Why the Therapy/Enhancement Distinction Should Still Ground Gene Editing Policy

How do we know when a person is dead? For most of human history, the line between life and death was generally considered obvious. If a person was breathing and their heart was beating, they were alive. Then came the invention of the mechanical ventilator. Suddenly, a person with no brain function could have their heart and lungs sustained indefinitely by a machine, allowing them to present the traditional signs of life, while being, in a sense, no longer present. Hospitals filled with cases of patients to whom the old definition of death could not be applied, leaving doctors, families, and courts to decide whether such patients were alive, dead, or something in between.

The solution was not a sharper definition. It was an institutional change. The 1968 Harvard Brain Death Committee proposed criteria for irreversible coma, the Uniform Determination of Death Act formalized brain death as a legal standard in 1981, and hospital ethics committees emerged to handle the cases that statutes could not resolve in advance. On the surface, this might look like a definitional shift, from cessation of breathing to the cessation of brain function. But brain death itself remained contested at its edges, with ongoing disagreements over whole-brain vs. higher-brain criteria and a steady supply of cases that no definition could fully resolve. What kept the concept of death operative was not new criteria alone but the institutional infrastructure built around them: hospital protocols, neurological examination standards, ethics committees, and statutes that delegated judgment to specialists. The concept of death did not become any cleaner, but rather, it became administrable. Instead of changing the metaphysics around death, the medical community adopted procedural infrastructure capable of handling conceptual ambiguity.

A Recurring Pattern

This pattern is not unique to death. Whenever biotechnology destabilizes a morally weighty distinction, the same response tends to follow: not a return to definitional purity, but the construction of institutions that can do regulatory work despite the fuzziness of the underlying concept. Gene editing is the latest instance in this pattern, with the distinction between therapy and enhancement being put under pressure. Therapy refers to interventions that restore or preserve normal function in response to disease or impairment; enhancement refers to interventions that extend capacities beyond the baseline of normal function. This is the working distinction used across the medical and regulatory community, from the FDA to the World Health Organization to national bioethics councils. CRISPR has made it possible to edit the human genome with precision never seen before, meaning that the difference between editing to treat disease and editing to improve a person is not a metaphysical thought experiment but an active policy concern. Critics argue that the therapy/enhancement line is too conceptually unstable to ground policy and should be replaced by frameworks built around consent, risk, and distributive justice.

The therapy/enhancement distinction should remain foundational to gene editing policy, but on revised grounds. The line is not a claim about a sharp natural division between two kinds of intervention. It is a policy scaffold that coordinates medicine, regulation, and public expectation around a shared sense of what medicine is for. The proposed replacements, examined on their own terms, each fail to do the regulatory work the distinction currently performs. This distinction also captures two moral concerns that no alternative framework can replicate: that future persons cannot consent to being enhanced and that an unregulated enhancement market would deepen existing social inequalities. The hard cases at the edges of this distinction are real, but procedurally traceable in the same way that the hard cases at the edges of life and death turned out to be traceable once the right institutions were built. The question is not whether the therapy/enhancement distinction is perfect. It is whether anything better is available. Nothing is, and imperfect indispensability is a respectable basis for policy.

The Distinction as Policy Scaffold

Most attacks on the therapy/enhancement distinction target a version of it that no serious regulator would use. Critics assume that the distinction should track some intuitive natural boundary and then point to cases where the boundary blurs and conclude that the distinction is incoherent. This argument is only successful against a metaphysical interpretation of the distinction. However, this metaphysical interpretation is not the one that does the regulatory work. Rather, the distinction functions as a policy scaffold. Its job is to coordinate a network of actors, including physicians, regulators, insurers, patients, and researchers, around a shared sense of what medicine is for. Erik Parens describes this distinction as a working concept used to clarify what medicine is for, not a metaphysical definition (Parens S1). The line marks where the internal purpose of medicine ends and consumer preference begins, and that boundary is morally significant even when its precise location is contested.

