Bias and ignorance in demographic perception

@dggoldst and i was thinking about your tweets from some years ago about works like: https://t.co/H3FuUDKY9w

@simardcasanova https://t.co/NKrHAM6bQB

Critique of this chart (haven’t read the paper yet):

@David_J_Bier My sense is that these numerical misperceptions are much less important than the widespread ignorance of immigration policies you highlight. Some even argue people overestimate small groups and underestimate large groups in general regardless

@sundersays @CommSenseUK yep it's misleading, particularly about v small/rare items. Best systematic explanation is from psychophysics, eg from David Landy et al in this https://t.co/tYnmdjalPt Basically people hedge to the middle with uncertainty, resul

Bias and ignorance in demographic perception https://t.co/roYmqYZ2lP

@expresso Quando se pede a uma pessoa para estimar a proporção de algo, ela tende a sobreestimar o pouco frequente e subestimar o muito frequente. Há um enviesamento para o 50%. https://t.co/8WxjlGwXsz https://t.co/neZQ936C0Q https://t.co/tSJBABD4Cj

Quando se pede a uma pessoa para estimar a proporção de algo, ela tende a sobreestimar o pouco frequente e subestimar o muito frequente. Há um enviesamento para o 50%. https://t.co/8WxjlGwXsz https://t.co/neZQ936C0Q https://t.co/TiF6GilHqP

@o76923 @Travon Cheers - the guesses are just so crazily off base, but there’s an interesting embedded link that offers an up the nice, gentle hypothesis that it’s mostly due to a combination of personal familiarity w a topic + “correction to the mean” bia

RT @j2bryson: Bias and ignorance in demographic perception https://t.co/OYJQ8B6Rm4 Biased estimates of eg %age LGBT or Islam are congruent…

RT @j2bryson: Bias and ignorance in demographic perception https://t.co/OYJQ8B6Rm4 Biased estimates of eg %age LGBT or Islam are congruent…

RT @j2bryson: Bias and ignorance in demographic perception https://t.co/OYJQ8B6Rm4 Biased estimates of eg %age LGBT or Islam are congruent…

@paulrauwolf

Bias and ignorance in demographic perception https://t.co/OYJQ8B6Rm4 Biased estimates of eg %age LGBT or Islam are congruent w/other known such biases, so not nec. due to topic-specific explanations eg misinformation, media bias, or xenophobia. via @kstrum

Mais en fait, il s'avère qu'il y a des mécanismes de pensée très généraux qui expliquent ces erreurs. Par exemple, en situation d'incertitude, on tempère notre estimation en la rapprochant de l'estimation qu'on ferait en l'absence d'information. /8 https:/

Overestimating minority positions and underestimating majority positions is a feature of human psychology regardless of the subject matter https://t.co/Gie2FSxOr0 https://t.co/Doe9Q4cZTf

@biggestjoel Seems to be a legit thing that happens, tho https://t.co/wQZ6omvAMt https://t.co/W9I6r6eMpU

@TheHelloFreaks people suck ass at estimating percentages, especially close to zero https://t.co/b8tdwbRL1Z

RT @BrendanNyhan: @gilmored85 @fawfulfan Read @BrianMGuay https://t.co/UO52HoThUO https://t.co/rKGQRWBaJY

paper for the academically inclined: https://t.co/iXNLhXzijy TL:DR the estimation errors in this survey stem mostly from either: - Bayesian priors that don't align with reality - proportions are represented mentally as log-odds, and survey responses are a

@ryancbriggs https://t.co/lFwehUEGeN

@no_earthquake that's bc of bad survey design, ppl consistently error towards even odds when asked to estimate percentages like this https://t.co/iXNLhXifhy

and this: https://t.co/jpdgxdrsln and this: https://t.co/YqMBcbgyej @DrDavidLandy @TylerMarghetis

RT @BrendanNyhan: @gilmored85 @fawfulfan Read @BrianMGuay https://t.co/UO52HoThUO https://t.co/rKGQRWBaJY

