16/09/2021 (Thursday) 13:00-14:00 E21B-G002

Taming human subjects: Reducing variation in behavioral experiments

The experimental method is often touted as the gold standard of scientific inquiry based on its ability to enable causal attribution. Valid causal attribution, however, can only be obtained through meaningful comparison across conditions, and this requiresadequate control and filtering of noise. This paper argues that control and filtering pose particularly significant challenges for the behavioral and social sciences because these disciplines are characterized by within-category variation, fuzzy category boundaries, and complex interactions. Through analysis of the experimental literatures on psychology and economics as well as related methodological discussions, this paper demonstrates three common filtering strategies for coping with those difficulties. The analysis shows that, although these filtering strategies work to some extent, they are unwieldy tools and the filters they create around experimental systems tend to be either not restrictive enough or so restrictive that the results cannot reveal much about the phenomenon of interest. This lack of precision precludes unequivocal causal attribution and leaves the door open to constant contestation, making consensus hard to emerge.


Carol Ting is assistant professor at the Department of Communication at the University of Macau. Her research interest is in the social aspects of quantitative research methodology.