Related Experiment Video
Updated: Oct 8, 2026
Estimate the Cognitive Load Using Electrocardiographic Measure: A Human-AI Collaborative Task
Published on: December 5, 2025
Which algorithm hurts most? Differential associations of task allocation, scheduling, and performance monitoring with
Xingyao Xiao1,2, Yihong Cheng3
1Graduate School of Education, Stanford University, Stanford, CA, USA. Doria.Xiao@acer.org.
Objectives:
Algorithmic management (AM) is rapidly diffusing from platform work into traditional sectors, yet empirical evidence on its implications for worker psychosocial well-being remains limited and often treats AM as a unitary construct. This study investigates whether three distinct AM functions (algorithmic task allocation, scheduling, and performance monitoring) exhibit differential associations with multiple dimensions of worker psychosocial well-being.
Methods:
We analysed European Working Conditions Survey 2024 data on 26 224 workers in 27 EU countries. Outcomes included the WHO-5 Well-Being Index (0-100), self-reported stress, emotional exhaustion, loneliness, sleep problems, and anxiety. Linear regression models with country fixed effects controlled for demographics, occupation, workplace size, and digital tool use. Robustness checks included random intercept multilevel models, within-occupation subgroup analyses, and stratification by national digitalisation regime.
Results:
The three AM dimensions showed qualitatively different patterns. Algorithmic task allocation was not associated with adverse outcomes and was inversely associated with anxiety ([Formula: see text], [Formula: see text]). By contrast, algorithmic scheduling showed the strongest associations with emotional exhaustion ([Formula: see text], [Formula: see text]) and loneliness ([Formula: see text], [Formula: see text]), while performance monitoring showed the strongest associations with stress ([Formula: see text], [Formula: see text]) and anxiety ([Formula: see text], [Formula: see text]). The WHO-5 index showed no overall association with AM ([Formula: see text], [Formula: see text]), consistent with offsetting dimension-specific associations. All patterns held across occupation groups and across high- and low-digitalisation countries.
Conclusions:
Not all algorithmic management functions are equally associated with poorer worker well-being. Algorithmic scheduling and performance monitoring warrant priority in occupational psychosocial risk assessments, while general well-being instruments may underestimate dimension-specific adverse associations and should be complemented by specific indicators.