Conformity and confirmation: group decision-making in foreign policy

World leaders typically do not make history-altering foreign policy decisions on their own. They evaluate options and set foreign policy in consultation with small groups of experts and advisors. In this dissertation, I explore the effects that groups of advisors have on foreign policy decision-making. We lack understanding about how biases among advisors can affect group decisions. To this end, I study small group decision-making through individuals who display two types of biases that are common among elite, small group decision-makers: confirmation and conformity. Confirmation bias is the tendency to interpret new information to support one's original beliefs and conformity bias is the tendency to seek out an agreement within the group. Both biases are well-documented among elite decision-making groups. Often considered pathologies, I explore when, within a group context, these prominent biases help or hinder foreign policy decision-making. I argue that conformity and confirmation biases lead to low-quality group decisions when alone, the presence of these biases together can produce high-quality group decisions. Advisors make high-quality decisions because they are more willing to follow others’ options due to conformity, but only willing to consider options that are clearly better than their initial positions due to confirmation. This tug and pull creates situations where a subset of advisors can find high-quality decisions and bring the rest of the group with them. These tendencies then can increase the probability of finding high-quality decisions. This study contains three distinct parts. First, I build a computational model to capture the dynamic interactions that occur in group decision-making. Designed to reflect a discussion among foreign policy advisors, the model inputs conformity and confirmation biases at the individual-level and allows individuals to consider different positions until they find positions that are satisfactory to them. Group decisions are then produced using aggregation rules. Simulation results show that groups with a mix of conformity and confirmation biases can lead to higher quality decisions and to more information search among agents compared to groups with stronger tendencies for conformity or confirmation. Second, I use small group laboratory experiments as an application of the computational model. This experiment asks groups of students to discuss and decide on U.S. climate policy after they are given incentives for conformity, confirmation, or a mix of conformity and confirmation. Results from the small group experiment show that groups treated with conformity reach decisions more quickly than both confirmation treatment and equal treatment groups; groups treated with conformity have more consensus, and there is no detectable difference for decision quality between the different groups. These results also vary depending on whether there are 3 students or 4 students in a group. Third, I test the implications of the model empirically using text analysis of historical cases. I collect an original dataset of U.S. National Security Council conversations scraped from transcripts during the Nixon and Ford Administrations. I develop speaker-level and group-level measures for conformity and confirmation using phrase repetition among speakers; I also develop measures for information search using questions and briefings in conversations. With a loess curve fit, I find a non-linear relationship for one of the conformity and confirmation measures, where conversations with slightly more conformity than confirmation have more information search compared to conversations that are only high in confirmation or only high in conformity. However, this fit is indistinguishable from a linear fit. Exploring the speaker-level data, I find that possible conversation dynamics, such as the lack of information search when Nixon was present or the tendency for more information search, higher conformity, and higher confirmation when Kissinger was present, could help explain why the non-linear relationship seems weak. Together, this research suggests that a mix of confirmation and conformity biases may be beneficial in foreign policy decision-making.U of I OnlyAuthor requested U of Illinois access only (OA after 2yrs) in Vireo ETD syste