The Hedgehog and the Fox

Philosopher and political theorist Isaiah Berlin wrote a 1951 essay called The Hedgehog and Fox, based on ancient Greek poet Archilochus: "the fox knows many things but a hedgehog one important thing." Berlin used this to analyze Tolstoy's War and Peace, essentially asserting Napoleon was a hedgehog (only invading Russia was important) while Russian General Kutuzov was a fox (and of course won). Berlin believed Tolstoy was a fox wanting to be a hedgehog. He also classified people from history as one or the other: Plato a hedgehog and Aristotle a fox, for example. Many authors used the hedgehog/fox dichotomy for a variety of purposes.

John Lewis Gaddis used the hedgehog/fox analogy in On Grand Strategy (2018) beginning with Xerxes's (a hedgehog) invasion of Greece in 480 BC, with advisers (foxes) counseling restraint because of the many unforeseen circumstances. One of the most interested comparisons was St. Augustine (saint and hedgehog) and Machiavelli (sinner and fox); dogged persistence versus eclectic cunning. When it comes to statecraft, realism is key, not idealism: single mindedness versus restraint, analysis and updating new information. Machiavelli believed politics was the art of seizing and holding power (this was Renaissance Italy after all). The hedgehog alternative is formulaic solutions to ill-defined problems.

Philip Tetlock's influential Expert Political Judgment found that dichotomy most useful for evaluating the ability of experts to make predictions and judgments on a variety of events (over 28,000 by 284 experts from the years 1988-2003). Granted, experts have more interest in explanations rather than predictions; they also are okay with predictions as long as no one keeps score. He found that, on average, experts had the same forecasting ability as a dart throwing chimpanzee (that is, they did no better than chance). He discovered some interesting results. Not surprising, short-term forecasts are more accurate than longer term predictions. The characteristic that really stood is hedgehogs predictions derived from grand schemes of theory were relatively poor (note that fits into my description of the blinders hypothesis), while fox predictions based on analysis of diverse sources of information were statistically better. Foxes were particularly better at longer-term forecasts. Characteristics included self-criticism, an open mind, and willingness to reject preconceptions. Presumably this results in what Nassim Taleb (Black Swan guy) referred to as "mediocristan" versus the potential "extremistan" of hedgehogs.

The key perspective is psychologists' interest in the psychology of forecasting by 1) measuring differences in performance by posing questions that are resolvable in a reasonable time frame and 2) learning by giving precise feedback. Part of the analysis was Bayesian ability to update viewpoints. That is, the change forecasts based on new information (the probability of a recessions increases with economic shocks such as a trade war). It turns out that people, including experts, are not good Bayesians, but foxes are still better than hedgehogs. It turns out it is more important how people (experts) think than what they think; Tetlock calls this multi-method triangulation using a full range of methods. For example, foxes are more likely to change their minds, based on new information. This suggests that hedgehogs suffer from cognitive dissonance (putting up belief-system defenses; not to mention closed-mindedness and hubris). There is the open question of whether social science should search for predictive laws. Four key obstacles to forecasting: 1) preference for simplicity, 2) aversion to ambiguity and dissonance, 3) need to believe we live in an orderly world, and 4) ignorance of the laws of chance (p. 37). We prefer deterministic thinking rather than probabilistic strategies (order in random sequences). "Foxes do better because they are moderates who factor conflicting considerations--in a flexible, weighted-averaging fashion--into final judgments; "hedgehog opinion was in greater demand from the media" (p. 118).

Interesting comments: History is one damned thing after another, a random walk with upward and downward blips without continuity (p. 19); but possibly a random walk with an upward trend (p. 33). Algorithms will beat experts in open systems (p. 26). One interesting question was the rise of the West out of the Middle Ages. Apparently, there were small advantages for growth and co-evolving institutions that encouraged rule of law (especially property rights), tolerance for free inquiry (and therefore a common pool of knowledge), and a competitive state system (p.28-9). Somewhat related are tipping point models. But is history just an extraordinary string of coincidences? The NTSB analyzes plane crashes in detail and can explain specific causes, but cannot predict the future; what they can suggest is not to repeat discovered mistakes (p. 35). Experts have both cognitive biases (pervasive misunderstandings) and motivational biases (p. 43). Bayesian calculations (p. 122), egocentricity gap or cognitive conservatives, but better for foxes (experts are not natural Bayesians). Hindsight bias, more pronounced in hedgehogs (they think they got something right when in fat they got it wrong, p. 138). Hedgehogs more top-down, deductive arguments, foxes to bottom-up inductive arguments (p. 147). Hedgehogs as prisoner of their preconceptions (p. 163); foxes have a more balanced style of thinking about the world (p. 118). Hedgehogs fit my blinders hypothesis, although I suspect virtually all of use have blinders based on culture, income levels, profession, and so on. Psychological and personal traits can also exacerbate the hedgehog/fox dichotomy.

[Note that Tetlock spends considerable time describing his methods of testing predictions in a game setting, such as value, controversy, difficulty, and fuzzy-set adjustments. His basic measures, beginning with probability scoring, are in the Technical Appendix, starting on p. 273).]