Responsible People Analytics for Remote-Work Decisions: A Machine-Learning Benchmark for Classifying Perceived Productivity

Authors

• Ruth Menjivar, CSU San BernardinoFollow
• Nima Molavi, CSU San BernardinoFollow
• Narges Mashhadi Nejad, Angelo State UniversityFollow

Document Type

Article

Abstract

This study examines whether employee-perception survey data can support responsible people-analytics decisions about remote-work productivity. Using the public New South Wales (NSW) Remote Working Survey 2021 (N=1,512), the study benchmarks statistical and machine-learning classifiers for self-reported perceived productivity classes (same, less, or more productive when working remotely relative to onsite work), not objective output, under default, class-weighted, and resampling protocols. Main evidence comes from 5×5 repeated stratified cross-validation using macro F1 and balanced accuracy with fixed model specifications. Class-balanced separability is modest. Random Forest, CatBoost, and LightGBM form a leading cluster with overlapping confidence intervals (macro F1 ≈0.51–0.52). Affective/well-being items, especially feeling better when working remotely, carry disproportionate signal (well-being block macro F1 ≈ 0.485 vs. full model macro F1 ≈ 0.522), consistent with construct overlap and possible common-method variance. Resampling does not materially change this conclusion. Managers should not use survey classifiers for individual employee evaluation, return-to-office enforcement, or surveillance. The framework supports aggregate survey auditing, measurement evaluation, and more cautious interpretation of people-analytics dashboards. The study provides a decision-oriented reporting standard for responsible HR analytics using employee-perception survey data.

Recommended Citation

Menjivar, Ruth; Molavi, Nima; and Mashhadi Nejad, Narges (2026) "Responsible People Analytics for Remote-Work Decisions: A Machine-Learning Benchmark for Classifying Perceived Productivity," Journal of International Technology and Information Management: Vol. 35: Iss. 1, Article 3.
DOI: https://doi.org/10.58729/1941-6679.1662
Available at: https://scholarworks.lib.csusb.edu/jitim/vol35/iss1/3