As human beings we have a unique ability to recognize patterns, even when confronted events that are completely random. In fact sometimes it's easier to see patterns in random noise. We pull narratives out of the randomness and use them to predict the future. Unfortunately the future is unpredictable and even when we have detected a pattern the outcomes end up being very different than what we expected.
The US-backed regime in Afghanistan lasted 9 days from the taking of the first provincial capital to the taking of Kabul. After the withdrawal of the Soviets in 1989. That government lasted over three years. What was the difference? Why after spending two trillion dollars and twice as long in the country did we do so much worse? Francis Fukuyama has asserted that the are no ideologies which can compete with liberal democracy. And everyone seems to believe that, but if that's so why did we have such a hard time implanting it in Afghanistan?
Philip Tetlock has been arguing for awhile that experts are horrible at prediction, but that his superforecasters do much better. If that's the case how did they do with respect to the fall of Afghanistan? As far as I can tell they didn't make any predictions on how long the Afghanistan government would last. Or they did make predictions and they were just as wrong as everyone else and they've buried them. In light of this I thought it was time to revisit the limitations and distortions inherent in Tetlock's superforecasting project.