Algorithmic Platform Management and Risk-Taking Behavior among Chinese Food Delivery Riders
Description
1. Research Overview This dataset explores how algorithmic platform management shapes food delivery riders’ risk-taking behavior (e.g., traffic violations). Based on Cognitive Appraisal (CAT) and Persistent Cognition Theories (PCT), we proposed a dual-path model: algorithmic management influences behavior via an "online" path (Perceived Algorithmic Control during work) and an "offline" path (post-work Work Rumination), with Self-Control Resource Depletion (risk factor) and Learning Agility (protective factor) as boundary conditions. 2. Key Findings Data supports the dual-path model, revealing a double-edged sword effect: Algorithmic management increases risk-taking via Perceived Algorithmic Control and Emotional Rumination, but reduces it via Problem-Solving Contemplation. Riders’ risk-taking is "silent compliance," driven by both active reward pursuit and passive penalty avoidance. Self-Control Resource Depletion amplifies Perceived Algorithmic Control’s negative impact; Learning Agility mitigates the effects of Perceived Algorithmic Control and Emotional Rumination. 3. Data Interpretation Algorithmic management’s influence extends beyond work hours via offline cognition. Behavioral outcomes are co-determined by algorithmic pressure, individual cognitive resources, and adaptive capabilities. Enhancing rider safety requires algorithm optimization (e.g., flexible time buffers) and programs to boost Learning Agility. 4. Data Collection Source: Two-wave matched questionnaire survey of food delivery riders (Meituan, Ele.me, etc.) in Changsha, China. Timeframe: T1 (Apr-May 2024), T2 (3 weeks post-T1). Sample: 320 valid matched responses (full-time/part-time riders). Design: Multi-wave offline collection (via station managers, field visits) to reduce common method bias. 5. Data Usage Variables: Constructs (APM, PAC, ER, PSC, RTB, etc.) measured with adapted 6-point Likert scales. Application: Ideal for gig economy, algorithmic management, occupational safety, and work psychology research (verification, secondary analysis, methodological reference).
Files
Steps to reproduce
1. Research Design & Participant Recruitment Adopted a two-wave time-lagged survey design (Changsha, China) to strengthen causal inference and reduce common method bias. Participants were full-time/part-time food delivery riders (Meituan, Ele.me). Recruitment methods (April 30–May 30, 2024): Station-Manager Facilitated Sampling: 4 delivery station managers distributed paper questionnaires during rider meetings. Field & Snowball Sampling: Researchers visited Changsha’s delivery hubs/merchant areas to invite riders; participants referred other eligible riders. All provided informed consent and received small cash incentives. 2. Data Collection Procedure & Protocol Data collected in two phases: Time 1 (T1, Apr–May 2024): 371 questionnaires measuring demographics (gender, age, tenure), Algorithmic Platform Management (APM), Perceived Algorithmic Control (PAC), Self-Control Resource Depletion (SRD). Time 2 (T2, 3 weeks post-T1): 344 questionnaires (same riders) measuring Emotional Rumination (ER), Problem-Solving Contemplation (PSC), Learning Agility (LA), Risk-Taking Behavior (RTB). 3-week interval: Based on prior research to capture lagged stress effects on persistent cognition. 3. Instruments & Workflows Instruments: All constructs used adapted mature scales (e.g., APM from Norlander et al. 2021/Yang et al. 2023; RTB from Zhang et al. 2015/Li et al. 2022) via translation-back-translation. 6-point Likert scale (1=Strongly Disagree to 6=Strongly Agree) to avoid neutral bias. Matching: Last 4 digits of riders’ mobile numbers as anonymous unique codes for T1-T2 matching. Data Cleaning: Invalid samples (mismatched codes, straight-lining, missing data) removed; 320 valid matched pairs (effective rate: 86.25%). 4. Software & Analysis Workflow Software: SPSS 27.0: Descriptive stats, correlations, reliability (Cronbach’s α), Harman’s single-factor test (common method bias), hierarchical regression, bootstrap (5000 resamples). AMOS 28.0: Confirmatory Factor Analysis (CFA) for measurement model validity. SPSS PROCESS Macro (Models 7,14): Test moderated mediation (conditional effects at ±1 SD of SRD/LA). Analysis Steps: Preliminary checks (common method bias, reliability/validity). Hypothesis testing: Hierarchical regression + bootstrap (direct/mediation effects); PROCESS (moderated mediation).
Institutions
- Hunan Agricultural University