Will Lowe
@conjugateprior.org
Señor Research Scientist, NPC at the Hertie School in Berlin 🇩🇪 via Princeton, Mannheim, Edinburgh and a bunch of other ivory towers that will probably be billiard balls and decorative boxes by the end of the decade.
Rome Statute appreciator. ..
more
Señor Research Scientist, NPC at the Hertie School in Berlin 🇩🇪 via Princeton, Mannheim, Edinburgh and a bunch of other ivory towers that will probably be billiard balls and decorative boxes by the end of the decade.
Rome Statute appreciator.
Computer science
32%
Political science
28%
Pinned
Will Lowe
@conjugateprior.org
· Apr 15
For Monty Hall problem aficionados: a #causal DAG, with explanation in the alt text.
Who knew that the M in M-bias stood for Monty?
Who knew that the M in M-bias stood for Monty?
Sometimes Harvey enters and participants see him from the left. Sometimes Harvey enters and participants see him from the right.
Attenuation, heterogenous treatment effect, or plain old SUTVA violation?
Attenuation, heterogenous treatment effect, or plain old SUTVA violation?
and still others will turn out to be simply out of scope for psychological theory altogether, like phlogiston as an explanation of colour.
So bring it on. We'll see what's standing when the dust settles.
3/3
So bring it on. We'll see what's standing when the dust settles.
3/3
Others think they show explanatory causes up close, like phlogiston or caloric.
But if thermodynamics is a guide, we should expect (hope, even) that in a truly mechanical theory, many will turn out distinct only because of our interests, like burning; others will just drop out, like caloric;
2/3
But if thermodynamics is a guide, we should expect (hope, even) that in a truly mechanical theory, many will turn out distinct only because of our interests, like burning; others will just drop out, like caloric;
2/3
And this. Cognitive dissonance can no longer be regarded as established.
journals.sagepub.com/doi/full/10....
journals.sagepub.com/doi/full/10....
A Multilab Replication of the Induced-Compliance Paradigm of Cognitive Dissonance - David C. Vaidis, Willem W. A. Sleegers, Florian van Leeuwen, Kenneth G. DeMarree, Bjørn Sætrevik, Robert M. Ross, Ka...
According to cognitive-dissonance theory, performing counterattitudinal behavior produces a state of dissonance that people are motivated to resolve, usually by...
journals.sagepub.com
Whether or not you think replication failures show that cognitive dissonance (or dunning-kruger or grit or...) is not a thing, the big question is: where do 'effects' belong in psychology?
Some think they're just practically-derived motivating phenomena, like fire and rusting for thermodynamics
1/2
Some think they're just practically-derived motivating phenomena, like fire and rusting for thermodynamics
1/2
Will Lowe
@conjugateprior.org
· 17h
imho, as a long time computer person, it should feel like using with a vacuum cleaner. Maybe a clothes dryer. That it's occasionally like handling a firearm is horrible, but still not as bad as realizing it's like negotiating with a hostile bureaucracy somehow taken over by an advertising agency.
Reposted by David Papineau, Will Lowe
A Multilab Replication of the Induced-Compliance Paradigm of Cognitive Dissonance - David C. Vaidis, Willem W. A. Sleegers, Florian van Leeuwen, Kenneth G. DeMarree, Bjørn Sætrevik, Robert M. Ross, Ka...
According to cognitive-dissonance theory, performing counterattitudinal behavior produces a state of dissonance that people are motivated to resolve, usually by...
journals.sagepub.com
And this. Cognitive dissonance can no longer be regarded as established.
journals.sagepub.com/doi/full/10....
journals.sagepub.com/doi/full/10....
Will Lowe
@conjugateprior.org
· 21h
Ooh, this page has the English mnemonics on it too. My favourite, which you have to pronounce in Garamond with contextual ligatures, is
My Very Elegant Mother Just Sat Upon Nine Porcupines
My Very Elegant Mother Just Sat Upon Nine Porcupines
Will Lowe
@conjugateprior.org
· 21h
"Talk to the gas giant"
Will Lowe
@conjugateprior.org
· 21h
Yeah, research infrastructure and academic software development are, looking back on twenty plus years of it, one of my regrets. Doing things that need doing, without asking why they're not already being done, and pursuing what you enjoy & are good at are basically bad ways to make career choices.
Will Lowe
@conjugateprior.org
· 21h
Voice training
Will Lowe
@conjugateprior.org
· 22h
How'd they get away with that, I wonder.
Will Lowe
@conjugateprior.org
· 22h
I can't even remember the top five Generals, to be honest. One of them must have been Monty in shorts.
