Perceived algorithmic control and courier safety risk in last-mile delivery: a transparency paradox

Purpose Drawing on job demands–resources theory, this study examines how perceived algorithmic operations control (PAC) relates to couriers' safety-compliance risk. PAC denotes algorithmic oversight: feeling monitored, directed and constrained by platform algorithms. The risk is physical: unsafe riding that can cause crashes. Two mechanisms are tested: system-triggered interruptions that fragment attention (a job demand) and reduced autonomy to adapt safely (a lost resource). Boundary conditions are dependence on platform income and visibility-based transparency, a widely advocated remedy. Design/methodology/approach In an exploratory sequential mixed-methods design, interviews and field observations informed the measures. A cross-sectional questionnaire survey of 344 food-delivery couriers in China was analysed using regression-based moderated-mediation tests. Riders' forum posts, produced independently of the survey, served to triangulate (cross-check) the survey constructs. Findings Stronger PAC is associated with more self-reported risky riding (incidents were not measured), partly through more interruptions and lower autonomy. Economic dependence weakens the interruption pathway but strengthens autonomy loss. Transparency strengthens both pathways, linking PAC to more interruptions and greater autonomy loss. This is the transparency paradox: a presumed remedy unintentionally amplifies risk. Originality/value Algorithmic management operates through adjustable system parameters: how often platform data interrupt couriers and how much discretion they retain. To cut risk without losing efficiency, platforms can pace notifications, freeze mid-route reassignments, restore discretion through exception handling and pair transparency with actionable options. For regulators, disclosure mandates alone may not make delivery safer.