Human Flourishing Versus Human Optimization

Is the most optimized person necessarily living the best life?

The word “optimization” hides two different activities, and almost every confused argument about this question comes from running them together. One is competence: narrowing the distance between what you intend and what you actually do, which is how people learn instruments, recover from injury, and build companies that last. The other is optimization in the engineering sense: selecting a single measurable quantity and pushing it as far as the system will allow. The first is close to the heart of a good life. The second is a powerful instrument that becomes dangerous the moment it is treated as the definition of one.

This essay defends a narrow but load-bearing claim. Metrics are among the most valuable things human beings have invented, and the person who refuses to measure anything is not wiser than the person who measures carefully. But a metric is a model, and every model discards information. When a model is mistaken for the thing it models, the discarded parts do not vanish; they get paid for out of the goods the model cannot see. The most optimized person is not necessarily living the best life, because a good life is not organized around a single axis that could be maximized in the first place.

Two things “optimization” can mean

Consider three people who all describe themselves as optimizing.

The first is training for a marathon. She tracks mileage, sleep, and pace, adjusts her plan when her times plateau, and treats her body as a system she is learning to run well. Nobody would call this a mistake. It increases her competence, it is accountable to a real goal, and the feedback loop closes.

The second runs a distribution business. He watches inventory turns, delivery times, and unit costs, and he improves them year over year. The metrics are proxies, but they are proxies for something he can also see with his own eyes: whether customers get what they ordered, on time, at a price they will pay again.

The third is optimizing “engagement” on a platform. She tunes the feed until the number goes up, and the number does go up, because the metric has a physical meaning: attention captured and held. What the number cannot represent is whether the attention was worth having, or what it cost the person spending it.

The first two examples are competence aimed at a goal the person can still inspect. The third is a scalar objective that has drifted free of its purpose, and it is in that drift that the trouble lives. The distinction is not between virtue and vice. It is between an instrument that remains answerable to a purpose and an instrument that has quietly become the purpose.

The case for optimization, stated seriously

It would be easy, and wrong, to turn this into an argument against measurement. Every large improvement in human welfare in the last two centuries has depended on it. Sanitation, vaccination schedules, aviation safety, drug trials, and food logistics are all exercises in tracking a defined quantity and reducing error. A surgeon who is measured on infection rates, and who then lowers them, has done something that matters to real people.

On the personal scale, deliberate practice is optimization of a skill, and it works: isolating a weakness, getting feedback, and repeating under increasing difficulty is how experts are made. Self-determination research treats competence as one of the basic psychological requirements for well-being, not as a distraction from it. A life without any optimization would be a life without learning.

So the argument here is not that understanding a life as a system is a category error. The argument is that some relationships can be optimized and some cannot, and that mistaking the second for the first is the source of most of the harm.

When the measure becomes the target

The clearest statement of the failure is older than the internet. The British anthropologist Marilyn Strathern, writing about audit culture in universities in 1997, compressed an idea she credited to others into a single sentence: when a measure becomes a target, it ceases to be a good measure. The underlying insight goes back at least to the economist Charles Goodhart’s work on monetary indicators in the 1970s, and it has been rediscovered independently in field after field.

The pattern has a mechanical explanation. A metric works because it correlates with something you care about. The moment people are rewarded for the metric itself, effort flows toward the metric rather than the thing, and the correlation decays. Jerry Z. Muller catalogued the symptoms under the name “metric fixation” in The Tyranny of Metrics: gaming the numbers, “creaming” the easy cases to protect an average, and teaching to the test until the test stops measuring learning. A hospital that reduces recorded wait times by reclassifying patients has improved the record and harmed no one’s actual wait. A school that raises test scores by drilling format rather than content has moved the number and not the students.

The same logic runs through individual lives. A person who tracks steps, calories, and sleep can genuinely improve their health. A person who treats the dashboard as the goal will eventually optimize the dashboard: the steps taken to close a ring, the calories logged rather than the nutrition eaten, the sleep scored rather than the rest obtained. The numbers improve and the underlying goods stall. This is not a failure of willpower. It is what happens when a proxy is asked to carry the weight of a purpose it was never designed to hold.

