Before You Regret It

Sally Ahmed; Kye Shimizu; Jan Enkmann; Ivy Yip; Vincent Beerman; Ayse Alomar; Pattie Maes; PhD; Falk Uebernickel

Before You Regret It

Predicting Problematic Phone Sessions with In-the-Wild Wearable Data

Sally Ahmed, Kye Shimizu, Jan Enkmann, Ivy Yip, Vincent Beerman, Ayse Alomar, Pattie Maes, PhD, Falk Uebernickel

Abstract

A 7-day in-the-wild study predicting problematic phone sessions from passive smartphone logs, low-cost smartwatch physiology, and session-level surveys.

Users often feel regret after using social media, making regret a more ecologically valid target than screen time¹ for understanding when phone use becomes problematic. Existing self-monitoring tools cannot anticipate regret before it occurs, and prior physiological work on social media use has been confined to the lab with research-grade sensors and curated content, leaving the question of in-the-wild prediction open.

We deployed a 7-day in-the-wild experience sampling study² with 21 participants, combining passive smartphone logging, a low-cost consumer smartwatch (Bangle.js 2, $80), session-level surveys (1,445 sessions), and exit interviews to investigate when and why social media sessions become regretful, and whether regret can be anticipated before a session begins.

Sessions per participant and daily session volume across the 7-day deployment.

Median session duration is 4.6x longer for social media than for non-social-media apps.

Findings

Three findings stand out. First, the gap between intended and actual use predicts regret far more strongly than session duration itself — duration’s apparent effect collapses once intention is modeled³. Second, regret is amplified when a session displaces a valued alternative, particularly at night and following productivity-app use. Third, pre-session contextual features generalize across participants while physiological signals add person-specific lift, pointing toward a two-layer architecture for just-in-time adaptive interventions⁴.

Per-participant regret by day vs. night and by predecessor app (productivity vs. other), with posterior distributions of the population mean.

Within-person ROC for predicting regret from smartwatch physiology combined with phone context (mean AUC 0.740 across 50 seeds).

Acknowledgements

This work was supported by the HPI × MIT Morningside Academy for Design Grant. We thank the 21 participants for tolerating a week of session-level surveys without selecting their phone for them, the MIT Committee on the Use of Humans as Experimental Subjects for protocol review, and our colleagues at the Fluid Interfaces group and the Hasso Plattner Institute for repeated rounds of feedback on the study design and the regret operationalization. Special thanks to the Bangle.js 2 community for keeping a hackable open-source smartwatch alive — without it the physiology arm of this study would not have been feasible.

Research Groups

Fluid Interfaces, MIT Media Lab
Media Lab Research Theme: Connected Mind + Body
Hasso Plattner Institute

Gallery

Screen time is what most operating-system dashboards and prior HCI literature optimize for, but two hours of intentional video calling with family and two hours of doomscrolling produce identical screen-time readings. Regret captures the user’s post-hoc evaluation and is far less confounded with content type and intention. ↩
Experience sampling (ESM) interleaves brief, in-the-moment surveys throughout the day to capture momentary states before they decay into reconstructed memory. We constrained surveys to immediately follow phone sessions so that participants rated specific sessions rather than aggregate impressions. ↩
This is a cautionary point for screen-time research more broadly: studies that report a relationship between duration and wellbeing without controlling for intention may be measuring intention all along. The same hour-long session is innocuous when planned and regretful when unplanned. ↩
Nahum-Shani, I., Smith, S.N., Spring, B.J., Collins, L.M., Witkiewitz, K., Tewari, A., & Murphy, S.A. (2018). Just-in-Time Adaptive Interventions (JITAIs) in mobile health: Key components and design principles for ongoing health behavior support. Annals of Behavioral Medicine, 52(6), 446–462. ↩

Credits

Sally Ahmed (MIT Media Lab) — co-first authorKye Shimizu (MIT Media Lab) — co-first authorJan Enkmann (HPI) — co-first authorIvy Yip (MIT Media Lab)Vincent Beerman (HPI)Ayse Alomar (HPI)Pattie Maes, PhD (MIT Media Lab)Falk Uebernickel (HPI)

Grants

HPI x MIT Morningside Academy for Design Grant(2024)