I recently joined a neuroscience/cog sci/philosophy of mind project on social influence… This post will probably become an amalgamation of notes that I jot down in my notebook, digitally organized. This post, like all my posts, is just for me, but if anyone else finds it interesting or informative, that would of course make me happy 🙂
Some Notes
It is generally agreed upon that external input can be very helpful whe one is presented with uncertainty in a decision one must make. One external input type that is understood to be very helpful: other people’s beliefs. There can be cons to incorporating other people’s beliefs, e.g. groupthink, herding, information cascades, failure to explore sufficiently, suppression of useful information. Thus, it is considered to be important to be able to balance the pros and cons of incorporating other people’s beliefs into one’s decision making.
One factor that affects the value of other people’s beliefs with regards to impacting one’s decision making is their informational independence. I.e. Having ten sources which provide you with reason to do X is not better (in fact could be worse) than having one source provide you with reason to do X if the ten sources get their information in favor of X from one common source.
We would like to know if people are sensitive to informational independence when incorporating other people’s beliefs into their decision making. Some studies show that people are sensitive to informational independence (Mercier & Miton, 2019) while other studies show that people are not sensitive to informational independence (Yousif, Aboody, and Keil, 2019).
Both studies have flaws. I choose not to take time to spell out these flaws, and instead note conceptually important things for the rest of these notes:
– Assumptions about neutrality or source expertise may compete with source independence in evaluating social information.
– There is a difference (obvi) between measuring beliefs and measuring confidence in beliefs. Good to think of both.
– The process of beliefs being affected is a point of interest, thus only measuring initial and final beliefs or confidence in beliefs seems inadequate; instead, we would like to do a multi-stage approach to measuring beliefs or confidence in beliefs.
– There are stable individual differences in people’s tendency to make use of social information, and this tendency can also be context-dependent.
– There may be ‘evolutionary valid cues’ which are cues that human beings’ brains have evolved to attend to and process; some people think heresay could be one such cue.
– Some recent work shows that the ‘friend of a friend’ attribution is a strong factor in social transmission of information (Altay, Claidiere, & Mercier, 2020).
Some hypotheses we are looking to test:
(H1) If people’s evaluation of social cues is sensitive to informational dependency, participants will adjust their estimates more when sources of social information are diverse than when they are homogeneous;
(H2) Participants will be sensitive to whether sources are intentionally selected vs. randomly assigned;
(H3) The effect of informational dependency will be larger for people’s likelihood ratings than for their confidence ratings;
(H4) By including a measure of baseline social information use as a covariate in a regression, we will be better able to detect an effect of condition.
Interesting background:
- Kahneman and Tversky (1972) showed that people do not typically incorporate sample size into their judgments of probability.
