Thursday I went to Mtskheta with some other volunteers, where we looked at the cathedral and trudged our way up a 5k ice and snow covered hill, looked at the outside of Mtskheta Jvari, where St Nino apparently preached to the ancient Kartvelians, were unable to get inside on account of a huge padlock, and shuffled our way back down again for supper of xinkali and katchapuri. I felt vaguely uneasy afterwards, I think because I’m so uncertain about social relations within TLG, never knowing where people will be or if I’ll ever see them again, that I’ve begun to treat every encounter as though it may be the only time I ever see someone, and respond accordingly, staying a little later than I know to be prudent, being a little sadder to leave than I otherwise would, and so on. It’s also kind of funny that the two main options in Georgia are a) visit an ancient church, or b) go to a bar or a supra and drink, usually in that order; the supras are way better.
I haven’t written because I’m not sure what to write. The weather has been unusually cold and snowy al through Europe and into Asia this year, and whereas freezing is normal, 6f is not, as the temperatures were reported to be last night. We’ve had two country-wide snow days yesterday and Thursday, which is apparently almost unheard of here, and they have some plan to make up the time un Saturdays; we’ll see whether anyone shows up. I found out that if I come back in the fall, I do get a flight somewhere in the summer, which is an unexpected benefit I’ve found that while one obviously can’t save any money working for TLG, they are remarkably generous in respect to freedom and encouraging travel (as much as anything I reckon it’s because they aren’t certain what to do with us for more than a few hours a day, and hope that we’ll figure out something on our own).
I’ve been reading Thinking, Fast and Slow by Daniel Kahneman, a statistical psychologist. I didn’t know his specialty when I bought the book, and if I had perhaps I wouldn’t have done so; he’s an interesting, entertaining writer, with many concrete illustrations, descriptions of studies, and personal anecdotes. The book itself is fairly easy to read because Kahneman is obviously writing for a popular audience, and knows how to do so properly, but at the same time I find it a challenge to make sense of what he’s saying, because the questions he’s trying to answer are mostly ones that I wouldn’t have thought to ask, or wouldn’t have pursued even if I did know about them. They’re mostly questions about predictive accuracy and statistical validly in decision making — but when he writes about them they sound much more interesting than that. I’m half way through the book, and am not sure how long it actually is, but it seems to be longer than either the introversion or the neuroscience books I read in the past couple of weeks, but that might just because I skipped significant chunks of the former and really enjoyed the latter.
What stands out to me as both important and obvious, and which tends to be overlooked by statistical psychologists, among others, is that the “unknown unknowns” are not only in the means that we employ, but also in the ends that we desire. For instance, Kahneman says that in many cases a simple algorithm can produce more consistent results (which to him means a correlation of .3 or .4 instead of 0) than human judgement — and the kinds of judgements he’s dealing with are usually of the sort where it’s an unmitigated good to be right: investments, this year’s wine, admitting soldiers into the Israeli army, making a correct inference on a test, or calculating the probability that a randomly selected person will be likely to approximate the base rate probabilities in his category. I’m willing to believe that statistical validity can be helpful in these instances — but I’m hard pressed to recall an instance in my own life where I had to make a decision that was in any way like that. The main difference between the decisions I’ve actually encountered and those studied by statistical psychology, is that they tend to ask different questions — and in real life it’s much harder to know what the most relevant question is. He’s dealing with uncertainty in making well defined predictions, and most people most of the time are not. For instance, wine collectors want to know how good this year’s wine will be a decade from now, and so it makes sense that they should use an algorithm based upon this year’s weather patterns rather than judgement, if it is proven that there is a higher likelihood of successful prediction that way; they’re only really interested in one thing (how good will the wine be?), and so predicting that more accurately is all to the good.
The kind of thinking emphasized in this book is that which is needed by certain specialists, especially collectors, investors, and business owners. It can more accurately answer questions of “if I want this kind of quantitate result, what is most likely to produce an acceptable outcome?” Tucson Unified School District has started conducting interviews with a panel asking pre-determined questions and rating them in separate categories, which is the same method Kahneman developed for screening entrants in the Israeli military. The interviewers hated it, but it’s the military, and efficiency is more important than the feelings of the interviewers: people’s lives might eventually depend on good placements. If a company feels itself to be under some kind of treat, it makes some sense that they might act similarly, and Arizona schools do tend to be threatened a lot. If the panel rating separate categories ad tallying the score algorithmically has been shown to be able to predict a little better whether someone will succeed in the particular areas the interviewers are looking for (I’m guessing things like “use of cutting edge teaching methods,” “increasing test scores,” and “cultural sensitivity” in TUSD), and the administration feels threatened if their employees can’t perform very well in those areas, then it makes sense that they would adopt that method, even if it requires terribly dull and irritating interviews.
As it happens, I’m not a manager, a collector, or a gambler, and I’ve never had to take a test rating probabilities whether someone’s cat in a bag is or isn’t dead. I have mixed feelings about efficiency. For instance, my current program is undoubtably inefficient in a myriad of ways, and is especially efficient in arranging the kind of quality interaction between foreigners and schoolchildren that is most important for acquiring a language. I would like it much better if we were provided more situations of that kind: focused conversation for practicing well defined conventions and skills while exploring interesting topics. In fact, the program doesn’t require that we do that at all, ever, unless we come up with opportunities on our own initiative — though it’s exactly those situations that lead to well developed language skills. At the same time, the kind of program that could organize those sorts of interactions is also the kind of program that could monopolize its participants’ time and leave no room for pleasant surprises or unplanned involvement. It’s the kind of program that wouldn’t leave me with any time and energy to be interested in Nikozi, or iconography, or reading books about statistical psychology, or silly walks up frozen hills.