@gilmored85 @fawfulfan Read @BrianMGuay https://t.co/UO52HoThUO https://t.co/rKGQRWBaJY

@BrendanNyhan @FabianNeuner @prof_mirya here are some! https://t.co/tDnINjcISU https://t.co/tqirf3ZXws https://t.co/jpdgxdrsln

(2/2) There is a great paper illuminating this phenomenon. It seems to not mainly be a bias related to specific populations. It is a domain-general phenomenon where people overestimate small proportions and underestimating large ones. Source: https://t.co

@BarneyFlames There is a great paper on this. "Our central thesis is simple: Demographic proportion estimation is just one specific kind of proportion estimation, and thus is subject to the same psychological processes that humans use to estimate other p

Reminds me of how humans tend to err in their estimation of proportions

@kelvmackenzie Shriver writes nonsense. 1: Other factors such as xenophobia/homophobia in play apart from media. 2: If she’d bothered to report research she’d find that ‘Small values are overestimated, and large values are underestimated, regardless of th

RT @profcikara: Really cool paper! Also dovetails with a lot of cool work on overestimation of size of small demographic groups https://t.…

RT @profcikara: Really cool paper! Also dovetails with a lot of cool work on overestimation of size of small demographic groups https://t.…

RT @profcikara: Really cool paper! Also dovetails with a lot of cool work on overestimation of size of small demographic groups https://t.…

RT @profcikara: Really cool paper! Also dovetails with a lot of cool work on overestimation of size of small demographic groups https://t.…

Really cool paper! Also dovetails with a lot of cool work on overestimation of size of small demographic groups https://t.co/dzG7SHlJJ4

RT @NoahCarl90: A lot of people have been sharing this graph from a recent YouGov survey, and wondering why people's guesses are so far off…

I sometimes see this tendency in my psychophysics data for effort so I wondered if this was really due to the general way people perceive proportions despite being perfect observers (https://t.co/StWSsFPTlC) or really reflected some inaccuracy...

@JSEllenberg Awesome paper about the idea at least part of the reason why these overestimates are high is because of a general feature of how we estimate stuff: https://t.co/dI64B9sead

@HC4N Read the linked research study — you would appear to be exactly correct! https://t.co/LzcErNXGLK

@UlrikeHeiss The argument is that this mid-estimation curve is a function of cognition (how people make estimates), and not if culture — so we would expect Germans to be similarly inaccurate. Be interesting if they weren’t, certainly! https://t.co/LzcErNX

@rachelcoldicutt Mm, the argument they’re making is that it’s not really sociological in this way. It’s not an artefact of beliefs about demographics, but occurs regardless of domain:

…which finds this estimation error occurring with basically anything, even just dots (!) https://t.co/LzcErNXGLK https://t.co/vfWnFt5ed4

Finding this estimation gap “Consistent with prior research” links to this 2017 study: https://t.co/LzcErNXGLK

RT @NoahCarl90: A lot of people have been sharing this graph from a recent YouGov survey, and wondering why people's guesses are so far off…

RT @NoahCarl90: A lot of people have been sharing this graph from a recent YouGov survey, and wondering why people's guesses are so far off…

@salonium There's research to support something like what you're suggesting: https://t.co/dI64B9sead

RT @NoahCarl90: A lot of people have been sharing this graph from a recent YouGov survey, and wondering why people's guesses are so far off…

RT @NoahCarl90: A lot of people have been sharing this graph from a recent YouGov survey, and wondering why people's guesses are so far off…

RT @NoahCarl90: A lot of people have been sharing this graph from a recent YouGov survey, and wondering why people's guesses are so far off…

RT @NoahCarl90: A lot of people have been sharing this graph from a recent YouGov survey, and wondering why people's guesses are so far off…

RT @NoahCarl90: A lot of people have been sharing this graph from a recent YouGov survey, and wondering why people's guesses are so far off…

RT @NoahCarl90: A lot of people have been sharing this graph from a recent YouGov survey, and wondering why people's guesses are so far off…

@DavideTomasello @real_fabristol In effetti proprio a proposito di questo oggi leggevo un articolo secondo il quale sembra che sì, forse non siamo bravi a stimare le proporzioni! https://t.co/QmyYIDvjj4

RT @jakehofman: @eredmil1 @JessicaHullman i’d defer to @DrDavidLandy’s work on this one, my read of which is that a plausible explanation i…

I liked because I didn't know the study and found the results interesting. I had no idea of the uncautious interpretations of this study that got out in public. Turns out that this article analyzing the results is much more vital than I thought. This isn't

@eredmil1 @JessicaHullman i’d defer to @DrDavidLandy’s work on this one, my read of which is that a plausible explanation is this is about estimation error in general, less so specifically about biases around demographics. https://t.co/ZHGxV9p0tu

RT @NoahCarl90: A lot of people have been sharing this graph from a recent YouGov survey, and wondering why people's guesses are so far off…

RT @NoahCarl90: A lot of people have been sharing this graph from a recent YouGov survey, and wondering why people's guesses are so far off…

RT @NoahCarl90: A lot of people have been sharing this graph from a recent YouGov survey, and wondering why people's guesses are so far off…

RT @NoahCarl90: A lot of people have been sharing this graph from a recent YouGov survey, and wondering why people's guesses are so far off…

A lot of people have been sharing this graph from a recent YouGov survey, and wondering why people's guesses are so far off. The phenomenon, termed "uncertainty-based rescaling", is explained in this paper by David Landy and colleagues: https://t.co/eoxgfX

ok i read the paper that yougov linked in their article, turns out this is actually is exactly what you'd expect. Also learned 'log odds' are a thing. What's interesting then are the estimates that are far from the expected inward bias, like gun ownership

RT @BrianMGuay: In our past work, we show that when you ask people to estimate the size of a group (e.g., of population that is foreign bor…

In our past work, we show that when you ask people to estimate the size of a group (e.g., of population that is foreign born or black), they overestimate the size of smaller groups and underestimate the size of larger groups. https://t.co/jpdgxdqUvP

@Leesplez Yeap, these are almost perfectly in line with the pattern we find. Another one of our papers on this here: https://t.co/jpdgxdqUvP

@AlecStapp @AlecStapp These patterns match what you'd expect if people are just estimating percentages in the way they normally do. i.e., the errors people make here are not specific to the quantities being estimated https://t.co/tDnINjcb3m https://t.co

Fascinating. Certainly evident in perceptions of risk. Gigerenzer warned about this 29 years ago

@EPoe187 Yes, but also there is the fact that humans estimate demographics in problematic ways. @NoahCarl90 has talked about it before. See this study. https://t.co/Ggv1YRovgB

This suggests impressively accurate perceptions, given peoples' tendency to over-estimate even the most benign small proportions. (see @BrianMGuay on this: https://t.co/Jhd1aiZMyV)

@rckhff @roderickgraham @JoeFeagin Here is a far more parsimonious and plausible account https://t.co/7F80EQRqph

This is a nice piece. https://t.co/m2kiLz2yt2

A great paper on the subject. There's nothing Covid-specific about this result. https://t.co/6DBevQXy0Q

RT @chriscurtis94: The public always overestimate low prevalence things https://t.co/3bOU0EY3go https://t.co/O3LhuEqZfI

The public always overestimate low prevalence things https://t.co/3bOU0EY3go

@seriousstats @JeffRouder So we aren't making any strong claims about the mind 'being Bayesian', but we *do* use an actual instantiated Bayesian updating model to capture the phenomena. It works well (e.g., here: https://t.co/Ub3x4g3TRt)

@jhkirkland1 Neat! Remember to take into account the fact that people systematically overestimate small proportions and underestimate larger ones, even when they have perfect underlying information. https://t.co/jpdgxdqUvP

RT @NoahCarl90: @jessesingal Hi Jesse, this paper provides an explanation for people's tendency to overestimate minority population shares:…

RT @NoahCarl90: @jessesingal Hi Jesse, this paper provides an explanation for people's tendency to overestimate minority population shares:…

@jessesingal Hi Jesse, this paper provides an explanation for people's tendency to overestimate minority population shares: https://t.co/eoxgfXkqSO

Really good paper, recommend it.

