Overcoming Bias

Share this post

Types of Partiality

www.overcomingbias.com

Types of Partiality

Robin Hanson
Apr 15, 2023
19
7
Share
Share this post

Types of Partiality

www.overcomingbias.com

Yesterday I did a set of nine Twitter polls, each time asking respondents on which side of particular divide (e.g. tall vs short) they fell, and if they try to favor their side of this divide. This table shows the % on each side, and what fraction of each side said they try to favor their side. (These are sorted from a small to big favoring ratio.) 

I see four types of divides here. First there are “meh” divides about which few are partial, and the two sides are similarly partial. Such as on tall vs short (where ~23% are partial). I expect there are a great many more divides on which I’d get similar results. 

Next are “fighting” divides where people are more partial to their side, yet both sides are similarly partial. For example, gender (~29%), politics (~57%) and humans (~80%). Note that the gender fight is relatively mild.

Third, I see “up-down” divides where only one side, such as youth, introverts, poor, shape rotators, has an elevated (43-47%) partiality. These correspond to common perceptions that one side is “down”, and is allowed to “punch up”, while the other “up” side is not supposed to “punch down”. 

That leaves one last divide, race, which is usually discussed as if were an up-down sort of divide. Here the usual “up” side keeps the usually low partiality, but now the “down” side says that it has an even lower, record low in fact, partiality! Not sure what is going on here; seems worth a deeper dive.   

Added 16Apr: I checked on the odd race results with a new set of polls, polls that also collect some data on when people disfavor their side of a divide. This time the results made sense in terms of race as an up-down divide:

19
7
Share
Share this post

Types of Partiality

www.overcomingbias.com
7 Comments
TGGP
Apr 15

I didn't vote in this poll, and I'm curious about the non-human respondents (and specifically how many of them are saucer-flying reptilians from deep within the hollow earth). I expect that your non-white respondents are highly unrepresentative of the US population.

Expand full comment
Reply
1 reply by Robin Hanson
Arby
Apr 15

94pct male skew could be impacting responses to other questions too no?

As for race, a) subsplit of non white (eg black vs asian) is likely to affect response and b) i suspect it's also the question where people are most likely to vote their opinions on what they "should" think rather than what they actually think

Expand full comment
Reply
5 more comments…
Top
New
Community

No posts

Ready for more?

© 2023 Robin Hanson
Privacy ∙ Terms ∙ Collection notice
Start WritingGet the app
Substack is the home for great writing