Passive or active? Behavioral changes in different designs of search experiments

While search experiments are available in several designs, accumulating ex- perimental evidence suggests that individual search behavior depends on design details. This paper reports the first classification and comparison of several search experiment designs widely accepted in search studies. These designs can be categorized as passive, quasi-active, and active. We found individual- and aggregate- level significant differences in the results across designs, despite identical models. In the passive design, subjects tended to be more reluctant to search than in the active design, and risk-averse subjects quickly terminated their search. Our results highlight the importance and potentials of designing search environments in practice.