In 1994, a NASA advisory committee offered what seemed like a straightforward working definition: life is a self-sustained chemical system capable of undergoing Darwinian evolution. Three decades later, that definition remains contested, riddled with counterexamples, and quietly insufficient for the very missions it was designed to guide. The question it tried to answer—what is life?—turns out to be one of the most stubborn problems at the intersection of philosophy and the natural sciences.

This isn't merely a semantic puzzle. When we send probes to the subsurface oceans of Europa or analyze the atmospheric chemistry of exoplanets, what counts as a biosignature depends entirely on what we take life to be. When researchers engineer protocells or design software that exhibits open-ended evolution, our assessment of whether they've created something genuinely alive hinges on a definition we haven't secured. The ontological stakes are real, and they ripple outward into epistemology, methodology, and the allocation of billions in research funding.

What I want to argue here is that the persistent failure to produce necessary and sufficient conditions for life isn't a sign that we need to try harder. It's a sign that we're asking the wrong kind of question. Drawing on developments in philosophy of biology, astrobiology, and artificial life research, I'll suggest that life may resist classical definition not because of our ignorance, but because of the kind of concept it is—and that recognizing this has profound practical and philosophical consequences.

The history of attempts to define life reads like a series of increasingly sophisticated failures. Consider the three dominant approaches. Metabolic definitions identify life with self-maintaining thermodynamic processes—organisms as dissipative structures that sustain themselves far from equilibrium. Reproductive definitions foreground the capacity for self-replication with heredity. Evolutionary definitions, like NASA's, require participation in Darwinian evolution—variation, heredity, and differential fitness.

Each captures something important. And each falls to well-known counterexamples. Fire is a self-sustaining chemical process far from thermodynamic equilibrium, yet no one calls it alive. Mules are paradigmatically biological organisms that cannot reproduce. Viruses undergo Darwinian evolution but lack autonomous metabolism—and their status remains genuinely contested. A single organism in a species of one could be alive without participating in any population-level evolutionary dynamics.

The counterexample problem is deeper than it first appears. It isn't that these definitions are approximately right and need minor refinement. The counterexamples often target the very feature each definition takes to be essential. Crystals grow and replicate their structure. Prions propagate with heredity. Computer viruses reproduce with variation. The boundaries between the living and the nonliving, approached from any single axis, turn out to be disturbingly porous.

Some theorists have responded by conjoining criteria—life requires metabolism and reproduction and evolution. But conjunctive definitions face their own difficulties. They tend to exclude edge cases that biologists intuitively want to include, like the earliest proto-organisms that presumably preceded the emergence of full Darwinian evolution. And they become increasingly ad hoc, patching each counterexample with an additional clause rather than illuminating any underlying principle.

What's philosophically significant is the pattern of failure itself. When decades of rigorous effort by biologists, chemists, and philosophers fail to produce a definition that withstands scrutiny, we should entertain the possibility that the target concept doesn't have the logical structure we've been assuming. Perhaps life isn't the kind of thing that admits of necessary and sufficient conditions—and perhaps that tells us something important about the natural world, not just about our conceptual limitations.

Takeaway

When every attempt to define a concept through necessary and sufficient conditions fails in systematic ways, the failure may reveal the structure of the concept itself rather than the inadequacy of our efforts.

The philosophical alternative that best accommodates this persistent definitional failure is the notion of a cluster concept—sometimes called a homeostatic property cluster, following Richard Boyd's influential analysis. On this view, life is not defined by a single essence but by a constellation of properties that tend to co-occur: metabolism, reproduction, heredity, responsiveness to environment, cellular organization, growth, homeostasis, evolution. No single property is strictly necessary, and no specific combination is sufficient. What makes something alive is exhibiting enough of these properties in the right kind of causal configuration.

The analogy to Wittgenstein's treatment of game is instructive but imprecise. Games share family resemblances without a common essence. But life's cluster properties aren't merely accidentally correlated—they are causally interrelated. Metabolism supports reproduction. Heredity enables evolution. Evolution shapes metabolic pathways. This causal cohesion is what Boyd's homeostatic property cluster account emphasizes: the properties cluster together because underlying causal mechanisms tend to produce them as a package. The cluster is held together by nature, not by our classificatory convenience.

