Skip to main content

Grab a couple of dice. Roll them.

If you get below 5, those are rookie numbers. Shout at the dice, let them know they're underperforming.

If you get above 9, that's what we want to see! They're good dice, and you should acknowledge that.

Repeat that and keep a record. You'll notice that negative feedback often results in better performance on the next roll. Positive feedback, conversely, can make them get lazy.


When you truly understand why this method of dice management works, you are ready to give feedback to people.


reshared this

I remember reading about this effect in The Drunkard's Walk, including how people cited this very effect refusing to believe the psychological research showing positive feedback is more effective: improvement in performance after negative feedback, and a dip after positive feedback, all explained by normal chance variations in performance.
@ljwrites it's the closest thing I know to mathematical proof that the devil exists! Such a powerful effect, coming only from basic maths and a little bit of human psychology
Yeah, it trips basically all our fallacy wires xD

statistics is the true devil 's science, even experts who should know better will get it wrong often.

See Monty Hall Problem Wikipedia page.

3/2 RIP Daniel Kahneman, who could explain things with both lucidity and warmth
Is this a genuine psychological effect?

@cdonnellySRE it's a genuine statistical effect! If there's a higher than average score, the chances are the next randomly-chosen one will be closer to the average. Same applies for lower scores - the next random score is more likely to be closer to the average (i.e. higher).

How you end up feeling about that, *that's* the psychological effect. The dice don't care 😀

@cdonnelly @серафими многоꙮчитїи This is something we learned in introduction to Psychology, as like, an important example of what's called The Armchair Effect in how people perceive psychological studies.

The idea is that a lot of people either reject data because they have lived experience, or feel like gathering data is worthless because it is lived experience, and this is the standard example courses use (at the University of Waterloo a decade and a half ago) of something where people often reject the data because of their lived experience.

@silverwizard @cdonnellySRE this is interesting but kind of the opposite of what I was getting at if I understand you right: if you follow the procedure and record the data, you absolutely will see a correlation between berating the dice and rolling higher next time!

(Of course this thread was about being misled and is therefore the perfect place for me to grab the wrong end of the stick)

@серафими многоꙮчитїи @cdonnelly I mean, yes, it's obvious when you look at the data, but so many people just don't do that, so it's a great example of how doing data analysis can make a lived experience evaporate.
@silverwizard I don't see the original post as being about dismissing lived experience, more about confusing correlation (reversion to the mean) and causation (my feedback has an obvious impact!)
@silverwizard I see. Maybe it's a regional dialect, but I would never use [or understand] "lived experience" to mean "cognitive bias"; I've typically heard it in the context of e.g. "a lived experience of racial discrimination".

@silverwizard @cdonnellySRE I never actually come out and say it directly in the main toots: I wanted to leave it as a fun little puzzle for people because I certainly find it counterintuitive.

It got a few boosts and I'm now self-consciously wondering how many people were thinking "this guy is trying to show why it's important to give negative feedback (and he's bad at statistics)" or even "this jabroni thinks he can influence dice by shouting". 😂

What is this even saying? That an individual's workplace performance is a stochastic process? That a skilled-enough manager can do the equivalent of "rolling" a person such that their performance is biased towards being better-than-average?

I mean, I get the surface message, but I don't think it holds up for precisely the reason that people aren't dice.

Yes, and that's why I'm asking. Reading 2/2 charitably, I was hoping you could help me "fully understand" what you meant by the dice analogy.


(a) the dice are not responding to your feedback: they're dice

(b) nonetheless, the stats "show" that your negative feedback was more effective than your positive feedback (this is a genuine thing you can try if your Friday nights are as exciting as mine...)

People are a little stochastic, although that's not really the point: we have good days and bad days influenced by factors that we can't predict well. In fact, the first time I saw this effect referred to it was in the context of actual human beings, and it occurred to me that it would be amusing and maybe clearer to put it in the context of inanimate objects.

It's a really powerful cognitive illusion caused projecting an emotional response onto regression to the mean, and a similar effect can definitely cause people to believe it's in some way effective to mistreat others. To be clear, I don't think it is.

I guess it's not really an analogy, it's a thing you can actually do, and the dice don't even have to represent anything other than their cubical selves.

I'm going to link to this excellent explanation (with code!) in case any future travellers get here

Inspired by @derwinmcgeary's post ( from a couple of days ago, I wrote a quick blog entry about teaching, feedback, and regression towards the mean. I hope I didn't get the explanation wrong and that I understood the point correctly!