COMPUTATIONAL SOCIAL SCIENCE -The ‘Hidden Trump Model’: Modeling social desirability bias through ABMs – Stephen Davis & Hannah Zontine
COMPUTATIONAL SOCIAL SCIENCE FRIDAY SEMINAR
Department of Computer Science
University of Mary Washington
The ‘Hidden Trump Model’: Modeling social desirability bias through ABMs
Social desirability bias is a tendency people have to lie about their opinions if they perceive they will be judged or rejected. We present an Opinion Dynamics model in which agents may not be truthful about their opinions when they interact with their social circle. We model two processes through which agents influence one another: an online anonymous process in which agents can interact with anyone and do not fear social rejection, and a face-to-face process where they interact only with friends and may feel compelled to conform. In a political setting, this would apply to a race in which one of the candidates bears a social stigma and therefore some agents are reluctant to voice support for him or her. The results that these nonlinear and asymmetrical processes will have on the overall electorate are not obvious, and are well-suited to an agent-based study.
We hypothesize that this model will produce a “poll bias” of the kind we saw in the 2016 Presidential election — i.e., a significant difference between the number of agents who say they will vote for a candidate and the number who actually do so on election day. We present an analysis of this “Hidden Trump model” and describe the way in which poll bias depends on the strength of the various interaction processes.