories—global-workspace, integrated-information, recurrent-processing, higher-order, predictive accounts—but no accepted final theory.
David Chalmers argues that even a complete explanation of cognitive function leaves a question about why there is subjective experience at all: the “hard problem.” Dennett thinks this residual demand rests on confused intuitions about qualia. This is not a fight that neuroscience has simply won for either side; it is a live dispute about what counts as an adequate explanation.
Dennett’s compatibilism treats the freedom worth wanting as a natural capacity: anticipating consequences, learning, responding to reasons, and altering conduct. It makes room for coercion, trauma, ignorance, impairment, and rehabilitation. Modern neuroscience can illuminate those capacities, but it does not by itself declare compatibilism true or false.
The sharp contemporary contrast is between a competence-based view (Dennett; Christian List) and basic-desert skepticism (Derk Pereboom; Robert Sapolsky in a public-science register). The practical question is not “brain or person?” A person acts through a brain. It is: which capacities justify answerability, protection, repair, or punishment—and which forms of retribution can survive a causal understanding of human lives?
Dennett’s meme was a useful provocation: culture can preserve, transform, and select information, helping minds become less miraculous. But a 2026 account starts with mechanisms rather than meme slogans. Who learns from whom? Is the transmission imitation, teaching, law, writing, platform ranking, a professional standard, or coercion? How faithfully is it retained? What pays off—truth, prestige, revenue, solidarity, compliance, attention?
Culture is often reconstructive rather than a high-fidelity copy. It can also be cumulative: language, tools, diagrams, archives, and institutions preserve improvements no individual could reinvent alone. A notebook, search engine, curriculum, or platform may become part of a coupled cognitive system, but only under particular conditions of access, reliance, and control.
Memetic success is not operational power. Public faces, slogans, and charismatic people may be highly visible, while standards, budgets, procurement rules, interface defaults, and administrative discretion do much of the causal work. “It went viral” is not yet an explanation.
Modern AI makes Dennett’s questions concrete. We can often use intentional language to predict a model’s output, yet a mechanistic story of training data, optimization, architecture, and deployment may tell a very different causal story. Evolutionary computation shows that selection-like search can find solutions; it does not show that all optimization is Darwinian, that a trained model understands, or that it feels.
Can it do a task robustly? Test with evaluation, errors, and real-world conditions.
Does it use grounded representations with stable causal roles? This remains theoretically contested.
Does it have experience? There is no agreed decisive test for current systems.
Who controls, benefits, and bears harms? This is institutional and ethical, not a benchmark score.
Natural selection can help explain why a detector tracks something, why a social practice persists, or how creatures capable of giving reasons arose. It does not automatically establish that a belief is true or that a value is binding. This is the central boundary for a disciplined naturalism.
Orientation, mapping, and inquiry have deep biological roots. Yet a survival-linked need for orientation does not prove that Socratic dialogue is the unique or highest cultural response. Ritual, narrative, meditation, scientific inquiry, and political struggle may organize orientation differently. Evolution makes naturalistic accounts of meaning a serious research program; it does not close the argument about truth or value.
These are routes through questions, not a ranked canon. A documented disagreement means authors directly contest a position; a productive tension means the pairing is curatorial.
SELECTION & EXPLANATION · EXTENDS
Darwinian processes come in degrees: identify a real population, inheritance, variation, and differential reproduction before calling any change “selection.” He also makes animal minds and agency comparative, graded questions.
Read: Darwinian Populations and Natural Selection (2009) · accessible start: Other Minds.
SELECTION & EXPLANATION · REVISES
Sober separates selection from drift and adaptationist storytelling; Okasha models selection at genes, organisms, and groups. Their shared question: which level and mechanism actually does causal work?
Sober, The Nature of Selection · Okasha, Evolution and the Levels of Selection.
DEVELOPMENT · REVISES
Developmental plasticity and niche construction alter the variation and environments selection encounters. They do not replace selection; they challenge a picture of organisms as passive recipients of genes and environments.
