The reason I was reading Scientific Revolutions had to do with my conception of economics as a science. A big question I am set on resolving is the nature of the very clear distinction between the natural sciences, on one hand, and the social sciences, on the other. What makes physics or chemistry so much more science-y than economics or — wince — sociology? Now I feel a lot closer to having an answer.

The last chapter in particular was elucidating. Perhaps (along with some methods, mindsets, and other similarities) the common point among the sciences is the homogeneity of their adherence to one particular paradigm. Science seems to progress so linearly because of an almost Orwellian process of rewriting history from the standpoint of the dominant paradigm.  When all a field’s practitioners are trained using one set of textbooks, and after each revolution the textbooks are rewritten, is it any wonder that this adherence of all a field’s scientists to one paradigm is achieved? From the other direction, is it a black mark on macroeconomics to have enough factions, and enough different text books that are not agreed upon, that there is not one dominant paradigm in use that constitutes the field?

More on this to come.

Freezepop has a song called “He says, she says” which accurately portrays the sorts of situations arising when we don’t have the above Life Indexer. Imagine how useful it would be to go back and accurately rebut: “No look, I actually said this, not that.” I believe it would be insanely useful.

My biggest issues arise from data sources I can’t look up on Google or Wikipedia. When I make a commitment with someone, I can only search through the mementos I write myself stemming from the social interaction we had that generated the commitment. That means when we disagree, it’s my word against theirs.

You can’t do science on unfalsifiable statements.

I’ve found one that’s very, very sneaky. I call it “measurability bias”, which I’m pretty sure I read somewhere but now I forget the exact reference. It comprises a few similar phenomena. For one, it is a tendency for researchers — notably those who pride themselves as scientists — to spend more time and resources focusing on problems where data are readily available, rather than what they identify as the most interesting or important problems.

A more general conception of measurability bias is when a decision maker weights more heavily the set of things that are quantitatively or accurately measurable when making decisions. For example, when Oreos increase in price by 20%, that’s very easy to see, but I’m not as good at noticing if the increase in quality leads to a greater-than-20% increase in the satisfaction I derive from said Oreos.

Cost-benefit analyses suffer from this bias a great deal. For example, in the case of climate change, the costs of carbon reduction programs are known relatively accurately, whereas things like the mitigated risk from rising sea levels, loss in biodiversity, chronic water scarcities in developing countries, etc. are much harder to measure. In this case, measurability bias is used as an excuse to do nothing — since the costs are large and the benefits are uncertain, we should defer climate stabilization policies.

The point to take away is that just because something is difficult to quantify does not mean it is not incredibly important.

But what if I want to figure things out in the very important realm of human relationships — specifically the romantic kind? Tough luck. I mean, first of all, how can you do science on relationships? There are too few data points! Even if you are Mr. Playboy and date, say, one new girl per week, it would be nearly seven months before you could use the OLS large sample assumption. And that’s no good, because then you’d just be testing something about short relationships — probably not the kind you’re interested in studying.

There is no control group: you can’t clone yourself to see how you would act under two different relationships. Besides, you’re not a static person either. Each relationship you’re in changes who you are to a great extent. Even if you determined some profound result, it would probably only be valid on a past self.

You would need to use a between-subjects design. It would be virtually impossible to organize two groups of people in relationships and keep the treatments the same. How would you control for all the possible differences in peoples’ personalities and relationship dynamics?

No, doing science on relationships is all but impossble. We’ll probably have to settle for psychology or something. (See what I did there? Thinly-guised psychology insults are the first thing a budding economist learns.)

Friends, that isn’t science. Science is about testing hypotheses with empirical data. And who has the data? We have the data. Let me introduce the Web Ecology Project.  This project is, among other things, about basing qualitative analysis on firm quantitative footing. Check out our first report on the election in Iran to see some real Twitter research.

Already we’re getting some sweet press. It’s only going to get more awesome from here, folks. Stay tuned.