Topic modeling and clustering in the trace data-driven analysis of job demands among teachers

Abstract

Psychosocial work environment characteristics like job demands have traditionally been studied using survey data. We propose an alternative approach utilizing work related trace data collected from the information systems that employees use to achieve organizational goals. We analyze the job demands of teachers from two universities of applied sciences using trace data collected from the educational online platform Moodle over a period of 90 weeks. The data contain pairs of targets and actions (like message_sent) performed by teachers on Moodle. The timestamps of the target-action pairs allow us to study the dynamic nature of job demands, which is not possible by using periodically collected survey data. We show how trace data can be used to analyze processes related to job demands using data-driven approaches. We have identified topics, themes, temporal processes, and employee clusters from Moodle data representing the work tasks of teachers. The information obtained is action-oriented, context-specific, and dynamic, meeting the current needs for information about changing working life. The approach we have provided could be widely utilized in organizations as well as in research on occupational wellbeing. It is useful in identifying targets for intervention and it could be expanded to include prediction models on different outcomes.