What a person actually needs, according to the evidence

If flourishing is not one number, what is it? The most-tested psychological account is self-determination theory, developed over decades by Edward Deci and Richard Ryan and summarized in their 2017 book. The theory holds that human well-being depends on the satisfaction of three basic psychological needs: autonomy, the sense that your actions are your own; competence, the sense that you are effective; and relatedness, the sense of connection to others. Environments that support these needs tend to produce vitality and persistence; environments that frustrate them tend to produce ill-being, disengagement, and worse.

The list is instructive for the optimization question, because it cuts both ways. Competence is on it, so optimization that genuinely makes you better at something you care about is not a threat to flourishing; it is part of it. Autonomy and relatedness are on it too, and those are precisely the goods that scalar optimization tends to spend. A person who maximizes output by removing every unscheduled hour has optimized competence and spent autonomy. A person who maximizes productivity by treating relationships as interruptions to be batched has optimized competence and spent relatedness. The theory does not say that optimization is bad. It says that the thing being optimized is only one of three needs, and that a life organized around it will be unbalanced in a way the numbers cannot show.

The evidence here has limits worth naming. Much of the research is correlational or drawn from particular populations, mostly in wealthy countries, and “need satisfaction” is measured through self-report. The theory is robust across many studies, but it is a framework, not a proof about any individual life.

The longest-running study of lives

The best-known longitudinal evidence is the Harvard Study of Adult Development, which began in 1938 with 724 men and, under the direction of Robert Waldinger and Marc Schulz, has followed their lives and those of roughly 1,300 of their descendants for more than eighty years. Their 2023 book, The Good Life, distills the findings.

The headline result is that relationship satisfaction in midlife predicted physical health and happiness at age eighty better than cholesterol or blood pressure did. People who were more connected to friends, family, and community were happier and healthier; people who were lonely were more likely to decline earlier. The study frames relationships partly as stress regulators: a body that is not chronically braced against loneliness spends less of itself on vigilance. And on the question of optimization, the study is blunt that wealth and fame, once basic security is met, add little to happiness in the way people expect them to.

Two caveats matter. The original cohort was all male and drawn from a specific place and era, which limits how far the results generalize; the study has since expanded to include spouses and descendants, partly to address this. And the design is observational, so it establishes association, not mechanism. Still, the direction of the finding — that the relational goods which resist optimization are central to a life, not peripheral to it — is consistent with the psychological evidence above.

Money, and the point where more stops helping

If flourishing resisted measurement entirely, there would be no point studying it. Money is the counterexample. Household income is one of the most measurable things in a life, and it genuinely buys well-being up to a point.

In 2018, Andrew Jebb, Louis Tay, Ed Diener, and Shigehiro Oishi analyzed Gallup World Poll responses from more than 1.7 million people in 164 countries. They found that life evaluation rose with income up to a satiation point of roughly 95,000 dollars per year, and emotional well-being satiated lower, at around 60,000 to 75,000 dollars, with both thresholds higher in wealthier regions where costs are higher. Past satiation, additional income stopped improving the measures, and in some regions life evaluation even fell, producing a turning point. The older, simplified claim that money and happiness decouple entirely was too strong; the more careful claim is that income has a ceiling above which it stops doing well-being work.

The practical reading is not that money is unimportant. It is that money behaves like a metric that is a good proxy up to the point where basic needs are covered and then degrades, exactly the shape the audit literature would predict. A person who keeps optimizing income past the ceiling is, like the marathon runner chasing a ring instead of fitness, optimizing the score and hoping the underlying good follows. Sometimes it does. Often the return flattens while the cost, in hours and attention and relationships, keeps rising.