The content of this distinction comes from the idea that medicine has an internal purpose, that being the restoration of function and the relief of suffering, which enhancement does not share. This commitment has a long history in the philosophy of medicine. Christopher Boorse's naturalist account defines disease as deviation from species-type function, anchoring the concept of therapy to a biological baseline (Boorse 542). Norman Daniels develops a more nuanced version, arguing that normal function is partly shaped by social and historical context but arguing that it remains useful as a policy coordinating standard (Daniels, ch. 2). The distinction between these different definitions is not relevant to the argument. Rather, it only requires that some account of normal function be available to ground policy, and the fact that the term "normal function" remains consistent across every healthcare system on earth suggests that it is.

Critics might argue that the actual distinction used in policy is more metaphysically ambiguous than what is being defended here. However, sophisticated regulatory bodies would suggest otherwise. The National Academy of Sciences, in its 2017 report on human genome editing, explicitly treats the therapy/enhancement distinction as a policy concept rather than a natural kind, acknowledging its fuzziness while affirming its regulatory utility (National Academies of Sciences, ch. 6). The metaphysical reading of this definition is more of an esoteric philosophical critique rather than one that regulators actually defend. Therefore, using this distinction to ground policy is not a retreat to a weaker version, but rather a restoration of a version that's been used all along.

What counts as therapy and what counts as enhancements is not fixed. Vaccines were once treated as enhancements and are now standard care. Cosmetic surgery was once a medical specialty that has now turned largely consumer-oriented. Prescriptions for ADHD treatments flip back and forth across the line depending on diagnostic conventions and prescribing norms. Critics argue that the variability of distinction means it cannot stably ground policy.

However, a distinction that moves with medical practice, social consensus, and accumulated evidence, that is ever evolving and prone to change, is not failing but rather functioning correctly.

A strictly metaphysical distinction would remain fixed across all these changes. It's precisely the variability of this distinction that allows the same regulatory framework to handle vaccines in the 1900s and CRISPR in 2025 without requiring a new framework every time the medicine advances.

Why the Alternatives Fail

The three most commonly proposed alternative frameworks are risk, consent, and distributive justice-based frameworks. Each addresses a unique issue that arises within gene editing—the risk of potential mishaps or unintended consequences in the actual editing process, the fact that, by the nature of gene editing, those who are affected by it cannot directly consent to it, and the huge social and class ramifications that gene editing would have, specifically regarding who has access to it. But each of these issues is fundamental to gene editing and necessary to address; they are not the core factor to center the policy framework around.

The risk-based framework is centered around the idea that gene editing regulation should look at whether a procedure's benefits outweigh its risks, and that the morality surrounding therapy for enhancement is irrelevant in this decision. Savulescu's "procreative beneficence" argument is the clearest version of this view: parents have reason to select for any trait that improves expected welfare, regardless of whether the change counts as treatment or improvement (Savulescu 415).

It doesn't take much searching to find cases where this framework fails. Consider two interventions, both safe and reliable. One corrects a mutation to prevent sickle cell disease. The other raises a child's height by four inches. Under this framework, both of these operations would be treated equally since the risk profiles are identical. However, most people would agree that these cases feel intuitively different from a moral standpoint. This framework just doesn't make sense because it doesn't capture the real tension people feel around gene editing. While risk is a crucial factor, the discourse surrounding gene editing regulation is more focused on what will happen when the risk is minimal, if not eliminated.

The second alternative raises the issue of consent. It's built around the idea that interventions made with the explicit consent of the patient are permissible, and ones made without consent are not. This framework works for somatic gene editing in adults but collapses for germline editing. The affected person, in this case the future child, cannot consent to any intervention because, first, they have not been born yet, and second, the intervention will determine who they become and what kind of decisions they make. The solution is to develop a way to assume consent by asking what a reasonable future person would consent to. Jürgen Habermas develops the most rigorous version of the consent framework for germline cases and arrives at a striking conclusion (Habermas 63). The future person can be presumed to consent to disease prevention, because almost any rational agent would want their disease prevented. The future person cannot be presumed to consent to enhancement, because enhancement reflects parental preferences rather than the affected person's interests. Notice that using this definition, the therapy/enhancement distinction has snuck its way back in as the criteria for presumed consent. The consent framework actually depends on the therapy/enhancement distinction, meaning it can't replace it.

