Fwd: PDF - www.pnas.org

Ashley N. D. Meyer, PhD ashley.meyer2 at VA.GOV
Fri Aug 26 13:59:20 UTC 2016

Very interesting article and great summary Linda.

I think the findings are interesting: that similarity in accuracy of the group is key for the advantage of collective intelligence over either the group member's decision with the most confidence (the confidence rule) or the group's majority decision (the majority rule).

However, as Linda points out, the similarity in accuracy that is key for the collective intelligence advantage happens less often than the dissimilarity in accuracy that would prove disadvantageous for collective intelligence (in Figure 2, it looks like ~2/3 of the time collective intelligence becomes a disadvantage over the other methods- i.e., when similarity is less than .1). Since you will likely not know what the similarity of accuracy within a group is in the real-world, however, it might be better to go with either the confidence rule or the majority rule (the authors also found that both of these rules consistently outperformed the average individual performance within a group regardless of similarity of accuracy amongst the group members). A rule that does better than the average physician might be more amenable to real-world implementation.

Additionally, I wonder if it might be the boundary conditions (i.e., the difficult cases or the cases in which there is a lot of disagreement between physicians) that we might want to focus on more for the use of interventions like collective intelligence. I wonder how many cases in this bunch were relatively easy and devoid of overall disagreement. Those results might paint a different picture.

Ashley N.D. Meyer

Moderator: David Meyers, Board Member, Society to Improve Diagnosis in Medicine

More information about the Test mailing list