um I read a really interesting thoughtful blogpost about analyzing activity data and the comments on it were like FULL EUGENICS IQ MATH ZONE, the grey literature in tech is a trip
an interesting mix of horrific bias + a slight understanding of collider bias but not so much as to counteract the notions of the first thing but instead ending up at literally the opposite conclusion about what we should expect from observed performance, yikes
feels like I need to ask for hazard pay out here
In Your Ancient Comments, Thinking About Your Beliefs like some kind of parody of a psychologist here
Also everybody involved needs to read this piece:
You cannot understand performance observed in an unbalanced field without understanding unbalanced selection and how it operates DIFFERENTLY for the groups penalized
Finishing this year literally as I started it (I was reading some January writing of mine where I was upset about human ability beliefs re: "programming brains")
My heart just fucking breaks for everyone in this industry who has to be exposed to the rancid garbage of these essay length comments about "women and IQ distributions" (and this is what people are willing to say out loud in a way I can read years later!). So terribly flawed. Using the faint suggestion of Math And Science to intimidate. Reading this stuff is totally wretched but informative. It reminds me in every way of the arguments I grew up around.
@grimalkina My experience has been that most software engineers I meet do not have an especially deep math background but do place a lot of value on math, so using even slightly sophisticated math tends to carry a lot of weight with many of them. Unfortunately, this remains the case even when the argument is completely wrong.
@grimalkina I think the other thing is that most software folks are coming from an engineering mindset, whereas a scientist needs to be more like a blend of an engineer and a philosopher, grappling with questions of epistemology and ontology. Hopefully, most scientists in our training learn that discerning what's really going on in a system is far more complicated than just making some measurements and doing some math; some of the most pivotal questions are around what to measure, how to measure, and what assumptions to make in analysis.
In a science like psychology, this is presumably much more pronounced, due to the irreducible complexity of the phenomena and the limited ability to control experiments (requiring much more sophisticated thinking about experiment design and analysis). And added on top of that you have the problems that come with mitigating one's own biases on questions that touch upon one's own sense of identity and beliefs.
Which is all to say that I think many software engineers (and engineers of other types) are especially prone to buy into mathematically-based pseudoscience, especially in social and behavioral areas. And it's exactly why we need people actually trained in those areas, like yourself, to rely upon rather than amateur hour.
