Saturday, March 5, 2011

Biases in Pastcasting

One of my most vivid memories of college is a warning I received during an introductory course in history (which ended up being my minor): Don't impose the judgments of the present on the past. We are all biased by our personal experience, which is shaped by a variety of influences, most of which we aren't even conscious of; this can be a great impediment to the historian's job of finding and accurately describing how people in earlier times were living their lives.

I found myself remembering this warning as I pondered the apparent disconnect between my projections of low happiness in the past with evidence of more happiness reported by small, isolated communities in the present. As I speculated in “Practical Revolution,” this was likely due to the isolation of those communities. Even the countries in the data, none of which is individually isolated, have considerable variation around the statistical relationship between happiness and per-capita consumption (technically, the ecological footprint). The individual differences have suggested the possibility of replicating the conditions of the happiest, least consumptive countries, for the whole world in the hope of salvaging something like our civilization in a future severely constrained in resources. If you've studied my work, you will recall that the correlation I found between population and per-capita consumption has this same characteristic, which is why I've stressed the condition of total isolation for its validity, and presumably the validity of the correlations between per-capita consumption and both happiness and life expectancy. Now, take this reasoning a step further, which is what I did when I generalized these relationships to the past and the future, and you can see why the warning about just such generalizations popped into my head, especially given the suggestion of an exception in the present-day.

In the case of my math, I'm making a clear statement of bias when I claim that the relationships I found are generalizable to the Earth as an isolated system over all time. I've never knowingly suggested that this claim shouldn't be tested, just that it matches the available data that I'm aware of. Because my projections into the past (“pastcasting”) are only really applicable to the last 2,010 years, such testing could possibly be done by historians, or a combination of historians and archaeologists. I expect that the trends I've identified will be vindicated on a statistical basis for societies living in that span, which is why I've labeled my pastcasts as “simulations” (such as the most recent effort, which goes back to the 1800s). Specific details like those I've identified so far should be falsifiable by a search of the evidence; a preliminary search, like in my recent effort, at least shows that they're plausible.

No comments: