Imagine, for a moment, that Newton had approached gravity not as a description but as a prescription. Objects ought to attract each other in proportion to their masses. Planets should orbit in ellipses. Apples are morally obligated to fall downward. This sounds absurd, and it is — because we understand physics as a descriptive enterprise. We observe regularities, formalize them as laws, and then continuously refine those laws as new observations come in. Newton's gravity was refined by Einstein's, which will presumably be refined by whatever comes next. The "laws" of physics are not commands. They are our best current descriptions of how things actually behave.

Now consider ethics. We have moral "laws" — do not murder, do not steal, treat others as you wish to be treated. But we treat these not as descriptions to be refined but as commandments to be obeyed. They come from somewhere — God, reason, social contract, evolutionary psychology — and our job is to follow them, not to study them. What if this framing is exactly backwards? What if moral regularities are more like physical regularities: patterns we observe, formalize, and continuously refine?

Prescriptive vs. Descriptive Ethics

The dominant ethical traditions are all fundamentally prescriptive. Deontology says: here are the rules, follow them. Consequentialism says: here is the metric (utility, welfare, happiness), maximize it. Virtue ethics says: here is the kind of person you should be, become them. These frameworks disagree on content but agree on structure — ethics is about what you should do, and the philosopher's job is to determine what that is.

A descriptive approach to ethics would look radically different. Instead of asking "what should people do?", it would ask "what do people do when they are at their best?" Instead of deriving moral laws from first principles, it would observe moral regularities in human behavior — across cultures, across centuries, across contexts — and formalize them as provisional descriptions, subject to revision.

We do not criticize Newton for being "wrong" about gravity. His description was extraordinarily useful for centuries and remains so for most practical purposes. It was refined, not discarded. Why should we treat Aristotle's ethics, or Kant's, any differently? Not as eternal truths to be defended, but as useful approximations to be improved.

This is not moral relativism. Relativism says there are no moral facts — just preferences that vary by culture. Descriptive ethics says there are moral facts, in the same sense that there are physical facts, but that our access to them is empirical, provisional, and subject to error. The moral law is real. Our formulation of it is approximate.

Formalizing Moral Observation

If ethics is a descriptive science, we need tools for formalization. Utility functions are a natural starting point. Define a function $U: S \to \mathbb{R}$ that maps states of the world to real-valued measures of human flourishing. The "moral law" at any given time is our best estimate of what kinds of actions tend to increase $U$ across agents and over time.

The difficulty, of course, is that $U$ is not directly observable. We infer it from behavior, testimony, and outcomes. This is exactly the situation in physics — energy is not directly observable either. We infer it from measurements and use it because it makes the equations work. The fact that $U$ is inferred rather than measured does not make it less real. It makes it harder to study, which is precisely why moral knowledge advances more slowly than physical knowledge.

Game theory offers a more precise lens. Consider two agents choosing actions, each with a utility function. A Nash equilibrium is a pair of strategies $(s_1^*, s_2^*)$ such that:

Neither agent can improve their outcome by unilaterally changing strategy. This is a descriptive result — it tells you where rational agents will end up, not where they should be. But notice something interesting: many Nash equilibria correspond to what we intuitively recognize as moral outcomes. Cooperation in repeated games, truth-telling in markets, reciprocal altruism in communities — these emerge as equilibria, not as commandments.

The Golden Rule is not a divine command. It is a Nash equilibrium in the repeated game of social life. Treat others as you wish to be treated, because in the long run, defectors are excluded and cooperators thrive. The morality is descriptive — it describes what actually works — and it is no less binding for being so.

The Comparison

It helps to see the two approaches side by side:

The "error handling" row is particularly important. In prescriptive ethics, moral disagreement is a crisis — someone must be wrong, and the stakes are ultimate (sin, damnation, moral failure). In descriptive ethics, moral disagreement is data. If two cultures arrive at different moral conclusions, that's an observation to be explained, not a battle to be won. Perhaps one is closer to the truth. Perhaps both capture different aspects of a complex regularity. Perhaps the disagreement reveals a boundary condition where our current moral "laws" break down — analogous to the way Newtonian mechanics breaks down at relativistic speeds.

The Learning Morality

The deepest implication of this framing is that moral knowledge is cumulative. Just as physics has progressed from Aristotle to Newton to Einstein to quantum field theory, moral understanding can progress — not by discarding previous insights but by refining and generalizing them. The prohibition against murder is not wrong. It is a low-resolution description of a more complex principle about the conditions under which human flourishing requires the protection of life. As our understanding deepens, the description becomes more precise.

This means ethics is a learning enterprise. We are not trying to perfectly implement a known set of rules. We are trying to discover the rules — the deep regularities that govern human flourishing — through the same iterative process of observation, hypothesis, test, and revision that has been so spectacularly successful in the physical sciences.

The expected value of adopting a descriptive stance can be framed as an optimization problem. Let $\theta$ represent our current moral framework and $\theta^*$ represent the true moral regularities (which we may never fully know). The loss function is:

where $U_i$ is the flourishing of agent $i$. A prescriptive approach fixes $\theta$ and minimizes compliance error. A descriptive approach minimizes $L$ directly by continuously updating $\theta$ toward $\theta^*$. The prescriptive approach is only optimal if you already have $\theta = \theta^*$ — if your current moral framework is already perfect. The descriptive approach is optimal in all other cases, which is to say, always.

I am not proposing that we abandon moral conviction. I am proposing that we hold moral convictions the way a good scientist holds physical theories: with confidence proportional to evidence, with openness to revision, and with the understanding that being approximately right and getting more right over time is better than claiming to be exactly right and never updating. The moral law is not less real for being discovered rather than decreed. Gravity did not become less binding when we stopped thinking of it as God's will and started thinking of it as the curvature of spacetime. If anything, our description became more useful, more precise, and more beautiful. I suspect the same will be true of ethics, if we have the courage to look.