Sunday 7 June 2009

Gerd Gigerenzer on 'rules of thumb'

Orthodox economics asks the question 'What is rational decision-making?' It recommends allocating scarce resources among our competing interests according to how much we value each interest and their probability of being brought about by our choices. By multiplying these together we should get a single ordering of our choices in terms of the subjective value of the outcome that will be brought about, and then we should choose the highest ranked option.

Of course that's not very realistic - humans can't manage to put precise values on our interests or calculate probability accurately. But experiments also revealed that even in simple situations humans deviated significantly from the orthodox theory. Behavioural economics therefore appeared in the economic firmament from the 1980s to ask a quite different question: 'Why do humans make irrational decisions?' That is, they were concerned to identify the systematic cognitive biases inherent in human nature that prevented us from achieving the economic ideal of rationality. If one could map those biases, then one could correct for them and full rationality could be achieved!

Gerd Gigerenzer takes a different approach. He suggests that economists are too concerned with constructing theories for 'small worlds' (with few choices, clear causal mechanisms, easily calculated probabilities - the kind that exist only in theories or experimental set-ups). But the real world is open and pretty chaotic, and there is no reason to presume that a finely crafted decision theory that works in a small world will be at all helpful for making good decisions in the real world. So his question is 'How do humans make decisions?' And his answer is that we employ heuristics ('rules of thumb') which are very simple and only consider part of the information available. His argument is that less can often be more because the real world is so uncertain (radically unpredictable) that basing decisions on all the information available is not only technically infeasible but may also lead to worse predictions than merely focussing on a tiny set of salient information.

Gigerenzer's new book 'Gut Feelings: The Intelligence of the Unconscious' is peppered with well-rehearsed and fascinating anecdotes supporting this role for heuristics. To name just one, Harry Markowitz won a Nobel prize for proving a very complex theory of optimal investment portfolio management, yet when he made his own retirement investments he used the simple 1/n rule of equal investment in each kind of asset. (Incidentally this simple 1/n rule not only saves on computing time, but also outperforms more complex information-greedy theories in the real world.)

So i) heuristics are not a second best alternative in the real world, ii) they are how we think (since evolution has stuck us with them in any case). But Gigerenzer's heuristics remain somewhat mysterious. They operate at the level of behaviour, reasoning and perception. They can be unconscious reflexes, conscious reflexes (available to introspection), or deliberately chosen (perhaps with the aid of higher level heuristics?). But heuristics can only be considered more or less 'ecologically rational' i.e. effective for certain situations. When they go wrong - with racism for example - or conflict, there seems to be some scope for deliberately evaluating, updating and choosing heuristics in the light of their overall efficacy or other concerns. But Gigerenzer's antipathy to the power of reasoning makes him unable or reluctant to explain how this would work.