West-Eberhard, Developmental Plasticity and Evolution · Laland et al. on the Extended Synthesis.
PLURALISM · CHALLENGES
Dupré treats organisms as multi-level processes rather than fixed machines; Longino treats objectivity as public, critical inquiry among differently situated investigators. Both resist a single explanation swallowing every other one.
Dupré, Processes of Life · Longino, Studying Human Behavior.
MIND · EXTENDS
Clark’s extended mind treats tools and environments as possible parts of cognition; Seth’s predictive account treats perception and selfhood as controlled, embodied construction. Neither implies that constructed experience is unreal.
CONSCIOUSNESS · DOCUMENTED DISAGREEMENT
Chalmers argues that explaining function and report may still leave why there is experience—the hard problem. This is the cleanest philosophical challenge to Dennett’s strategy, not a rejection of neuroscience.
AGENCY · DOCUMENTED DISAGREEMENT
List defends persons as genuinely reasons-responsive higher-level agents; Pereboom rejects the basic-desert responsibility that compatibilists hope to retain. Both reject a ghostly soul; they disagree about what causal agency can justify.
List, Why Free Will Is Real · Pereboom, Living Without Free Will.
CULTURE · SPECIFIES
Dual-inheritance and cultural-evolution work replace memes as a default unit with learning biases, teaching, networks, artifacts, demography, and institutions. Culture can change genes—and can persist for prestige, coercion, or lock-in rather than biological benefit.
Boyd & Richerson, Culture and the Evolutionary Process · Henrich, The Secret of Our Success.
LEARNING · REVISES
Sterelny emphasizes apprenticeship and constructed cognitive niches; Heyes argues that many sophisticated human “cognitive gadgets” are culturally assembled. Minds may be built by culture, not merely hosts for culture.
Sterelny, The Evolved Apprentice · Heyes, Cognitive Gadgets.
AI & COMPLEXITY · QUALIFIES
Mitchell separates impressive task competence from general intelligence; Crawford reconnects AI to labor, minerals, energy, data, and institutional power. Together they ask what an intentional description hides.
SITUATED SCIENCE · REFRAMES
Subramaniam traces colonial, racial, and gendered inheritances in evolutionary narrative; TallBear shows why genetic ancestry cannot simply settle Indigenous belonging, kinship, or sovereignty. In contexts of ancestry testing and Indigenous governance, classification has social and political consequences as well as technical ones.
Subramaniam, Ghost Stories for Darwin · TallBear, Native American DNA.
MULTISPECIES & CARE · REFRAMES
Haraway, Kimmerer, and Tsing are not a single school or a technical alternative to population genetics. Read them for different projects: Haraway on situated multispecies relations; Kimmerer on reciprocal responsibilities alongside plant science and Potawatomi teachings; and Tsing on capitalist supply chains, labor, and life in damaged landscapes. These are ethical and historical lenses, not evidence that revises a population-genetic mechanism.
Haraway, Staying with the Trouble · Kimmerer, Braiding Sweetgrass · Tsing, The Mushroom at the End of the World.
A useful historical caution is that contemporaries can overread coordination and intention from shared conditions. That supports an analogue of “design without a single designer.” But it does not make actors puppets, nor turn a world machine into an agent.
The better practice is a selection audit: name candidate variants; the inheritance channel; the filter or retention mechanism; the material and institutional substrate; the relevant timescale and level; evidence; alternatives; and deviations. Preserve branching interpretations and visible departures from expectation rather than force anonymous convergence around a perfect model. This is a historical research heuristic, not a claim that institutional change is biological natural selection.
Three tempting but invalid moves: “it persisted, so it is good”; “it is visible, so it caused the outcome”; “it has a history, so it was inevitable.” Selection, transmission, and complexity are descriptions of processes. They do not award truth, legitimacy, or moral worth.