A Look into How a Human-In-The-Loop Experience May Alleviate Algorithm Aversion
To study the phenomenon of algorithm aversion, this study incorporated a human-in-the loop experience to observe if this element would encourage model usage further. This was done by asking participants to sort information as either “least important” or more important” based on cases in which they were asked to make forecasts. A 3 × 2 design was utilized with three domains (credit card approval, happiness ranking, and university ranking) and two conditions. Those in the control condition had the option of viewing a generic model, and those in the experimental group had the option of viewing a model which was tailored based on their sorting of variables. Participants were asked to forecast the outcome of scenarios in each domain and then report their confidence in their decision. Results found a significant effect of domain in frequency of viewing and following the model, as well as significant differences in confidence between domains. An unexpected opposite effect was found for condition with those in the control group both viewing the model and following it more frequently when compared to the experimental group. The results suggest that there is further research to be done on what may be done to alleviate algorithm aversion.
Quantitative psychology|Clinical psychology
Hussein, Yasmin, "A Look into How a Human-In-The-Loop Experience May Alleviate Algorithm Aversion" (2021). ETD Collection for Fordham University. AAI28715616.