Reducing uncertainty is costly, and sometimes costs freedom
The last part of the book is quoted less often. Yet it deserves attention: Knight reminds us that reducing uncertainty is not desirable regardless of cost. What is gained in predictability has to be weighed against what is sacrificed to obtain it.
One might imagine that a rational economy always seeks to reduce uncertainty as far as possible. Knight rejects this conclusion, and the argument is worth following through to the end, because it applies as much to a project as to an economy.
An optimisation problem, not a moral one
Increasing knowledge requires research, data, time and resources. Strengthening control over the future calls for procedures, reserves, institutions and sometimes very heavy investment. Every reduction in uncertainty therefore has an opportunity cost: what could have been done with the same resources.
This is a classic optimisation problem, with an equilibrium point. The first feasibility study removes a great deal of uncertainty at modest cost. The third independent review of the same subject removes very little at a comparable cost. The marginal return on information declines, while its cost remains linear. There is therefore a point beyond which knowing more destroys value, and that point comes sooner than one might think.
The political cost of predictability
Knight adds a cost of a different kind, discussed more rarely. Consolidating decisions within large organisations reduces uncertainty by concentrating authority and coordinating more activities. But this concentration also reduces the freedom of those who delegate their choices. The same mechanism that improves predictability alters the distribution of power.
The reasoning applies at every scale. Within an organisation, the person who accepts a fixed salary trades uncertainty for security, but also trades autonomy for subordination. This is neither good nor bad, it is a trade-off, and Knight simply asks that it be seen as such rather than presented as progress without a downside.
Finally, he puts forward a deliberately provocative idea: slowing progress would remove a large share of uncertainty, since change is its main source. But the price would be obvious: less innovation, less experimentation, less improvement. The goal therefore cannot be "zero uncertainty". We have to choose how much uncertainty we accept in order to obtain other benefits.
Project governance as a case in point
This logic is directly useful in running projects. More control, reporting and sign-off reduces certain errors. But beyond a certain point, the system consumes time, slows decisions and kills precisely the adaptability it was meant to protect. Perfect compliance with an old plan is not a success when a new world has arrived in the meantime.
The symptom is easy to recognise: a project whose steering bodies produce more documents than decisions, where every deviation triggers a justification rather than an adjustment, and where the time spent demonstrating control exceeds the time spent on delivery. The uncertainty reduction system has then become the main source of risk.
The right question is therefore not: how do we eliminate uncertainty? It is: which uncertainty deserves to be reduced, at what cost, and what share has to be accepted because eliminating it would destroy too much value? It is a question of proportion, not of purity.
In an age when the abundance of data sometimes gives the impression that one more dashboard will solve metaphysics, Knight remains useful. He reminds us that information is a resource, that control is an organisation, and that predictability is a good that has a price. Even the obsession with certainty has to go through a business case.
Frank H. Knight, Risk, Uncertainty, and Profit (1921), chapters XI and XII. Editorial adaptation, not a quotation.