The "dark" side of engagement in stated choice experiments: exploring the role of time and complexity with an example on food

Online stated preference surveys have become increasingly common in recent years. However, concerns about respondents' engagement and data quality remain critical. Our study proposes a robust modelling framework that addresses the role of respondents' survey engagement while addressing key limitations in existing approaches. We implement a hybrid modelling framework incorporating engagement before and during the task as latent variables on data from a food choice experiment. The models consider engagement valence and temporal aspects while accounting for the order in which the task was displayed and for individual task easiness and complexity, based on respondents' preferences. Our results show the importance of distinguishing between the time respondents spend before and during the choice tasks, as each phase has unique implications for how respondents process the information. Moreover, the time respondents spent on the initial and later tasks has a different association with engagement, confirming the role of ordering effects, experience learning and changes in decision-making strategies. Finally, we highlight the key roles of easiness and task complexity in engagement. Such results offer insight into how task design can influence respondent scale heterogeneity and data quality.