Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Bayesian analysis of ego-depletion studies favours null hypothesis

onlinelibrary.wiley.com

Post reply on HN
Print view View original

21–30 of 89 posts

Re: Bayesian analysis of ego-depletion studies favours null hypothesis

#21

Post by mikraig »

dbecker wrote: post #3

It's disappointing to see links to the abstracts on the front page of HN (given the paper itself is behind a paywall). In this case, the find some non-zero effect, but call it zero because the difference was not statistically significant. That's likely a reflection of their small sample size rather than evidence for the null hypothesis as the title suggests. But it's hard to know, or even have an intelligent discussi…

In Bayesian analysis, it's possible to actually accept the null, rather than just the reject/not reject that is common to frequentist statistical inference.

Re: Bayesian analysis of ego-depletion studies favours null hypothesis

#22

Post by averagewall »

OskarS wrote: post #9
gbear605 wrote: post #4

Earlier quoted context omitted.

Ego depletion refers to the idea that self-control or willpower draws upon a limited pool of mental resources that can be used up [1]. Previous studies have found that it is real. This study did a complex statistical analysis and found that they didn't have enough evidence to prove that it is real. [1]: https://en.wikipedia.org/wiki/Ego_depletion

To give some extra context: this is another study confirming the "replication crisis"[1] in psychology and social psychology (and other sciences as well, but these are the most prominent fields) where longstanding results that were thought to be rock-solid are falling like dominoes. Basically, the science backing, like, half of all TED talks and most episodes of Radiolab is crumbling before our eyes. [1] https://en.w…

It seems like the least competent people to be designing and analysing psychology studies are psychology researchers. Statistics should be the number one skill required. The psychology part of these high-profile studies is usually trivial so that anyone with no psychology background could have thought of it, or at least easily understand it. I don't really see the need for a specialist psychology researcher to do this kind of work. Maybe we need a new field - "studies of interesting everyday phenomena" that's mostly about doing statistics right.

Re: Bayesian analysis of ego-depletion studies favours null hypothesis

#23

Post by mizzao »

krashidov wrote: post #13
colemannugent wrote: post #5

Earlier quoted context omitted.

Basically, there was a theory that exercising self-control would "deplete" ones "self-control reserves". The idea was that exercising your self-control, to say avoid a junk food, would temporarily decrease your ability to resist that temptation again. This paper reports that not only were they unable to verify any of the prior research into this so called "ego depletion" effect, but they found more persuasive evidenc…

This is reassuring. Every time someone told me this I figured it had to be bullshit. I think people used it as an excuse to make bad decisions.

...and you're exemplifying confirmation bias right there, which is the whole reason this happened in the first place.

Re: Bayesian analysis of ego-depletion studies favours null hypothesis

#24

Post by debacle »

colemannugent wrote: post #18

I always get the feeling from reading academic material on psychology that people are trying to apply the scientific method on top of unscientific principles. It feels rather like alchemists trying to explain fire via phlogiston: an attempt to quantify high-level phenomena when you don't understand any of the lower levels. All the data comes from psychology students at a college in the US who either were able to use…

Data collection in soft sciences definitely turns it into a Hardmode, but that doesn't mean all findings are bunk.

https://vimeo.com/118188988

Re: Bayesian analysis of ego-depletion studies favours null hypothesis

#25

Post by closed »

KirinDave wrote: post #12
nonbel wrote: post #10

Earlier quoted context omitted.

I agree, there is no reason to mention "bayesian" in the title other than using it as a hype term like "blockchain" or "deep learning". Also, the real difference is whether they test a "null hypothesis" vs testing an "actual hypothesis". Like almost all psych, this falls in the first category.

You say this but then you go read a paper with a cruddy t-test and then injure your eyes with the spasmodic rolling that it induces, like I did. It's kinda remarkable how many people don't actually use modern bayesian analysis techniques, which is why it's such a policy war. It's the difference between "older statistical methods uninformed by the glut of modern processing power" and "things which require a computer t…

I don't think it's an issue of old versus new methods. Daniel Lakens has an great string of blog posts / publications on how a lot of the same things can go wrong with Bayesian anaylses [1], and how p-values can be viewed as a "poor man's Bayesian updating function" [2].

[1]: http://daniellakens.blogspot.com/2017/03/no-p-values-are-not...

[2]: https://daniellakens.blogspot.com/2015/11/the-relation-betwe...

