I promised I’d post more information about the things I study, and I don’t, usually due to lack of time and copyright problems with the books and texts I read. That, and me doing other less cerebral things like making giant KitKats.

Anyway, one of the things I study is scientific controversies and the problems with researching a disupted topic. Coinciding with a conference on climate change taking place this week, I recently read a good paper on uncertainty and politics in modelling climate change (PDF), which, unlike the majority of reading I do, is not just restricted to dusty journals, but available on the interweb for free.

I recommend reading the paper (it’s 35 pages but double-spaced) if you have the time, as it neatly encapsulates the problems of modelling climate change, not only the technical ones (modelling is enormously expensive and complex to do on a computer), but the very problem of ‘proving’ a theory. What’s known as the “experimenter’s regress” comes into action: the problem of being forced to act totally rationally when experimenting or modelling. In order to prove that we get the right results that truly reflect reality, we need to have the right experiment (or model); however, we cannot decide what the right experiment is until we get one that produces the right results.

As a result, scientists get split, between those who focus on the how well the model’s construction fits the known and proven theory, with those who focus on how well the model does empirically. And the politicians will follow suit, as they pick the scientists whose findings support their own interests.

This circularity gets even stickier the deeper you look into how modelling and data collection actually are done - the climate models are calibrated and tested on existing empirical data (making them “data-laden”), but the data themselves have to be shaped and corrected, using knowledge of the models (making them “model-laden”). This circularity and co-dependency could lead one to the scary conclusion that what’s being done in climate change modelling isn’t “good science”, or even “science” at all.

But then, what is “science”? In truth, most revolutionary research and paradigm-setting is performed under similar conditions of controversy. Often the truth isn’t the final arbiter for confirming a new theory, because we’re unable to agree on what evidence actually reflects the truth. Climate change is not the exception but more the rule. Even the elementary stuff like measuring temperature is up for interpretation - are satellites or ground thermometers the best? How do we compensate for the (relative) lack of measurements in remote areas, or different altitudes? How reliable are measurements taken 100 years ago? How can we measure the temperature from 1000 years ago?

In short, science is horrible. It’s rarely straightforward and nearly always squabbled over. The “truth” (or at least, the best theory we can come up with that reflects the truth) usually wins out at the end of the day. Though with something like climate change, hanging round 100 years or so to see whose theory gets confirmed (like waiting to see whether Bangladesh gets submerged under six feet of water) is a pretty risky business, given the stakes involved. But considering these fears, deciding whether to err on the side of caution or risk it, whether to trust in one side or the other, is to introduce a degree of irrationality into our judgement - the only way of breaking the experimenter’s regress that rationality presses upon us.

After some shitbag tried spamming the site with nearly 150+ trackback requests, I’ve disabled the feature until I can find some better way of implementing them.

(Public Service Announcement ends)