Peter Ebbes from HEC Paris held a talk on how to identify the information content of research subjects as part of our Department’s Research Seminar Series. To overcome the problem of dishonest or inattentive respondents in experimental online studies (such as in Choice Based Conjoint Tasks) the presented research introduces a mixture modeling framework which clusters subjects based on variances in a choice based setting (multinomial logit models). This model naturally groups subjects based on the internal consistency of their answers, where Peter and his colleagues argue that a higher level of internal consistence (hence lower variance) reflects more engaged consumers who have sufficient experience with the product category and choice task, to have well-formed utilities. This approach provides an automated way of determining which consumers are relevant and is argued to be more effective in identifying “Gremlins” in the data than previous approaches that have been proposed in the literature. We thank Peter for his visit and his insightful talk and discussions. For information on the upcoming guest speakers at the Research Seminar Series, please check the agenda.