The third alternative would ground policy in a distributive justice framework. Proponents of this framework would argue that gene editing should be regulated according to the advantages and disadvantages it gives to people. Rawlsian frameworks supply the most developed version, with the difference principle requiring that inequalities benefit the least advantaged (Rawls, sec. 13). However, contradictions arise when applying a Rawlsian framework to gene editing. One might argue that genetic enhancement should be banned because it gives advantages to children of only the wealthy. A Rawlsian could equally argue that genetic enhancement should only be given to disadvantaged children, on the grounds that genetic intervention is one of the few mechanisms capable of correcting inherited inequality at its source. Colin Farrelly has shown how distributive principles applied to genetics generate exactly these kinds of contradictory recommendations, depending on which inequality is treated as primary and which counterfactual is taken as baseline (Farrelly 46). Distributive justice is essential to thinking about gene editing, but it cannot ground policy on its own because it does not specify which inequality matters most. The framework is too permissive in one direction and too restrictive in the other, and nothing inside the framework chooses between them.

Each of these alternatives captures one dimension of what the therapy/enhancement distinction holds together but becomes too restrictive or too permissive once it is asked to do the regulatory work alone. Risk-benefit captures the welfare dimension but loses the distinction between need and preference. Consent captures the autonomy dimension but loses traction in exactly the germline cases that matter most. Distributive justice captures the equality dimension but generates contradictory recommendations depending on framing. The therapy/enhancement distinction does not replace these considerations. It coordinates them. It is the scaffold that lets welfare, autonomy, and equality each do their proper work without any of them having to bear the full regulatory weight.

But coordination is not the only thing the distinction does, and the failure of the alternatives is not the only reason to keep it. Each alternative breaks on the same intuition: that there is a moral difference between preventing sickle cell disease and adding four inches to a child's height, even when risk, consent, and distributive effect are held constant. The frameworks could not absorb this intuition because the intuition is not reducible to any one of them.

When Intuition Outweighs Argument

There is a similar phenomenon in ethics. A moral framework can be coherent, consistent, and validly argued from reasonable premises and still get rejected because of where it lands. Take the surgeon who could save five dying patients by killing one healthy visitor and harvesting the organs. A pure cost-benefit calculation says to do it: five lives outweigh one. Almost no one accepts this, and the refusal is not treated as a failure to follow the argument. It is treated as a verdict on the argument.

When a clean theory and a firm intuition collide, the intuition is the evidence and the theory is what gives way. When a principle yields a monstrous result, it shows that something is wrong with the principle, not with us.

The therapy/enhancement distinction is precisely an intuition of this nature. The alternatives to it are all coherent frameworks. It's the difference we feel between curing a disease and choosing a trait that keeps colliding with them. Therefore, the case for the distinction rests on a different question: when is an intuition worth trusting, and is this one?

Testing the Intuition

To conclude whether an intuition is worth trusting, we have to examine what sits underneath the feeling of wrongness. Let's start by understanding why exactly enhancement feels wrong under the three alternative frameworks, starting with distributive justice. Picture an enhancement market that is legal and priced like any other elective procedure. Families who can afford it will buy height or stamina or IQ points or whatever traits the labor market at the time rewards. The families who can't won't. The issue here is not so much that we are creating people who are biologically "superior" to others, but rather that the advantages already concentrated by wealth will gain a new way to compound. The results would be devastating, especially in a society that rewards inherited resources and whose institutions already fail to close the gap, a dynamic that Buchanan, Brock, Daniels, and Wikler argue tends to widen existing inequalities (Buchanan et al. 16). However, in such a society, therapy would also be unevenly distributed, not just enhancement. While true, this is a different type of injustice. The uneven distribution of therapy is the uneven distribution of a good that everyone is presumed to want or need. The uneven distribution of enhancement is the uneven distribution of skills or traits that are only valuable because others do not have them.

The line here is between correcting disadvantages and creating new advantages. No one would argue against the implementation of therapeutic or restorative procedures simply because they are unevenly distributed. We wouldn't stop all vaccinations or the use of antibiotics because there are some people who can't afford them. Instead, we would continue to use them while striving to make them accessible to all. The same cannot be said for enhancements. Making an enhancement available to everyone doesn't fix the problem the way making a vaccine available to everyone does. A vaccine protects you whether or not your neighbor has one. An enhancement doesn't work that way. If everyone can afford the extra height or the extra IQ points, the advantage cancels out, and all we're left with is the cost and risk of the procedure with nothing to show for it. If only some can afford it, the advantage stays tied to wealth. Wider access either erases the benefit or deepens the gap, which is why uneven distribution counts against enhancement in a way it never counts against therapy. That is the first thing sitting underneath the intuition: not a worry about who receives the good, but a worry about a good whose entire value depends on others going without. The real tension here is not the distribution of gene edits but the distribution of gene enhancements, and therefore, the real consideration is in the therapy/enhancement distinction and not in any sort of distributive justice.

The same realization emerges when we move from a society to a singular person. Consider a couple editing an embryo. In one version, they remove a mutation in order to prevent sickle cell disease. In another, they choose to add 5 inches of height or select an even temperament for their child. In both cases, the child can't consent to the procedure. However, the two choices are still asymmetrical. A person who will one day live in that body, when asked beforehand, would almost certainly agree to the prevention of a painful and life-threatening disease. The same cannot be assumed for any sort of modified characteristic, such as height or temperament. These changes don't correct any harm, but rather reflect the personal desires and expectations of the parents. Jürgen Habermas makes this point about freedom: A person whose traits were fixed by someone else's preferences is no longer the sole author of their own life or a full equal among people who were not designed (Habermas 79). This is the moral intuition stated precisely. It is also the piece that the consent framework was missing. The issues around gene editing don't arise from a lack of consent but a lack of consent regarding enhancement.

A further concern, the hardest of the three to state, lives not in fairness or consent but in the relationship between a parent and a child. Michael Sandel argues that genetic enhancement erodes the disposition to receive a child as they come rather than as a project specified in advance (Sandel, ch. 5). Stated as a claim about life being a gift, this invites an obvious reply, and the reply is fair: parents shape their children all the time through tutoring, music lessons, orthodontics, and careful nutrition, and none of that is thought to corrode parental love. But there is a difference between shaping a child and specifying one. For example, imagine a parent who wants their child to be a musician. Ordinarily, they would put their child in music lessons and see whether they take a liking to it, all while having to accept that they might not and turn out to be a swimmer instead. In this case, the pathway or aptitude for music was offered, and the child had the choice to either accept it or choose something else. However, editing a child to give them a natural proclivity for music is different because the inclination is installed rather than discovered. Lessons act on a person who already exists and can push back, while enhancement sets the terms before there is anyone to push back. This will not satisfy a critic who finds the difference empty. But notice where it lands. Curing a disease lifts an obstacle and returns a child toward a baseline no one would refuse; specifying a trait imposes a particular picture of who the child should be. A worry that owes nothing to fairness or consent, that comes only from what it means to be a parent, picks out the same boundary the other two did.

Each of these cases individually arrives at the same conclusion, that conclusion being that therapy is permissible and enhancement is not. However, this only tells us what the line is, not where it should be drawn.

The Sorites Problem

While this intuition has survived three independent tests, it has only been tested at the extremes. Sickle cell versus four inches of height are clean ends of the spectrum, and real-life cases are rarely so straightforward. Somewhere between preventing harmful diseases and customizing children lies a series of subtle intervals that slowly morph therapy into enhancement, with the worry being that no principled line can be drawn across it. This is arguably the most serious objection to the therapy/enhancement distinction.

