The Science of Decision-Making Under Radical Uncertainty · Ludwig
Frank Knight's distinction between risk and uncertainty, published over a century ago, remains one of the most important ideas in economics. Risk can be quantified; uncertainty cannot. Yet our entire apparatus of decision-making is built for a world of risk, not uncertainty.
The Limits of Probabilistic Thinking
Consider a simple question: what is the probability that artificial general intelligence will be developed in the next decade? Any answer you give reveals more about your assumptions than about reality. Unlike a coin flip, where we understand the underlying physics, AGI represents a category of problem where probability itself becomes ill-defined.
Yet we routinely apply probabilistic frameworks to such questions. Superforecasters assign credences to geopolitical events. Actuaries build models of catastrophic risks. These exercises can be valuable, but only if we understand their limitations.
The future is not only unknown but unknowable. The best we can do is prepare for a wide range of possibilities while remaining alert to signals that our assumptions are wrong.
Building Robust Strategies
Robustness differs from optimization in several crucial ways:
Optionality over optimization. Rather than finding the single best strategy, robust decision-making favors strategies that preserve flexibility.
Resilience over efficiency. Systems optimized for a single objective tend to be fragile to unexpected shocks.
Adaptability over prediction. Rather than investing heavily in forecasting, robust strategies invest in the capacity to detect and respond to change.
Applications to Investment
The traditional approach of estimating expected returns and covariances to construct optimal portfolios is fundamentally flawed when applied to long-term decisions. The inputs to such models are not stable parameters but moving targets subject to regime change.
In a world of radical uncertainty, humility is not just a virtue, it's a competitive advantage.