Feb 5, 2008

nassim nicholas taleb, at SALT

nassim nicholas taleb was the latest speaker in stewart brand's series of free public seminars about long-term thinking (SALT). the last lectures i attended were by vernor vinge (on the coming singularity) and frans lanting (on photographing time past in the present). taleb is the author of the black swan and fooled by randomness, two books which i've tried and failed to read through (like the hobbit, no doubt, which took more than four attempts). stewart brand always releases an excellent summary of the talk on the SALT blog, so there's really no point in me reinventing the wheel that brand will probably make today. though his talk was nominally about how the future has always been crazier than we thought, it was really an extended disquisition on the limits of knowledge both retrospectively and prospectively. what i thought was most apropos was his argument that the modelling sciences are generally in the business of inferring future states based on observed past states. this can be a flawed exercise because

  1. historical data is almost always incomplete (records are ephemeral, recordkeeping is idiosyncratic, and coverage cannot be universal)
  2. any series of data can be explained by an infinite number of non-linear models
in consequence, particularly of the second point above, to be effectively predictive, quantitative models are best applied to non-complex systems. the interesting result of this reasoning is that it becomes a strong motivation for reintroducing qualitative analysis as a viable mode of understanding the kinds of complex systems that which occupy the researches of most social scientists: societies, cultures, and how they change. intelligent qualitative analysis is a mode thinking which acknowledges the effect of unobservables and idiosyncrasy rather than falling back on presumed certainty derived from highly-imperfect data.

in a sense, taleb's entire talk was about knowing the enemy and knowing the self (知己知彼,百戰不貽). pervasive rationality leads us to believe that we can extend the nature of causal connection from simple to complex systems (taleb would call it extending learnings from mediocristan to extremistan) when in fact this extension is fallacious. we need to have the ability to confront complex systems without requiring a highly-structured and deterministic analytic framework, by which i mean quantitative modeling and analysis of almost all varieties. i bring up keats's concept of negative capability and a.r. ammons's corsons inlet* repeatedly in situations like this, precisely because this is the ability and approach that they both advocate.

as a race, we're now in a position where the rate of growth in our power to change the world has far exceeded the rate of growth in our knowledge of how to use that power. as we increase our influence over complex systems like global climate patterns or ecozones, our basic action orientation has to include negative capability so that, as taleb puts it, we can understand why our default position in regard to complex systems should be a hyperconservative one. (ie, getting us both individually and collectively to a point where our actions and decision-making methods are appropriate to the systems our actions affect.)

now, getting off the soapbox, another question which this brings up is whether or not complex systems (like societies and other large groups) are ever susceptible to analysis by analogical comparison to self-similar -- but smaller and more bounded -- groups. i think the jury is still out on that one.

* both introduced to me by glenn adelson as part of bio95hfz: conservation, nature, and biodiversity -- sadly no longer taught at harvard, but available to the lucky women of wellesley in modified form as es242: war and environment.

No comments: