Discussion in 'General Discussion' started by Agrul, Aug 8, 2017.
I fucking loved me some Frasier.
That's basically frozen yogurt
Agrul is definitely whiter than I am.
Also: that is a fucking milkshake. It has 80g of sugar in it and packs 500 calories (good god). Are you trying to get diabetes?
I drink black coffee w/ the only variation being whether it's hot or iced.
it's 340 calories not 500
but yea it is a milkshake. if i want coffee i don't go to starbucks
also back like 5 years ago which was the last time i frequentedf starbucks prior to my resurgence of interest in it ~2 weeks ago i used to get their mocha light recipe instead
at the time i would mix in a bunch of protein powder w/ it and it was rlly good
cant do that anymore tho bc super lactose intolerant now
& it looks like they dont have an almond milk option for the mocha light which is weird
her point about google's employee base living in the tails is well made
That *is* a good point that I hadn't even considered.
This is also a good point that you can extrapolate more broadly to a lot of bro-academics, "classical" liberals, or "the younger generation/progressives have GONE TOO FAR" liberals, on a lot of different topics relating to diversity.
I just read this, too. I agree.
Vox articles jump the gun sometimes & come across as too righteous/preachy/misandrist, but this was not of them.
She raises a really good point that I haven't seen brought up yet:
It seems like the first thing we should be considering, along with the sociological/historical context that was ignored. You can't just shit out data from a scientific study without bothering to actually connect it to whatever it is that you're criticizing (the Google workplace).
that's what i was referring to w/ "google's employee base living in the tails" sear. it is extremely important and we here at tzt / most larger discussions i've seen of the 'manifesto' have largely overlooked it
it's not like it is impossible to study the tails, but most psych/soc/anthro studies certainly do not (altho the IQ literature has some notable exceptions to that general trend)
Agrul - My main obections all seem to revolve around explanations rather than procedure. Even in the most crude and obvious scenarios - e.g. that degrees centigrade is a poor measure of gravity - it really doesn't matter from a scientific standpoint, because the degrees celsius measure will predict different things than other measures, and that is all that really matters. It becomes problematic when you try to explain things in a way that fits in with preexisting scientific or social categories. But those categories are just subjective, organizational constructs, anyway, rather than things that exist and are real. So whether they fit in or not really doesn't matter form a scientific perspective. So I can't object to your points, even if the extremity you take it to seems fucking weird. I don't know if I explained that properly or not.
It's an important point because, anecdotally, the vast majority of discussions of scientific findings that occur here and other boards I have frequented seem to primarily revolve around "construct validity" in a loose sense. E.G. this isn't really prejudice, this isn't really intellifence, etc.
Also, on a tangent, for research purposes I wish we just had a universal IQ database that had everyone's IQ in it that we could just match to participants every single time we did research. Without adjusting for IQ, research results often have to be taken with a grain of salt. I'd like to be able to just take every study discussed during this controversy and adjust for IQ and see what the results say.
No one had considered that point I think when judging this guy. Which I only find problematic if you were crucifying this guy based on a caricature of what he was saying.
To her first point, I think his claims to support diversity in the abstract ARE important because it established some level of positive intent with his ensuing post. We can assume this guy is misguided, not malicious. As Agrul pointed out his conclusions aren't unreasonable based on the information he was working with.
I bolded the part I have a problem with here, has this been demonstrated that biological differences have a small effect? I can buy it for most occupations, and probably for even most tech jobs if not all, but is this really that decided as it seems to claim?
i agree that construct validity is especially perncious in lay conversations of psychological constructs, bc psych is still not really a very well-established 'science' & bc its objects of study obviously overlap messily with and share names with concepts ppl encounter every day so ppl feel empowered to give their all-important opinions about psychological constructs when the constructs' names sound familiar
but i think you see the same thing happen in the academic literature, & even if you didnt i think the preeminent position seemingly given to construct validity in textbook descriptions of psychometric validation is deeply problematic. validating 'construct validity' should be regarded as a messy but unavoidable psychometric step that involves translating non-scientific but still useful prior information into birthing good scientific constructs. i think that description makes both the strengths (draws on prior knowledge/intuition, which is unavoidable in all scientific work to some degree or other, & especially in early psych work) and weaknesses (it is fundamentally nonscientific & should be viewed as a heuristic guide, not a final criterion for mdoel judgement) of 'construct validity' much clearer than in standard descriptions of it, and places it in its correct role as an important heuristic building block for scientific models, but emphatically not as a final criterion for model judgment
A) we don't and technically can't ever know if 'biological differences have a small effect,' and appreciating that this is an intrinsically complex question to ask is part of what a careful reading of the heritability literature forces upon the reader. taking about the proportion of differences in behavior explained by genetic variation only makes sense in a given context; if you change the context, that proportion can go up or down, so the only way to decisively answer this question is to develop a model that convincingly tells us in all possible contexts what heritabiltiy estimates would look like. that is a much more complicated object than just a number that says 'biological is miportant' or 'biology is not important,' it is probably an impossible object to even write down in any useful fashion, & even if some way were available to us t owrite it down it is almost certainly mipossible for us to develop enough knowledge of biochemistry & biospcyhology to actually do so; in particular we obviously can't actually change to every single possible context and run a twins' experiment to get an estimate in that context, because there are way more than infinitely many contexts
B) although we are limited in our ability to talk about nature vs nurture, their relative importance, & their interactions in the sense expressed in A, i think we can confidently say that in the immediate present and recently experienced contexts:
i) heritability estimates for key personological traits (IQ, extraversion, neuroticism, conscientiousness, openness to experience, agreeableness) have accounted for quite substantial but not overwhelming proportions of observed behavioral differences (30%-80% depending on time period, trait, context, & the estimation procedure's specifics/particular defn of heritbaility),
ii) sex differences in behavioral variation have always been observed as substantial enough to be meaningfully estimable (i.e. there is some kind of difference worth remarking upon) but nevertheless quite modest (significantly less than 1 std deviation, i think? i never really gave a shit about the sex differences literature so my recollection here is shaky) and in particular inter-sex variation is generally found to be much smaller than intra-sex variation
iii) although we will presumably (short of us finding out biochemistry & our world are somehow dramatically simpler than oru current best guesses suggest) never have a knowledge of the relative importance of genetic & non-genetic influences in all possible contexts (which is what we would need to answer the 'how important is nature relative to nurture?' question in general; the right answer is a function that at the very least returns a percentage for any given context in which the question can be asked, not a number that's constant across all contexts), it is worth noting that considerable effort has been invested into studying for at least some traits how easily they can be altered through external shocks. in particular a great deal of time and money was spent trying to induce long-term changes in kids' IQs through temporary early educational programs, e.g. head start. most of these efforts showed initial positive results but quick regression to pre-program IQs when the programs were ended. i dont know that similar efforts have been put into studying the robustness of personality traits in the face of external interventions; culturally we would probably be less excited to fund someone trying to engineer your kid's personality than to engineer your kid's IQ, so i'm guessing less work has been done there
C) although we cannot hope to ever fully answer the nature vs nurtue question bc ofmy fundamental caveats stated in A and B and although we can confidently state what i confidently stated in B, we should further note that:
i) all of the stylized facts we're discussing might look radically different if instead've talking about the population psychologists happen to have heavily & completely nonsystematically sampled from -- which is to say almost purely shithead american college students in their early 20s. in particular if we're interested in a population we believe is very different from early 20s US college kids, like the population of google employees or the population of reasonably competent coders or w/e, then it is unlikely we or manifesto-man have access to the data necessary to even begin to make the kinds've claims he did/that we've been throwing about. vox-lady's 'but goggle ppls lives in the tails' counterpoint raises exactly this issue
ii) we have not lookde at any work that specifically tries to estimate the heritability of sex differences. the usefulness of genetic variation for explaining behavioral differences might be much higher or much lower when you restrict yourself to trying to explain differences between ppl of different biological sexes. this question is at the intersection of the work we've talked about & that manifesto-man talks about, but knowing what heritability estimates for trait behaviors look like w/o taking into account sex & knowing what trait-behavioral variation between the sexes look like does not mean u know what heritability estimates for sex differences in trait behavior look like
D) i had a D but i forgot what it was. food processors r for processing food
Well, I read that whole thing. You are welcome Agrul.
