32_undue_hate.md

Undue Hate

Errors

• Overestimate share of disliked groups, % holding disliked beliefs, % exhibiting bad behaviors, extremity of attitudes, ...

  • False Polarization --- pg. 35 + also ahler, malhotra etc.
  • Share of disliked groups ~ ahler and sood
  • Democrats and Republicans assumed that the levels of prejudice and dehumanization held by the other side (i.e., meta-prejudice and meta-dehumanization) were 50–300% [!] higher than what was actually expressed by a representative sample of outgroup political partisans. (Moore-Berg et al. 2020)
  • cor(bias, polarization) ~ fairly robust evidence
  • Nuance
    • Dimant asked subjects about their feelings about President Trump and classified subjects as Trump “lovers,” Trump “haters,” or indifferent. The haters contributed, on average, just under $5 to the public good when told they were paired with Trump lovers and a bit over $6.50 when paired with fellow haters. Trump lovers contributed around $5 when paired with hat- ers and just over $6 when paired with fellow lovers. Once again, nearly identical behavior was observed across the two groups, and this behavior was substantially prosocial even when subjects were paired with political opponents.
    • Trump lovers’ guesses were close, on average, to the actual contributions made by Trump haters when they were paired together. Trump haters’ guesses were too low when paired with lovers: haters guessed lovers would contribute under $4 on average (again, the actual average was $5), consistent with biased dislike.
    • Tappin and McKay found that, on average, choices were nearly identical across the two parties and most participants in both parties played collective interest, cutting their own payoff in half to help all of their group mates, even the out-partisans. For Democrats, 60 percent played collective interest, while only 26 percent played self-interest, and 14 percent played out-party hostility. For Republicans, 59 percent played collective interest, 26 percent played self-interest, and just 15 percent played out-party hostility. Republicans were actually more optimistic about Democrats playing collective interest than about their copartisans doing so: on average, Republicans guessed 43 percent of Democrats would choose this action and just 37 percent of other Republicans would.
    • Whitt and colleagues found that, on average, Democrats actually over- estimated Republican contributions (Democrats expected $2.67 on aver- age, while Republicans only gave $2.35). Republicans did underestimate Democratic contributions, receiving $2.91 on average from them but only expecting $2.43.
      • https://gojiberries.io/2020/07/31/gaming-measurement-using-economic-games-to-measure-discrimination/
      • Democratic respondents, on average, guessed 58.8 percent of Democrats would make the copartisan sacrifice to save the five out-partisans but that only 48.2 percent of Republicans would do this (versus a true value of 60 per- cent). Republican respondents guessed around 52.5 percent of Republicans would make the copartisan sacrifice but that just 48.6 percent Democrats would do this (versus a true value of 58 percent).

• Why?

  • Infer from Noisy Signal

    • We think people with “good” character traits are more likely to take “good” actions or hold “good” opinions in the future. And we feel warmer toward those people because we then expect them to be more likely to do good things going forward.
      • We can never perfectly understand ourselves, much less other people, about whom we have orders of magnitude less information.

  • Motivated reasoning

    • Motivated skepticism of claims/confirmation bias
      • We’re also too credulous of congenial information that’s uncer- tain or unverified.
        • why?
          • wired to ignore inconsistent data. chris chabris, etc.
          • see above
    • ultimate attribution error source: Hewstone’s metastudy implies that this tendency is “motivatedly” asymmetric: we attribute positive acts by in-groups to character traits and negative acts to bad luck and vice versa for out-groups.
    • Infer cynical motives: Politics is the art of making your selfish desires seem like the national interest.

  • Why?

    • make us feel good about ourselves and our identities
    • in-party is more likely to prevail in future elections and policy choices, which is a pleasant thought
      • Hostility toward the out-party can benefit the in-party via improved political motivation and engagement.
      • signal ... loyalty and contributions to the in-group
      • also the group is biased so exploiting those biases for your gain
      • helps you win arguments

  • WYSIATI what you see is all there is (whiz-ee-ah-tee) (Kahneman)

    • that we fail to think carefully about the information we don’t see and reality tends to be more complicated than it seems
      • correspondence bias (overinference about a person’s traits based on observed actions; see Ross and Nisbett, 1991

  • Tastes

    • mistake horizontal for vertical product (moral) characteristics
    • Naive realism can also make us overestimate the objectivity of our prefer- ences. If we think that if we like salads better than burgers, then salads must simply be (objectively) better than burgers, then we’ll think that people who choose burgers are simply wrong and not just different.
    • moral foundations as 'taste buds': fairness, care, loyalty, authority, and purity. Liberals value the first two higher than the others, and conservatives place approximately equal weight on all five.

  • False Consensus

    • Either way, overestimating the underlying agree- ment logically causes overinference of the chance of the other side having an ulterior motive for claiming to disagree.

  • Macro factors like technological change --- product differentiation, etc., geographic sorting

    • Media's incentives to highlight conflict
      • People can't guess (see Ben Enke)
    • I’ll follow Bail (2021) and Kim’s team (2021) and call this tendency to observe out-partisans online behaving in a misleadingly unlikable way the social media prism. (When we see copartisans express extreme opinions and act belligerently online, we don’t mind it so much and maybe even like them for it, a phenomenon called political acrophily; see Clark, 2021.)
    • The correlation between spouses’ party-feeling thermometer scores increased from 0.39 in 1965 to 0.77 in 2015. For parents’ and their children’s scores, this correlation increased from 0.20 to 0.64 over the same period (Iyengar et al., 2018).
    • Nuance
      • The paper’s main results were that ideological segregation was higher for online media than for traditional media, but not by much, and ideological segregation for both was quite low as compared to for offline interactions.

  • Unmotivated factors

    • Limited strategic thinking
    • Selection neglect
      • why we are unduly influenced by the social media prism
    • Correlation Neglect
    • Representativeness, Availability (plus media)
    • mere exposure effect: prefer content and ideas that we are exposed to repeatedly
    • illusory truth effect makes us more likely to believe false content when we hear it repeatedly