1/ I use and respect statistical and agent-based models. I'm a computational social scientist. But, associated with my respect for those methods is the appreciation of the limitations. Modeling people -- especially groups of people -- is **very** challenging.
-
Show this thread
-
2/ When people make claims about models of social phenomenon being objective, they're usually full of shit. The map isn't merely not the terrain. It's sampled from the set of possible maps that satisfy the modelers' demands.
1 reply 0 retweets 5 likesShow this thread -
3/ Describing reality is one such demand. It's not the only one. For people who claim objectivity, it's certainly the (imputed) principal one. Yet, researchers can't easily remove themselves from the process that conditions model selection.
1 reply 0 retweets 0 likesShow this thread -
4/ That is, the set of explored models isn't sampled uniformly over the space of possible models. Different methods preferentially select for different models and different data. And, since historical data is so desperately sparse, vast regions of that space are absent.
1 reply 0 retweets 0 likesShow this thread -
5/ Furthermore, publish or perish creates pressure to use data / methods that are similar to those that had "statistically significant results." This isn't just an academic problem. It's exists for anyone using models in arguments -- which, is basically anyone using models.
1 reply 0 retweets 0 likesShow this thread -
6/ Models are useful if you're able to carefully evaluate their limitations and power, in a particular context. Most political arguments drawing upon empirical models don't do this. They strip that part away.
1 reply 0 retweets 1 likeShow this thread -
7/ In the end, even a well-designed model gets used to bludgeon a (usually untested) counter-argument -- rhetorical ammunition with the pretense of mathematically exacting objectivity. Which is to say, their merit is evaluated and propagated subjectively. That's pretty weird.
1 reply 0 retweets 1 likeShow this thread -
8/ I keep tweeting threads in this vein because it bothers me more and more as time goes on. I wrote about this a while back in this link below. And, I'd really like to work on this in the future.https://dispatches.artifexdeus.com/what-is-your-political-loss-function-a32749c5c37f …
1 reply 0 retweets 1 likeShow this thread -
9/ But, I wonder if social science research would benefit from a loss function statement with abstract-level prominence. And, a statement of invalid contexts w.r.t applications.
1 reply 0 retweets 0 likesShow this thread
10/ With increased access to scientific literature, I think researchers need to appreciate and anticipate the potential misuse of their models. #scicomm shouldn't be seen as a particular type of job. But, as a role with varying degrees of commitment.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.