ur gawd damn rite u did
A rat model of depression isn't supposed to be a construct for rat depression, it's supposed to be a construct for human depression. We know what human depression is (sort of) and have measured it in various ways, but you can run an optogenetic study on the human brain. So you have to try to replicate it in rats. Whether or not the forced swim task is a valid construct for human depression (probably not, but whatever) is an important consideration when trying to study these phenomenon, because the goal is to try to generalize it back to humans. It's an important concept in psychology, because we are trying to find ways to experimentally test already observed and measured phenomenon.
I don't know if you actually need the concept of construct validity to do that, and you have already addressed a lot of what I wrote. But I doubt we are going to see it go anywhere, because, at the very least, it is too useful a shorthand to describe the generalizability of results as.
I offended a lot of people in RL by bringing it up, which means it's probably an important topic.
Good. All them bitch-ass psycho logists should get upset. I want them riled up so that when I send out my survey studying the correlation between "being a little bitch" and "loving u some weak-ass construct validity" my results will properly reflect that I already know going into this that loving u some weak-ass construct validity makes u a little bitch & all results to the contrary r just evidence that I accidentally studied the wrong constructs.
Srsly tho I agree it's an important concept. I just don't think its proper role is well-described in modern psychometrics, and I think it is yet another gaping hole in psychological methodology through which researcher degrees of freedom can rush in and continue to destroy all hope of psych ever achieving widespread replicable respectability. In a nutshell my argument is just that it is emphatically not desirable to declare the most important kind of validity to be the kind that also happens to depend most dramatically on researcher intuition & subjectivity. That's just asking for a bunch more shitty Psychological Science pubs that say nothing except that the researcher in question contorted their model and made a bunch of messily data-dependent but analytically unacknowledged choices until the noise started to look like a pony and lo' they were excited because the pony was their construct... until someone tries to replicate it and finds the pony was statistical static all along.
I don't really follow the animal models counterpoint. Attempting to generalize by taking lessons learned when studying models of rat depression to make more general claims about mammal depression or something is a perfectly fine/tremendously necessary/good thing to do, but I don't see why you need 'construct validity' to describe those challenges.
I think the key point is that he is criticising the statistical approach to measuring discrimination and diversity goals in general. Then he unloads a bunch of armchair science to back this, particularly focusing on biological differences. This part contains some inaccuracies and is not very well made, not surprising considering he has little or no training in this area or in academic research in general. This doesn't invalidate his argument though, and what is biological or not is frankly not that important. You just need to weight the evidence and extent of behavioural differences vs. that of discrimination to figure out if gender statistics is a good indicator of discrimination. I strongly suspect it is not, and so aiming for 50/50 is more about social engineering than ending discrimination. Google can do whatever it wants, but their stock holders presumable want decisions made rationally even if the Twitter mob doesn't..
Interestingly, Peter Singer recently came out criticizing the firing of this engineer.
Peter Singer the vegan ethicist?
Aiming for a uniform distribution would obviously be a bit weird, but I think you dramatically overstate the simplicity of 'measuring discrimination,' Vel, and being as behavioral differences are context dependent as well, I'm not really even sure what you're proposing at all.
I picked the animal model examples because:
A) We have a pretty good, empirically tested concept of what the phenomenon we are observing looks like BEFORE creating a model to try to measure it. It's not putting the cart before the horse, as you said.
B) We are developing constructs that are SUPOOSED to model that phenomenon in humans.
C) We have an empirical way to test to "validity" of those construct (we can see how they correlate with human outcomes).
Having said that, I, again, don't know if "construct validity" is even important to that process at all. You can evaluate translational research simply by looking at correlations and not making any judgments as to the validity of the "construct" in any way. Construct validity seems superfluous at best, and harmful (for reasons you suggested) at worst.
Separate names with a comma.