The Impact of Algorithmic Recommendation on Time Preference: Evidence from Laboratory and Field Experiments

How to cite

Li, Guangrui; Gong, Zheng; Xiaoquan. 2025. "The Impact of Algorithmic Recommendation on Time Preference: Evidence from Laboratory and Field Experiments." AEA RCT Registry. AEARCTR-0015949. https://doi.org/10.1257/rct.15949-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Guangrui Li* , Zheng Gong , Xiaoquan

* Primary investigator

Registry ID
AEARCTR-0015949
DOI
https://doi.org/10.1257/rct.15949-1.0
First registered
Last updated
Status
Completed
Trial dates
2021-04-28 – 2023-05-21
Countries
China ()
JEL codes
D12, D90, L82, L86, I31
Registry record
http://www.socialscienceregistry.org/trials/15949

Abstract

Algorithmic recommendation systems on short video platforms significantly impact young adults' time preferences. Our laboratory and field experiments demonstrate that deactivating personalized recommendations increases patience, with pronounced gender differences: male users become substantially more impatient under algorithmic recommendations while female users remain unaffected. This disparity stems from gendered content consumption patterns, particularly males' preference for high-stimulation content. These findings reveal how digital environments can shape fundamental economic preferences, potentially influencing important life outcomes.

Intervention

The participants in the treatment group are asked to turn off the algorithmic recommendation functions on their short video apps.

Experimental design

Lab Experiment: Lab experiments were conducted at two Chinese universities (S1 and S2), recruiting current students via an online ad offering compensation and a lottery chance. Eligible participants completed a survey reporting their video app usage and availability. From these, users of short video apps with no prior related experiment experience were selected and randomly assigned to treatment or control. Treatment participants were asked to deactivate algorithmic recommendations. All participants completed a baseline survey, browsed short video apps for a fixed time period, and then answered questions on time preferences. S2 participants provided additional data on content types and viewing behavior.

Field Experiment: We conducted four-week field experiments at two Chinese universities (L1 and L2), recruiting participants two weeks prior via a baseline survey. Eligible users who hadn’t joined similar studies were invited to participate. On Day 0, they completed a pre-experiment survey covering screen time, well-being, and time preferences. Participants were randomly assigned to treatment (disabled algorithmic recommendations) or control (default settings), respectively, Weekly and post-experiment surveys tracked usage, activities, and repeated the CTB tasks.

Design details

Randomization method
Randomization done in office by a computer
Randomization unit
Individual
Clusters
3-4 universities
Observations
Lab: around 300 participants; Field: around 200 participants
Arms
In the lab experiment, we planned to have 50% participants into the treatment and 50% into the control. For the field experiment, we planned to have 33% participants into the treatment and 66% into the control, following Allcott et al. (2020).
Primary outcomes
Time preference (patience)
Secondary outcomes
Time preference (patience) across gender groups

Cited by

No citation data available for this registration yet.

Keywords