Re: Bayesian analysis of ego-depletion studies favours null hypothesis

#26

Post by pavement »

colemannugent wrote: post #5
piratebroadcast wrote: post #2

ELI5?

Basically, there was a theory that exercising self-control would "deplete" ones "self-control reserves". The idea was that exercising your self-control, to say avoid a junk food, would temporarily decrease your ability to resist that temptation again. This paper reports that not only were they unable to verify any of the prior research into this so called "ego depletion" effect, but they found more persuasive evidenc…

Gross. It sounds like a total failure to correctly model reality.

Like measuring how blue the sky is, based on how much "blue" is left in the universe. The idea they attempt to quantify is improperly described in terms of qualities. e.g. Do colors even work that way?

(...and keep your planck unit, M theory, quantum field eigenstate, history-of-the-big-bang, age of the universe remarks away from this hypothetical example, please. You know what I mean when I describe the sky as either blue or not blue in the colloquial sense.)

Re: Bayesian analysis of ego-depletion studies favours null hypothesis

#27

Post by tinymollusk »

crawfordcomeaux wrote: post #20
tinymollusk wrote: post #17

Earlier quoted context omitted.

To me, this is an interesting statement because it highlights the difference between true and useful. For several years, since I read the book Willpower (an intro piece to ego depletion), I've meditated on the idea. I've decided that even if it's not scientifically true, it can be useful to me. Some may see it as self-help tripe ("Eat your frogs in the morning!"), but it's a useful mental model for me to push myself…

I started using animism yesterday to motivate me to do the dishes, viewing them as the physical embodiment of souls that I, their god, allowed to get moldy/rusty through neglect. I think when we introduce choice/attention into the theory that we construct reality out of beliefs, emotions, culture (thought/behavior patterns), and intentions, it's possible to find all kinds of useful frameworks. Maybe the idea of "scie…

How do you think we could adapt that concept to subjective experiences? As I've gotten older (and learned to manage my mental state better), I've found great benefit from separating "scientifically true" and "useful to lead a good life".

I suppose that makes me a hypocrite in the same way that devoutly religious scientists are. But it makes my life more fulfilled, I experience less anxiety, etc.

It's an interesting ride, whatever it is.

Re: Bayesian analysis of ego-depletion studies favours null hypothesis

#28

Post by closed »

I'll have to look more closely at the paper, but am very surprised by this result! I've always figured the effect sizes were quite small, and there are good researchers who don't buy its main explanations, but the idea of ego depletion shows up in a lot of areas under different names (e.g. "resource depletion in working memory").

Pretty contentious area of research, so I'm sure this will be some nice fuel for the fire..

Re: Bayesian analysis of ego-depletion studies favours null hypothesis

#29

Post by ajross »

colemannugent wrote: post #18

I always get the feeling from reading academic material on psychology that people are trying to apply the scientific method on top of unscientific principles. It feels rather like alchemists trying to explain fire via phlogiston: an attempt to quantify high-level phenomena when you don't understand any of the lower levels. All the data comes from psychology students at a college in the US who either were able to use…

Modern experimental science is just garbage on the whole. Psychology has the disadvantage of a very poor theoretical foundation, making a lot of its pronouncements seem hokey and fabricated. But on the whole I don't think the prevalence of poor measurement and (espeically) bad statistics is any worse there than it is in, say, biology. Unreproducible experiments are the rule in modern science, not the exception.

Re: Bayesian analysis of ego-depletion studies favours null hypothesis

#30

Post by KirinDave »

closed wrote: post #25
KirinDave wrote: post #12

Earlier quoted context omitted.

You say this but then you go read a paper with a cruddy t-test and then injure your eyes with the spasmodic rolling that it induces, like I did. It's kinda remarkable how many people don't actually use modern bayesian analysis techniques, which is why it's such a policy war. It's the difference between "older statistical methods uninformed by the glut of modern processing power" and "things which require a computer t…

I don't think it's an issue of old versus new methods. Daniel Lakens has an great string of blog posts / publications on how a lot of the same things can go wrong with Bayesian anaylses [1], and how p-values can be viewed as a "poor man's Bayesian updating function" [2]. [1]: http://daniellakens.blogspot.com/2017/03/no-p-values-are-not... [2]: https://daniellakens.blogspot.com/2015/11/the-relation-betwe...

When I do not have a processor easily capable of bayesian mkmv sampling I'll consider using that method.

But I do and I guess this means I am rich? Sure.

Post reply on HN