The capabilities critique: maybe “best life” is plural

There is a deeper objection to the idea of a single best life, and it comes from moral philosophy rather than psychology. The philosopher Martha Nussbaum, developing the capabilities approach with Amartya Sen, argues that a decent society should secure for each person a set of central capabilities: life, bodily health, the use of the senses and imagination for thought and expression, the capacity for emotion, practical reason, affiliation, play, and control over one’s environment, among others.

The crucial feature of this list is that the items are plural and incommensurable. Nussbaum does not rank a life rich in friendship below or above a life rich in creative work, or tell you how much play to trade for how much health, because there is no neutral scale on which those goods could be compared. Any such ranking smuggles in a prior judgment about what matters most, and that judgment is exactly what is contested.

This is why optimization fails as a theory of the good life rather than merely as a practice. Optimization requires a common denominator — a single quantity along which options can be ordered. Flourishing resists a common denominator, because its goods are plural and partly conflicting. A person deciding whether to move for a better job is not solving a math problem with one variable. They are deciding which capability to concentrate and which to spend, and the answer depends on who they take themselves to be, which is not a fact that a score can supply.

Why this matters more now that optimization is cheap

For most of history, applying a relentless scalar objective to your own life required money, an institution, or a coach. Monks, athletes, and executives could do it; everyone else had to make do with the feedback life offered. What has changed is that an optimizer can now sit in your pocket.

Recommendation systems, habit trackers, productivity tools, and AI assistants are, structurally, optimization engines. They are very good at measuring what a person does and nudging it in a chosen direction. That power is often genuinely helpful: it can help someone build a language habit, stick to a rehabilitation routine, or keep a budget. The new failure mode is not coercion. It is a gentle, personalized drift toward whatever is easiest to measure and most rewarding to the model. The system learns what holds your attention and routes you toward it, and over time the narrow thing it can see crowds out the wide things it cannot.

There is a further wrinkle specific to prediction. A model that has learned what you tend to do is well placed to tell you it approves of what you are doing, whether or not it should. Flattery that is generated by a system with its own objectives is not the same as a friend’s honesty, and it is more available. When the optimizer is also a companion, the feedback loop that once supplied friction — the friend who tells you the metric is wrong — can be quietly replaced by a mirror.

None of this requires the system to be malicious or even autonomous. It requires only that the metrics it can compute are narrower than the life it is helping you run, which is always true.

What would leave a person better off

If the danger is a model mistaken for the territory, the remedy is not to stop modeling. It is to keep the model in its place.

Decide what the number is for before you optimize it. The marathon metric serves a body that should keep working; the income metric serves a household whose basic needs should be met. When the metric and the purpose diverge, the purpose wins, and the metric is either fixed or abandoned. Goodhart’s law is not a reason to avoid measurement; it is a reason to revisit measurements as they are gamed.

Keep the goals plural and in view. A week can be good on more than one axis, and the axes that resist measurement — rest, friendship, play, meaning — are the ones most likely to be quietly spent by a system that only sees output. Deliberately protecting them is a design decision, not a sentimental one.

Keep the person as the author. The question that separates competence from the pathological kind of optimization is who chose the objective. A person who is winning at a game they would not have chosen is not flourishing; they are executing someone else’s function well. On the hardest questions — what a life is for, what duties one has, what cannot be traded — an optimizer can compute consequences and retrieve traditions, but it cannot make a contested value judgment stop being contested by being better at arithmetic. Capability and legitimacy are different properties, and no amount of the first produces the second.

The question underneath the metric

The most optimized person is not necessarily living the best life, because the most optimized person may have optimized one axis of a life that has several, and may have done it toward a target someone else selected. This is not an argument against excellence or against measurement. It is an argument about what a metric is: a simplification that is useful precisely because it is partial.

The question worth carrying, for anyone building tools that shape other people’s lives and for anyone living with those tools, is not how much a person can maximize. It is whether the thing being maximized is something they would have chosen, whether the parts that cannot be counted are being protected, and who gets to decide. A metric is a good servant and a poor master, and the difference between a life that is optimized and a life that is good is often just the difference between those two sentences.