What I've Been Reading

The Ambiguities of Experience by James G. March. Brilliant book exposing the limits of experience, a must read. But let me be clear - this book is not a confirmation bias for people who refuse to learn from their own experiences, leave alone learning from wisdom of others. Thanks to FarnamStreet for recommending this book and for these brilliant notes.

Folk wisdom both trumpets the significance of experience and warns of its inadequacies. On the one hand, experience is described as the best teacher. On the other hand, experience is described as the teacher of fools, of those unable or unwilling to learn from accumulated knowledge or the teaching of experts. The disagreement between the folk aphorisms reflects profound questions about the human pursuit of intelligence through learning from experience that have long confronted philosophers and social scientists.


The ambiguity of experience has many causes and takes many forms, but a significant fraction of them can be summarized in terms of five attributes of experience:

• First, the causal structure of experience is complex. Many uncontrolled variables are involved, and their relations include multiple interactions and multiple colinearities. The relations among the variables may include numerous instances of feedback loops, a variety of time delays, and unknown functional forms. As a result, it is difficult to uncover the causal structure and to identify the effects of actions.
• Second, experience is noisy. The events of history are drawn from a distribution of possibilities, either because of errors in observation or interpretation or because the causal structure is truly stochastic. A particular realized history is likely to be a quite poor representation of the possibilities. As a result, learning from experience involves trying to learn not only from the actual events observed but also from the events that did not occur but might quite easily have occurred. The generation of such hypothetical histories replaces evidence with imagination, with all the invitations to error that such a substitution involves.
• Third, history includes numerous examples of endogeneity, cases in which the properties of the world are affected by actions adapting to it.History is a series of samples, but the sampling rates, and therefore the sampling errors, of the various alternatives are affected by the unfolding of experience. 
• Fourth, history as it is known is constructed by participants and observers. Individuals learn not from history but from historical stories, including the stories they tell themselves, that are concocted for a purpose. The proposition that the mean of lies converges to truth as the sample size increases is not easily demonstrated, either empirically or deductively. 
• Fifth, history is miserly in providing experience. It offers only small samples and thus large sampling error in the inferences formed.

“History will justify anything. It teaches precisely nothing, for it contains everything and furnishes examples of everything”

- French poet-philosopher-historian Paul Valéry