Self-selection and attrition biases in app-based persuasive technologies for mobility behavior change: Evidence from a Swiss case study. (December 2021)

• Record Type: Journal Article
• Authors: Cellina, Francesca; Vittucci Marzetti, Giuseppe; Gui, Marco
• Is Part Of: Computers in human behavior. Volume 125(2021)
• Publication Date: 2021-12
• Subjects: Persuasive technology -- Mobility behavior -- Self-selection -- Attrition -- App churn -- Behavior change; Interactive computer systems -- Periodicals; Man-machine systems -- Periodicals; 004.019
• DOI: 10.1016/j.chb.2021.106970
• Languages: English
• ISSNs: 0747-5632

Abstract

Abstract: App-based persuasive technologies emerged as promising tools to promote sustainable travel behavior. However, the opt-in, self-selection framework characterizing their use in real-life conditions might actually lead to wrongly estimate their potential and actual impact in analyses that do not rely on strict randomized controlled trials (RCTs). To investigate evidence of such biases, we analyze mobility data gathered from users of a persuasive app promoting public transport and active mobility launched in 2018 in Bellinzona (Switzerland). We consider the users' baseline mobility data: km per day (total and by car) traveled during the app validation period, when behavior change motivational features were not enabled. To estimate the possible self-selection bias, we compare these data with the reference population, using data from the Swiss Mobility and Transport Census; to study the possible attrition bias, we look at the relations between baseline mobility and the number of weeks of app's active use. We find evidence of neither self-selection nor critical attrition biases. This strengthens findings by earlier non RCT-based analyses and confirms the relevance of app-based persuasive technologies for mobility behavior change. Highlights: App-based persuasive technologies for behavior change are increasingly exploited to promote sustainable mobility. Self-selection and attrition due to the opt-in framework may affect their actual impact and bias the estimate of this impact. We analyze mobility data collected in an app-based behavior change intervention run in Southern Switzerland. Comparison with census data on mobility and analysis of app use over time show no evidence of critical biases. Further research is worth analyzing the app's actual impact and the reasons for the observed high drop-out rate.