The Detectability of Socio-Egocentric Group Speech: A Quasi-Turing Test
Scholars interested in group phenomena generally conceive of communication as either a conduit for, or as constitutive of, group decisions. Hewes's socio-egocentric model contends that we possess no unambiguous proof of any communicative impact on decision making. This study asks whether contrived socio-egocentric group speech is distinguishable from real group speech, and therefore whether the socio-egocentric model is even a plausible depiction of interaction. We developed a simulation that produces socio-egocentric speech and report on its use in a study that asked naïve raters to discern human-generated from simulated socio-egocentric group discussion. Results indicated that participants correctly judged the source of transcripts at rates worse than chance. Furthermore, heuristics employed by the participants can explain their poor performance, because criteria that produced accurate judgments of human transcripts produced inaccurate ones for the simulated transcripts and vice versa. We conclude that the model must be taken seriously as a depiction of plausible group interaction, and that in future studies a distinction between local and global features of conversation is important for studying socio-egocentric interaction.