Begin with the correction of a single mutation that would otherwise kill a child during infancy. Clear therapy. Next, correct a mutation for a disease that is serious but survivable. Still therapy. Now, correct a variant that does cause a disease but raises the risk of one. Correct a variation that only raises the risk slightly. Strengthen the immune system so that it fights infections more reliably than a natural one. Edit in resistance to a disease that the child had no special risk of inheriting, as was done by Chinese scientist He Jiankui in 2018, when he edited embryos to be resistant to HIV. Raise baseline lung capacity. While you're at it, raise stamina. Maybe some more muscle, and with the muscle, add height. Suddenly, we have arrived at enhancement. That means somewhere along that chain, the edits suddenly stopped being therapeutic. The question is where.

This scenario is a famous logical puzzle in philosophy known as the Sorites paradox. Imagine a heap of sand. Remove one grain of sand from the heap. Is it still a heap? Obviously, as the heap hasn't changed in any meaningful way. Remove another and repeat. Eventually, there will be one grain of sand that is not a heap. Therefore, it seems to follow that along the way, the removal of one grain of sand changed the heap into something else, maybe a pile or a handful or a collection. This puzzle illustrates what kind of complications arise when we apply broad terms to the real world. The endpoints remain clear and distinct while the middle becomes an uncertain gradient, which is challenging to define. A critic would point to this and argue that a distinction with no non-arbitrary boundary, the argument goes, cannot ground policy, because every place the law might draw the line is as defensible as the place a centimeter to either side, and a boundary that is arbitrary cannot bear the moral weight of deciding what may be done to a future person.

The argument is sound against a target the distinction was never supposed to be. It assumes that to ground policy, the line must mark a sharp natural boundary—that somewhere in the genome or the gradient there must be a real seam where therapy ends and enhancement begins and that if no such seam exists, the distinction is empty. But the line was never a claim about where nature divides. It is a policy scaffold, and policy scaffolds are not in the business of carving nature at its joints.

How Fuzzy Concepts Do Real Work

If the line need not carve nature, the question becomes how a line that carves nothing can still decide anything. The answer is visible wherever a fuzzy concept already does serious work, and two fields show it most plainly.

The law is built almost entirely out of concepts with no sharp edges. H. L. A. Hart described this as the open texture of legal language: every general rule has a settled core and a surrounding number of penumbras, which are not explicitly included or excluded from the rule. A rule banning vehicles from a park explicitly bans cars from driving in there. It says nothing about bicycles, skateboards, or electric scooters. This is not a careless oversight but rather a fundamental part governing unpredictable worlds using general rules (Hart, ch. 7). The law does not respond to these instances by sharpening definitions. Rather, it responds procedurally using a court to determine the outcome case by case and setting that decision as a fixed point for the next one.

Negligence is the clearest example. Liability is determined by whether the defendant exercised reasonable care, and reasonable care has no definition that can universally be applied in every court. In Vaughan v. Menlove, the court held that the standard is not what the defendant himself believed prudent but what a person of ordinary prudence would have done, and that whether that standard was met is a question for the jury. This allows for the problem to be solved procedurally. Jurors apply the standard to the facts given to them and render a verdict by verdict. There's no universal standard for what it means to be a reasonable person, yet no one has concluded that negligence is incoherent as a concept. The concept works because we have institutions designed to apply it, not because it is sharp.

Medicine faces a similar problem, but in a more literal form. Most variables that matter to health vary by a subtle degree, meaning that there is no clean natural line between normal and harmful. Instead, the line gets drawn in a similar way as it does in the law. A body of experts fixes the line, publishes criteria for it, and reevaluates it based on new evidence. In 2017, the American College of Cardiology and the American Heart Association lowered the threshold for hypertension from 140/90 to 130/80, reclassifying about thirty-one million Americans as hypertensive without one of them changing (Whelton et al.). A reading of 131 is not meaningfully sicker than a reading of 129, so the cut is arbitrary at the margin in exactly the way the critic warns about. It governs the treatment of tens of millions anyway, because a usable line drawn by a credible process is worth more than a perfect line that does not exist.

