"so for each belief is equivalent to a simple contagion model in all respects other than the influence of the individual’s internal state.
[ ] Change "individual's" to "potential adoper's" or something that emphasizes we are talking about the individual potentially adopting the belief.
This is quite remarkable, as post-hoc, individuals have both internal support for their beliefs (i.e. each belief is supported by many other beliefs) and external support (i.e. other individuals share both their beliefs and their justifications for believing so). This happens even though a priori we have no way to tell which sets of beliefs will become popular, and no ground truth.
[ ] Want to emphasize that there is support for some beliefs but not others - ie, differential, coordinated support for subsets f beliefs.
in which whole groups of people can collectively deceive themselves
[ ] Emphasize that this is coordinated between individuals, and that is why it is different from normal confirmation bias.
B.
Correlation between the initial number of susceptible individuals and the popularity of
each belief, given the same final distribution of adoption between conditions; detail
shows t = 9 popularity vs. number of individuals initially susceptible. C. Correlation
between the popularity of each belief and the popularity of the most popular belief it
shares a ’node’ with; sample t = 9 popularity of each belief overlaid on the knowledge
graph. D. Clustering coefficient of a hypothetical knowledge graph made up of the
most popular 10% of all beliefs; sample t = 9 network filtered on popularity.
[ ] clarify that after semicolon in each section refers to the inset. Maybe put them in parentheses?
By contrast, under interdependent diffusion each belief has a particular likelihood of adoption...
[ ] really, it's under this model. Interdependent diffusion can also happen when homophily is active, they are just separate mechanisms.
[ ] Fig 6 text spills outside of margins on "social network effec(t)", redo figure size
[ ] Fig 3 description in methods should discuss using phi coefficient as measure of similarity between individuals.
[ ] fig 1c2: put arguments on new line, spaces around equals sign for consistency with other rows
[ ] Fig 6: caption doesn't follow the same format as the other captions (A. B. C.)
[ ] really, it's under this model. Interdependent diffusion can also happen when homophily is active, they are just separate mechanisms.
[ ] Fig 6 text spills outside of margins on "social network effec(t)", redo figure size
[ ] Fig 3 description in methods should discuss using phi coefficient as measure of similarity between individuals.
[ ] fig 1c2: put arguments on new line, spaces around equals sign for consistency with other rows
[ ] Fig 6: caption doesn't follow the same format as the other captions (A. B. C.)