How Do People Process Information from Automated Decision Aids: an Application of Systems Factorial Technology

While many researchers have investigated the performance consequences of automated recommender systems, little research has measured how these recommendations can impact the user's decision-making process. In the present work, we measured how people process information when provided with an automated recommender system using the Systems Factorial Technology (SFT) framework. This research comprises two experiments that explore the circumstances in which people use one or all available information (Experiment 1) and process information serially or in parallel (Experiment 2). For each experiment, participants completed a speeded length judgment task with a reliable but imperfect aid. Participants demonstrated serial processing of information and likely used only one source of information when making decisions across all conditions. Integrating information on the display and accurate training were shown to lead to more efficient information processing. Display characteristics, performance incentives, and training play a role in how people use information from the automated aids which may lead to slow downs or speed ups in information processing. This research sheds light on how people gather and process information with an automation aid and suggests how we might design systems to improve decision performance.