[...] Skimming through the paper (particularly Figure 3b), seems like this optimal level of disagreement is roughly 20%.
"We started with the basic tools of probability theory to model how an individual would make the best decision given some information (both private and social). This is an idealized version of reality in which individuals are not assumed to be cognitively biased. That idealization is, however, useful because it gives a clear baseline to which we can add specific biases, such as conformity—the tendency to align one's judgment with a group."
[...] "From this relationship, we computed the probability that accuracy declined as agreement increased," explained Boussard. "We term this phenomenon 'credibility-enhancing dissent.' This is a pattern in which groups that appear more confident are in fact less trustworthy.
[...]
The team's findings suggest that when conformity bias is absent, the accuracy and trustworthiness of group decisions steadily increase along with agreement between members. However, when this bias is considered, unanimity or high levels of agreement are no longer reliable indicators of trustworthiness and decision quality.
Skimming through the paper (particularly Figure 3b), seems like this optimal level of disagreement is roughly 20%.
[...]
Skimming through the paper (particularly Figure 3b), seems like this optimal level of disagreement is roughly 20%.