This framework handles the problem cases with notable elegance. Viruses exhibit heredity and evolution but lack autonomous metabolism—they sit near the boundary of the cluster, which is exactly where intuitions become contested. Fire exhibits thermodynamic dissipation but lacks heredity and cellular organization—it falls clearly outside. Mules exhibit metabolism, cellular organization, homeostasis, and growth, but not reproduction—they're comfortably inside despite lacking one property. The cluster concept maps onto our actual biological judgments far better than any essentialist definition.

Critically, the cluster concept approach entails that life admits of borderline cases as a matter of metaphysics, not just epistemology. It's not that we don't yet know whether viruses are alive and will figure it out with more data. It's that 'alive' may not be a fully determinate predicate at those boundaries. This is a substantive philosophical claim about the structure of biological reality—that the living and nonliving shade into each other, and that the transition from prebiotic chemistry to biology was not a sharp threshold but a gradual accumulation of the properties in the cluster.

For philosophers of science, this has important implications for the broader debate between natural kinds realism and conventionalism. Life, on this analysis, is a real pattern in nature—the causal cohesion of the cluster is objective—without being a classical natural kind with sharp necessary and sufficient membership conditions. It occupies a middle ground that Daniel Dennett might recognize: real enough to ground scientific investigation, but resistant to the crisply bounded categories that essentialist metaphysics demands.

Takeaway

If life is a homeostatic property cluster rather than a classical natural kind, then borderline cases like viruses aren't puzzles to be solved—they're exactly what the structure of the concept predicts.

This isn't an exercise in academic taxonomy. The definition of life has direct consequences for how we design instruments, interpret data, and allocate scientific resources. Consider NASA's astrobiology program. Every biosignature detection protocol embeds assumptions about what life is. If we define life metabolically, we design instruments to detect chemical disequilibria. If we define it evolutionarily, we look for molecular structures capable of hereditary variation. The definition determines the search space, and a wrong or overly narrow definition could mean that we detect extraterrestrial life and fail to recognize it—or fail to look in the right places entirely.

The cluster concept approach suggests a different strategy: rather than searching for a single definitive biosignature, astrobiology should pursue multiple independent indicators corresponding to different properties in the cluster. Chemical disequilibrium plus complex molecular organization plus evidence of hereditary variation would collectively provide stronger evidence than any single criterion. This is, in fact, how the most sophisticated astrobiological thinking already proceeds—but it often does so without explicit philosophical justification. The cluster framework provides that justification and makes the reasoning transparent.

The stakes are equally concrete in artificial life research. When Thomas Ray's Tierra or Karl Sims's evolved virtual creatures exhibit open-ended evolution, metabolism-analogs, and reproduction, have they created life? A classical definition forces a binary answer that may distort assessment. The cluster concept allows a graded evaluation: these systems instantiate some properties in the life cluster while lacking others, and our assessment should reflect that granularity rather than forcing a yes-or-no verdict.

There is also a deeper methodological lesson here. The search for a definition of life often presupposes that definition must precede investigation—that we need to know what life is before we can study it scientifically. But the history of science suggests the opposite. Chemistry didn't wait for a satisfactory definition of 'element' before making extraordinary progress. The concept was refined through scientific practice, not prior to it. Similarly, the concept of life may be one that science progressively articulates rather than one that philosophy fixes in advance.

This reversal of the traditional order—from define, then investigate to investigate, then refine the concept—is itself a philosophical commitment. It reflects a naturalistic stance that takes science and philosophy as continuous enterprises, each informing and disciplining the other. The problem of defining life, far from being a preliminary question we need to settle before real science begins, is one of the most productive sites where scientific detail and philosophical analysis genuinely need each other.

Takeaway

The most consequential philosophical questions aren't always the ones we answer—sometimes they're the ones we learn to ask differently, letting scientific practice reshape the concepts that guide it.

The persistent failure to define life through necessary and sufficient conditions is not a scandal for biology or philosophy. It is an informative result—one that tells us something substantive about the topography of the natural world. Life, understood as a homeostatic property cluster, is a real and causally grounded pattern that nonetheless resists the sharp boundaries classical definition demands.

This has consequences that extend well beyond the seminar room. It reshapes how we design missions to search for extraterrestrial biology, how we evaluate claims in artificial life research, and how we understand the relationship between conceptual analysis and empirical investigation. The question what is life? doesn't need a final answer—it needs a better framework for being asked.

Perhaps the deepest lesson is that some of the most important concepts in science are precisely the ones that resist tidy definition—and that this resistance, properly understood, is not an obstacle to knowledge but a guide toward it.