What Gene Editing Requires

This pattern exists in both law and medicine, and it's what gene editing requires as well. The work that the therapy/enhancement distinction does will not be made better by a sharper definition. Rather, like law and medicine, it requires institutions and procedures that can draw the line case by case. The therapy/enhancement distinction has the concept without the right institutions yet. The embryo being edited against the disease it was never going to face is the same as a ventilator patient. They are both cases where no definition can settle them in advance, but rather where a standing body could adjudicate, weighing each intervention against the internal purpose of medicine the line was built to track.

The line will not always sit still. What we count as therapy will migrate slowly as technology progresses. Vaccination, once an enhancement of the body's natural defenses, has now become standard care. An edit we may consider genetic enhancement today could be standard care in fifty years. This migration is not the distinction failing, but rather it is demonstrating its ability to absorb new capability without creating a framework every time. It's in the decision-making around this line migration that the proposed alternatives, risk, consent, and justice become meaningful considerations. With therapy and enhancement as the fundamental line, these are the factors we turn to in determining where that line falls. The therapy/enhancement distinction is the structure that lets all three of these inform decision-making and policy without having to let one bear the weight alone.

This is why nothing can replace the distinction. Every alternative would inherit its fuzziness without the structure it provides. The case for the distinction was never that it is clean, fixed, or metaphysically deep, as it is arguably none of those. It is that the distinction is the one available concept that captures the moral asymmetry between curing and choosing, coordinates the considerations that are relevant to the hard cases, and inherits the ability to change over time.

The only aspect we're missing is the procedures built around it, procedures that would be created by institutions we've developed before and can develop again. The imperfect indispensability that is the distinction between therapy and enhancement is a respectable basis for the policy coordination around genetic editing.

Works Cited

Parens, Erik. "Is Better Always Good? The Enhancement Project." The Hastings Center Report, vol. 28, no. 1, 1998, pp. S1–S17.

Boorse, Christopher. "Health as a Theoretical Concept." Philosophy of Science, vol. 44, no. 4, 1977, pp. 542–73.

Daniels, Norman. Just Health: Meeting Health Needs Fairly. Cambridge University Press, 2008. Ch. 2, "What Is the Special Moral Importance of Health?"

National Academies of Sciences, Engineering, and Medicine. Human Genome Editing: Science, Ethics, and Governance. National Academies Press, 2017. Ch. 6, "Enhancement."

Savulescu, Julian. "Procreative Beneficence: Why We Should Select the Best Children." Bioethics, vol. 15, no. 5–6, 2001, pp. 413–26.

Habermas, Jürgen. The Future of Human Nature. Translated by Hella Beister, Max Pensky, and William Rehg, Polity Press, 2003.

Rawls, John. A Theory of Justice. Harvard University Press, 1971. Ch. 2, "The Principles of Justice."

Farrelly, Colin. "Genetic Justice Must Track Genetic Complexity." Cambridge Quarterly of Healthcare Ethics, vol. 17, no. 1, 2008, pp. 45–53.

Buchanan, Allen, et al. From Chance to Choice: Genetics and Justice. Cambridge University Press, 2000.

Sandel, Michael J. The Case Against Perfection: Ethics in the Age of Genetic Engineering. Belknap Press of Harvard University Press, 2007. Ch. 5, "Mastery and Gift."

Hart, H. L. A. The Concept of Law. Clarendon Press, 1961. Ch. 7, "Formalism and Rule-Scepticism."

Vaughan v. Menlove. 132 Eng. Rep. 490. Court of Common Pleas, 1837.

Whelton, Paul K., et al. "2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA Guideline for the Prevention, Detection, Evaluation, and Management of High Blood Pressure in Adults." Journal of the American College of Cardiology, vol. 71, no. 19, 2018, pp. e127–e248.