Will Lowe
@conjugateprior.org
· 22h
Oooh, I forgot General Studies (still love that name though). They'd just brought AS-levels in and I think they had us take GS as one of them, just kind of because they could, with no obvious preparation. Or that's how I remember it. Though it was quite a long time ago, so I could be misremembering.
Will Lowe
@conjugateprior.org
· 22h
It is still boggling to my wife that I have never, at any age, thought it would be cool to be an astronaut and that the only really interesting bit of physics is thermodynamics.
Will Lowe
@conjugateprior.org
· 22h
At the time you had to choose 3 subjects (4 at a push) at 16 to take to A-level at 18. No STEM subject made the cut – to the very obvious relief of a lot of my high school teachers.
Will Lowe
@conjugateprior.org
· 22h
Cool. There's probably an English one, but who gave up physical sciences at 16 and never looked back? 🙋
Will Lowe
@conjugateprior.org
· 23h
Monster: ...but you get to choose what.
Me [thinking furiously]: Aha! Feathers. I'd like to be hit by a ton of feathers, please.
Monster: You don't understand how weights works, do you?
Me [thinking furiously]: Aha! Feathers. I'd like to be hit by a ton of feathers, please.
Monster: You don't understand how weights works, do you?
You're welcome.
Prob the most useful thing is to use the definitions as thinking tool, and ask which of say {controlled direct, natural/pure direct, component} effects are interesting for explanation. Your description earlier suggests component effects might be the right *sort* of thing.
Prob the most useful thing is to use the definitions as thinking tool, and ask which of say {controlled direct, natural/pure direct, component} effects are interesting for explanation. Your description earlier suggests component effects might be the right *sort* of thing.
Right. I do get the plan. The great hope of the original mediation folk too was that they could somehow race pathways against one another. I think that's pretty well unraveled at this point, for reasons that are too convoluted to rehearse here, but not everyone agrees :-D
I think the (earlier) argument for the non-inclusion of mediators might be the part where an estimand would clarify. Sure, leaving out the mediator gets you a 'total effect' of D but there's 2-3 'direct'ish D effects you could want, depending on how you think about intervening on it & Z.
I confess I've read that paper a few times, and I still can't quite get what they're on about. So you could be doing exactly it, for all I know 🙂
I don't know Dowding, so this might well be just what they said, but at least one general critique is that "theoretically interesting comparisons" just _are_ causally identified comparisons, in the sense of being the ones that pick out theory-predicted effects of particular interventions. But ymmv.
True. I was running with the paper's graph that had a direct route, not my reply's hypothetical where there wasn't one.
I note in passing: in the no direct X->Y version, an X + Z model would condition on an instrument, which would be unfortunate for predictive power (plus helping Z->Y confounders)
I note in passing: in the no direct X->Y version, an X + Z model would condition on an instrument, which would be unfortunate for predictive power (plus helping Z->Y confounders)
And in the second shot, I wanted to ask: does "such causal mechanisms are detached from the demographics" just mean demographics are screened-off by conditioning on a mediator, or something else?
tbh I find the mix of predictive comparisons and causal structure without causal estimands v confusing
tbh I find the mix of predictive comparisons and causal structure without causal estimands v confusing
tbh I also wondered about some of these sections.
In the first screenshot, that ideology mediates demographics _rather than_ demographics confound ideology seem like a very odd claim. In this structure both are true for different estimands.
In the first screenshot, that ideology mediates demographics _rather than_ demographics confound ideology seem like a very odd claim. In this structure both are true for different estimands.
(Which is another way to say that you can't identify the effect of X by conditioning on Z and looking at predictive performance because Z is 'post-treatment'.)
In this structure, I think we'd want to say that both theories are correct – just fighting over how close to Y explanations should be – no?
In this structure, I think we'd want to say that both theories are correct – just fighting over how close to Y explanations should be – no?
Not sure I follow your logic.
Assume X is demographics, Z is ideology (as you do later) & that the X->Y path is negligible. Then X is a cause of Y, and Z is too, just nearer. Z screens off X from Y, so Z predicts Y best, X predicts Y but worse and a model with both is barely better than one with Z
Assume X is demographics, Z is ideology (as you do later) & that the X->Y path is negligible. Then X is a cause of Y, and Z is too, just nearer. Z screens off X from Y, so Z predicts Y best, X predicts Y but worse and a model with both is barely better than one with Z
Further work is needed to see how the effect varies with multiple batmen, when the pregnant woman is also wearing a batman costume, and when there is also basketball playing gorilla*. This is just how science works. I don't make the rules.
*we thank an anonymous reviewer for this suggestion
*we thank an anonymous reviewer for this suggestion