Displayed judgment certainty reorganizes source selection in self-formed networks

Collective intelligence in self-formed networks depends on which judgments people choose to observe, as well as on how they integrate those judgments. We tested whether displaying judgment certainty reorganizes source selection and improves subsequent accuracy. Across eight online sessions, 273 participants evaluated 14 true and false news headlines, formed costly directed observation networks, and revised their beliefs over five updates. During network formation, we manipulated the visibility of certainty, defined by the extremity of each source's private probability judgment, while belief direction and source accuracy remained hidden. Displaying certainty produced denser networks and redirected links toward more certain sources; current popularity also predicted selection. Visibility increased exposure to highly certain but incorrect judgments, without a clear improvement in the relative accuracy of selected sources. Accuracy increased across repeated judgments, but certainty visibility did not detectably increase this improvement and reduced reconstructed net scores under the task's payoff structure. Observations with selected sources showed greater improvement than those without, and improvement was associated with selected sources' initial accuracy. Belief revisions moved toward selected sources and decreased with participants' own certainty. Among six compact predictive models, certainty-reinforcing DeGroot provided the best penalized and aggregate held-out prediction, although it did not capture the large first update. These findings distinguish a cue's ability to attract attention from its ability to identify useful information: displayed certainty can reorganize costly information acquisition without a detectable average accuracy benefit.