Collective Wisdom and the Fermi Method: Improving the Accuracy of Deliberative Groups

LESCANO-CHARREAU, VICTORIA; BARRERA LEMARCHAND, FEDERICO; RUIZ, JULIETA; CÁCERES, NURIA; CARRILLO, FACUNDO; NAVAJAS, JOAQUIN

Congreso; 9th International Conference on Computational Social Science IC2S2; 2023

Understanding the conditions under which groups of people perform better than independent individuals is a fundamental challenge in the social and psychological sciences.(1,2) A central phenomenon in this field is the "wisdom of crowds" (the aggregate of independent estimates can be more accurate than the best individual judgment),(3,4) used in various areas, including politics(5), health(6), and business(7). While previous research suggests that group deliberation improves the accuracy of crowds, the mechanisms behind this remain unclear.(8) This study addresses this question by analyzing the use of the "Fermi method" in groups (breaking down complex questions into simpler ones and combining their solutions to reach the final answer).(9) A first study with 130 groups of 4 participants interacting in chat rooms (N=520) confirmed that group estimates are more accurate than individual ones, especially in those groups that use the Fermi method. This result was obtained by analyzing written discussions, both with human evaluators and automated methods. A second study (N=240) provided causal evidence of this phenomenon by explicitly requesting that some groups use the Fermi method, while asking others to employ estimation aggregation strategies. A third study (N=160) showed that collectively employing the Fermi method leads to superior improvement compared to using it individually. In summary, these results provide causal evidence that collectively employing the Fermi method improves group estimates, promoting its application in various domains, through either explicit instructions or automated detection methods