For a detailed discussion of uncertainty-based rescaling, see this paper by Landy and colleagues: https://t.co/eoxgfXkqSO

@StephenRowe0 I haven't. But other research *has* https://t.co/7F80EQRqph

@ZachG932 There's a very interesting paper on this: https://t.co/eoxgfXkqSO

RT @BrendanNyhan: @AdamBerinsky @justingrimmer I recently learned this is a general phenomenon that goes beyond the recent papers on overes…

@AdamBerinsky @justingrimmer I recently learned this is a general phenomenon that goes beyond the recent papers on overestimates of minority group pops. within country or parties - see https://t.co/UO52HpakWO: "Small values are overestimated, and large val

RT @KirkegaardEmil: @AlexStevensKent @GoodwinMJ Using small samples is not uncommon when doing aggregate level analyses, where measurement…

@AlexStevensKent @GoodwinMJ Using small samples is not uncommon when doing aggregate level analyses, where measurement error is lower. See this paper, for example (Fig 2C, n = 20): https://t.co/Ggv1YRovgB

@C_Kavanagh @PsychRabble There's some evidence suggesting that estimates of demographic proportions are actually a domain-general (i.e. unrelated to prejudice) phenomenon. https://t.co/xKM8SAdMLH

@FiveThirtyEight I'm a cognitive psychologist who studies numeracy (https://t.co/jGDKJpfu6d). As usual, 538 is reporting results without understanding the basic psychophysics in the situation. The results are therefore massively misunderstood and overblo

RT @ZachG932: @RyonCobb @JoeFeagin I also recommend you read this article https://t.co/7F80EQRqph. The overestimation of demographic groups…

@RyonCobb @JoeFeagin All the data presented in this article are inconsistent with Joe's claim https://t.co/7F80EQRqph. And you still haven't answered my question. You're instead making 'appeals to authority'.

@RyonCobb @JoeFeagin I also recommend you read this article https://t.co/7F80EQRqph. The overestimation of demographic groups is a pervasive cognitive phenomenon, and there's nothing uniquely 'white' about it.

@JoeFeagin You're missing the point Joe. Whites are not unique in overestimating the size of the black population (let alone that of other social groups https://t.co/7F80EQRqph).

"Les Belges surestiment fortement le nombre de musulmans" -Soir "Connaissez-vous vrmt le pourcentage d’immigrés en Belgique ?" -RTL Un grand classique dans les médias culpabilisants. Cette étude montre q le biais de surestimation s'applique à *toutes* les

RT @KirkegaardEmil: Bias in demographic estimates is not immigrant specific, it applies to all 'salient groups' (=media promoted), probably…

RT @KirkegaardEmil: Bias in demographic estimates is not immigrant specific, it applies to all 'salient groups' (=media promoted), probably…

RT @KirkegaardEmil: Bias in demographic estimates is not immigrant specific, it applies to all 'salient groups' (=media promoted), probably…

RT @KirkegaardEmil: Bias in demographic estimates is not immigrant specific, it applies to all 'salient groups' (=media promoted), probably…

RT @KirkegaardEmil: Bias in demographic estimates is not immigrant specific, it applies to all 'salient groups' (=media promoted), probably…

Bias in demographic estimates is not immigrant specific, it applies to all 'salient groups' (=media promoted), probably due to availability heuristic. https://t.co/Ggv1YRovgB https://t.co/Emjh2qTGaY