Via Overcoming Bias, je tombe sur cet intéressant article à rajouter dans la catégorie « la science et le conservatisme » (voir ici, là, ou là ou encore ici). L’article défend notamment l’idée que les institutions encadrant la recherche scientifique tendent à favoriser une recherche « d’exploitation » plutôt qu’une recherche « d’exploration » :
« A key problem, suggests mathematical physicist Eric Weinstein of the Natron Group, a hedge fund in New York, is that it is too easy for scientists in the “establishment” of any field to cut down new ideas, and to do so without really putting anything at risk, thereby leading to a culture that is systematically biased toward caution. “High-risk science is much more associated with figures from the past,” he says.
The result, he suggests, is that science is becoming less a “bottom-up” enterprise of free-wheeling exploration — energized by the kind of thinking that led Einstein to relativity — and more a “top-down” process strongly constrained by social conformity, with scientific funding following along fashionable lines. The publish-or-perish ethic, in particular, strongly rewards those scientists doing more or less routine technical work in established fields, and punishes more risky work exploring unproven ideas that may take a considerable period of time to reach maturity.
This is especially damaging given the disproportionate benefits that come from the most important discoveries, which seem to be inherently unpredictable in both timing and nature. As Taleb argues persuasively in The Black Swan, any sensible long-term strategy in a world dominated by extreme and unpredictable events has to accept, and even embrace, that unpredictability. He illustrates the idea in the financial context. People investing in venture-capital start-ups, for example, have to expect continual losses in the short term, and bank on the fact that they will ultimately make up for those losses by hitting on a few really big winners in the long run« .
Au passage, on peut noter que cet argument est à rapprocher de l’idée de diversification/décentralisation de la recherche scientifique comme stratégie évolutionnairement efficace que j’avais développé ici. L’article aborde par ailleurs un autre point particulièrement intéressant souligné par Robin Hanson sur Overcoming Bias, à savoir l’idée d’instaurer un équivalent des « prediction markets » au niveau scientifique, ce qui reviendrait à instituer un marché des « futures » des idées scientifiques :
« Weinstein suggests another idea — that we should borrow some ideas from financial engineering and make scientists back up their criticisms by taking real financial risks. You think that some new theory is utterly worthless and deserving of ridicule? In the world Weinstein envisions, you could not trash the research in an anonymous review, but would buy some sort of option giving you a financial stake in its scientific future, an instrument that would pay off if, as you expect, the work slides noiselessly into obscurity. The money would come from the theory’s proponents, who would similarly benefit if it pans out into the next big thing.
Weinstein’s point is that markets, in theory at least, work efficiently and — putting the current financial meltdown to one side — lead to the accurate valuation of products. They exploit the “wisdom of crowds”, as a popular book of the same title recently put it. Take the famous electronic prediction markets at the University of Iowa, which pool the views of thousands of diverse individuals and consistently seem to give better predictions than any expert. For example, they predicted last year’s US presidential election correctly to within half a percentage point.
Could the same not be done for weighing up the likely value of scientific ideas? Those ideas, Weinstein argues, do not get weighed fairly today. As he points out, mavericks get their papers routinely rejected for what they feel are unfair reasons, and they often feel suppressed by the mainstream community, while mainstream scientists think it is perfectly obvious that the ideas in question are ludicrous and should not waste the community’s time. Current research practice lacks any mechanism that would arrange a fruitful meeting between the two — letting the maverick’s ideas gain free expression while at the same time letting the critics take a real stake based on their own knowledge
Bringing such possibilities into play, Weinstein suggests, would move research practice closer to the “efficient frontier” — the place where ideas get judged fairly based on all available knowledge, and risk takers, rather than being suppressed by social conformity, get encouraged by those taking a financial stake in the potentially enormous consequences of their success. Such mechanisms, Weinstein argues, would help avoid the effective censorship that often afflicts peer review, and that currently keeps research on the cautious side of the efficient frontier« .
On remarquera qu’une pratique commence à se développer chez les blogueurs américains, celle de conclure des paris sur l’évolution future de certains indicateurs comme le chômage. Voir par exemple le pari conclu entre Bryan Caplan et John Quiggin. Ce genre de pratique est finalement une forme d’ébauche d’un prediction markets des idées scientifiques. Alors, qui est prêt à prendre le pari avec moi que cette nouvelle forme d’évaluation des idées scientifiques va bientôt se développer* ?
* note : je plaisante !.