The Right Direction | Amitoj Singh

When tin becomes scarce somewhere in the world, prices rise. Shippers adjust their routes without knowing why. Manufacturers find substitutes. Consumers reduce consumption. All of this happens without anyone needing to know why tin became scarce — the price signal carries the information that matters. Friedrich Hayek called this a "marvel" and meant it.

I find the achievement genuine. Billions of people with different goals, different knowledge, and different values manage to coordinate their activities through price signals alone. No central planner could aggregate the dispersed knowledge that markets process every second. The shipper doesn't need to know about the mine collapse in Malaysia or the new battery technology in Korea; she only needs to know that tin prices rose.

Hayek deserves the credit he gets for this insight. The coordination problem he identified is real, the mechanism he described works, and the alternatives — centralized planning, bureaucratic allocation — consistently fail where markets succeed. The empirical track record is clear enough that I don't think this is ideological.

But Hayek used a phrase when describing this system that contains the problem his analysis doesn't resolve. He said price signals tell participants "the right action" and "the right direction." What determines the right direction? At the level of any individual transaction, the answer looks clean: prices do. Move resources toward higher-value uses. Zoom out, though, and a question appears that price signals can't answer on their own. What makes one direction "right" in the first place? Against what standard is the price system optimizing?

The question here is directional: which problems deserve the mechanism's attention, and who decides.

Markets operate within frameworks they cannot generate. Property rights determine who can participate in exchange and what can be exchanged. Contract enforcement determines which agreements will be honored. Legal definitions determine which entities count as actors capable of making binding commitments. These frameworks don't emerge from trading — they make trading possible.

Consider what markets presuppose rather than produce. Most societies prohibit certain exchanges: you can't buy votes, sell children, or purchase organs in most jurisdictions. These boundaries around legitimate exchange aren't discovered through price signals. They're settled through political and social processes that stand outside the market mechanism entirely. The market can coordinate activity within the boundaries of legitimate exchange. Where those boundaries should be is a different kind of problem.

This matters because it reveals something about the relationship between coordination and objectives. Before you can coordinate, you need to agree on what you're coordinating toward. The question "how do we win?" assumes someone already answered "what game are we playing?" — and that prior question, what counts as success, which values take priority when they conflict, is a different kind of problem than the coordination problem markets solve. I want to be careful not to overstate this: markets are indispensable within whatever boundaries get set. The issue is about the boundaries, not the mechanism that operates inside them.

Take carbon pricing. We can now measure greenhouse gas emissions with remarkable precision. The measurement problem is largely solved. You might think this means we can design a market mechanism to allocate emissions efficiently and let prices do the work.

Designing that mechanism requires value commitments that measurement alone cannot generate. Should emissions rights be allocated on a per-capita basis, giving equal atmospheric rights to each person on earth? Should historical emissions count, acknowledging that industrialized nations filled most of the atmosphere's carbon capacity during their development? Should developing nations receive greater present-day allowances as compensation for foreclosed growth paths?

These are questions about how competing values should be balanced when they conflict — questions about what matters. The price mechanism operates downstream of these commitments. Better measurement sharpens the value questions that need settling. Each increment of precision makes the underlying choice more visible, and more urgent.

As our ability to measure things improves, value questions multiply.

When measurement was expensive, organizations bundled together two different kinds of problems. "Follow protocol" compressed "we can't verify outcomes" with "we haven't settled which outcomes matter." The first is an information problem; the second is a values problem. They look similar in practice because both produce the same organizational response: defer to procedure rather than assess results directly.

As measurement improves, this bundle fractures. The information problem becomes tractable. We can now measure hospital outcomes, educational achievement, environmental impact, organizational performance — all with unprecedented precision. That precision doesn't settle which measurements matter. We can measure patient mortality rates, infection rates, satisfaction scores, cost efficiency, staff wellbeing. Measuring them all with perfect precision doesn't tell us how to weight them when they conflict. I find that greater measurement precision sharpens these disagreements. The facts become undeniable, and the argument shifts entirely to what the facts are supposed to justify.

Epistemic progress and evaluative progress operate on different axes. Confirming that something is the case leaves open whether it matters and how much it matters relative to other things that are also the case. Each resolved factual question tends to branch into multiple value questions. Confirming that climate change is real opens questions about historical responsibility, generational equity, development rights, species preservation. The problem space expands where the fact space contracts.

There's a further complication. Coordination mechanisms shape the preferences they aggregate.

When university rankings become visible, institutions start optimizing for the ranked dimensions: faculty-student ratios, citation counts, graduation rates. Each institution that does this rises; each one that doesn't falls. Within a decade, the rankings increasingly measure how well institutions optimize for rankings. No one is acting irrationally — each institution is responding sensibly to the incentive structure in front of it. The distortion is structural, not motivational.

The same logic applies wherever actors optimize toward signals rather than underlying values: financial markets where asset prices become reference points that actors trade toward, social media where engagement metrics shape content, organizations where performance dashboards define what "good performance" means.

Once something becomes measurable, it tends to become salient, then legitimate, then binding — each step individually unremarkable, the aggregate effect constitutive. The coordination mechanisms meant to aggregate preferences end up shaping the preferences they aggregate, through the attention they focus and the legitimacy they confer.

The implication follows: allowing coordination mechanisms to operate without explicit governance of objectives answers the question of which objectives matter by default — through the constitutive power of whatever becomes visible and measurable.

If you don't settle what you're optimizing for, the optimization process will settle it for you based on what's easy to see, what's easy to count, what produces signals that can be aggregated. That outcome systematically favors dimensions that can be compressed into metrics over dimensions that resist compression: short-term over long-term, individual over collective, measurable over meaningful.

The question of objectives runs through the entire coordination process. It's the central strategic question. Performance presupposes objectives: "How well?" requires prior answer to "Toward what?"

Settlement here means the processes — political, negotiated, authoritative, conventional — that produce commitment even where disagreement persists. Settlement is active: someone, or some institution, commits to treating certain values as binding for purposes of coordination. The answer must be created through commitment.

I'm not confident this distinction is always clean. There are situations where factual discoveries genuinely narrow the evaluative question. Learning that a policy causes harm we didn't anticipate can shift what counts as acceptable. But the narrowing happens because the new facts engage with existing values — we already cared about that kind of harm — and the facts alone don't settle the priority question.

What strikes me as genuinely problematic is forgetting that the framework was chosen rather than discovered. When we treat institutional frameworks — property rights, commodification boundaries, performance metrics — as natural rather than constructed, we lose the ability to question whether the game we're optimizing is the game we should be playing.

Markets are good at solving problems. They can't decide which problems are worth solving. Price signals tell you how to win the game currently being played. Which objectives matter? Against which standard? Who decided, and how?

There is no algorithm for determining what the algorithm should maximize.