Skip to content

Practical Rationality and the Long Run

Standard accounts of practical rationality treat each decision as the locus of evaluation. An agent faces a choice; we evaluate her choice; we say whether she chose rationally. The unit of evaluation is the decision.

I want to suggest, building on a point I have made before, that the unit of evaluation is sometimes the pattern of decisions, not the individual decision. There are cases where every decision in the pattern can be locally rational and the pattern as a whole irrational, and the inverse: every decision can be locally suboptimal and the pattern overall rational.

Local rationality without global rationality

The first case is familiar from the literature on procrastination and weakness of will. The agent decides, on Monday, that working today is a small cost and the benefits of getting started can be deferred to tomorrow. The same calculation on Tuesday produces the same conclusion. The pattern extends across the week. Each individual decision, taken in isolation, appears rational: small cost today, larger benefit tomorrow. The pattern produces an outcome the agent does not endorse.

The standard analyses of this case appeal to time-preference inconsistency, hyperbolic discounting, or weakness of will. These analyses identify what has gone wrong at the level of the agent’s preferences. What they do not, I think, fully capture is the sense in which the irrationality is a property of the pattern rather than of any individual decision. Each decision, taken on its own, is defensible. The defense fails when we step back to see the pattern.

Global rationality without local rationality

The second case is less well-discussed. An agent makes a series of decisions, each of which is locally suboptimal, but which together constitute a rational policy.

Consider an agent committed to giving 10% of her income to charity. Each individual donation, considered locally, is suboptimal — she could have spent the money on something with more immediate utility for herself. The pattern is rational because the policy of giving is. The agent who breaks the policy in a single case is not acting more rationally than the agent who keeps it, even though her individual decision in that case might be locally rationalizable.

Policies of this kind are pervasive. Saving for retirement is a policy of locally suboptimal decisions. So is exercising. So is dieting. So is keeping promises in cases where breaking them would be locally easier.

Two senses of rationality

If both cases are real, we have two senses of rationality at work: a local sense, attached to individual decisions, and a global sense, attached to patterns. The two can come apart in either direction. Which is the “real” rationality?

I do not think we should answer this question. Both senses do real work, and both are appropriate in different contexts. When we evaluate a single choice, the local sense is the right tool. When we evaluate a life or a project, the global sense is. The mistake is to think one sense is more fundamental than the other, or to think that the local sense, scaled up to a series of decisions, automatically gives us the global sense.

What this implies

If we accept that practical rationality has both a local and a global dimension, several things follow.

First, accounts of rationality that focus exclusively on individual decisions are incomplete. They miss the dimension of evaluation that applies to patterns. This is, I think, a significant blind spot in the dominant decision-theoretic frameworks.

Second, the conditions under which an agent can be held rational vary with the level of analysis. An agent might be locally rational on every decision and globally irrational; we should hold her rational at the first level and ask, separately, whether her pattern shows the kind of incoherence that calls for revision.

Third, and most practically, the project of becoming more rational is not exhausted by improving individual decisions. It also requires attention to patterns — to whether the sum of one’s decisions, across time, constitutes the life one would on reflection endorse. This is an old thought, but I think it has not been adequately integrated into formal accounts of practical reason.

I welcome objections from those who think the local-global distinction collapses on reflection. The most promising version of that objection, I take it, is that what looks like irrationality at the level of patterns is really preference inconsistency at the level of decisions. I have my doubts about this reduction, but I think the burden is on me to show that the reduction fails. That is work for another post.

More from this kitchen

Reasons and Rationality

A Note on Decisive Reasons in Normative Ethics

The term “decisive reason” is used in two distinct senses in the contemporary literature, and the conflation between them is, I think,…

NatashaDecember 19, 2025

Add your two pesos

Your email address will not be published. Required fields are marked *