Algorithmic Work Demands, HRM Buffering, and Digital Wellbeing: Three-Study Examination of Algorithm-Work Demands and Workers’ Human Resource Management Practices in AI-Fuelled Indian Organizations

Abstract

Purpose: The present paper sought to offer theoretical grounds for formulating 14 hypotheses on algorithm-work demands as a source of strain. The study evaluated algorithm-work intensification, algorithm-work expectancy of digital availability, algorithm-work incursion, and algorithm-work information overload as resources for workers' digital flourishing, work-family technology boundary, psychological safety, and intent to turnover to reduce the negative consequences of algorithm-work demands. The authors examined if and how workers' human resource management (HRM) practices moderated the relationships between workers' algorithm-work demands and their outcomes. The authors paid attention to workers' responses to the challenges of algorithm-work demands. Thus, this paper provided an empirical validation of theoretical models of job demands – resources, conservation of resources, social exchange, and self-determination theories of algorithm-work demands and workers' HRM practices in India. This study intended to add to the understanding of the mechanisms and context in which algorithm-work demands related to workers' digital flourishing, work-family technology boundary, psychological safety, and intent to turnover and how such relationships were moderated by employees' HRM practices.

Design/Methodology/Approach: The authors used a cross-sectional survey of 421 employees in Hyderabad and Bengaluru, India, who worked in IT services, manufacturing, and financial services industries as secondary data for three studies. For the first study, the authors assessed the structural model with the use of partial least squares structural equation modeling (PLS-SEM) software with bootstrapping procedure (5,000 samples) and PLS predict cross-validation procedure. The results revealed that the Q 2 values for DF and WLTB were 0.591 and 0.567, correspondingly. Hence, it was possible to continue to analyze the associations in various contexts (IT versus manufacturing) and work arrangements (remote versus office) on the basis of the multi-group analysis (MG) approach for Study 2 with the permutation test (5,000 runs). For Study 3, the authors used an XGBoost regressor (eta – 0.07; max depth – 6; test R² = 0.724) and a Random Forest regressor (trees – 500; AUC-ROC – 0.944) to construct two predictive models. To interpret predictions, the authors analyzed the SHAP Tree Explainer values and PLS path coefficients. The basis for constructing a theoretical model of the demand-strain-outcome pathway was the Job Demands-Resources Theory, Conservation of Resources Theory, Social Exchange Theory, and Self-Determination Theory. The role of human resource management (HRM) practices was identified as a moderator of the relationship.

Findings: Digital work exhaustion is the strongest predictor of lower digital flourishing (β=−0.389, f²=0.191) and the most ranked SHAP feature (0.412). HRM buffering practices significantly moderate the exhaustion-to-flourishing path: at high HBP, the exhaustion slope decreases from β=−0.487 to β=−0.267 – 45% strain buffering effect. Remote workers have significantly higher digital availability demand impact on exhaustion (Δβ=0.127, p=0.013) and detachment impairment (Δβ=0.159, p=0.008) than in-office workers. Employees in the IT sector experience significantly higher AI monitoring effects on exhaustion than manufacturing (Δβ=0.113, p=0.024). SHAP-PLS Spearman convergence is rs=0.97. SRMR=0.049, NFI=0.942.

Practical Implications: HRM buffering practices are the only organizational tool confirmed to reduce the exhaustion-to-wellbeing damage caused by algorithmic work demands. Remote work MGA findings speak in favor of differentiated digital wellbeing policies by work mode rather than universal HRM standards. Sector-specific AI monitoring finding indicates that manufacturing-appropriate monitoring intensity thresholds are not suitable for knowledge work.

Originality/Value: This study represents the first empirical application of a combination of PLS-SEM, dual-contrast PLS-MGA (sector and work mode), and XGBoost+RF+SHAP validation to the issue of algorithmic work demands and digital wellbeing in Indian AI-enabled organisations, with HRM buffering tested as a moderator.