Discovering knowledge work tasks from sequential event data

Abstract

As the use of digital tools and systems increases in today's knowledge-work-focused organizations, improving the work performed in digital systems has gained more attention. Digitalization is a key driver for knowledge work productivity through factors like improved communication, task automation, and artificial intelligence assistants. Monitoring and analyzing the work performed in digital systems can offer insights into how to realize all the potential of digital solutions. The thesis aims to contribute to this topic by developing a new analytical solution for analyzing frequent tasks performed in digital systems. The thesis is implemented in collaboration with a case organization and aims to address an organizational problem defined by the case organization. The case organization collects event data from digital systems using background software installed on knowledge workers’ computers. The research problem addressed in the thesis is how to identify frequently performed tasks from the collected data. Furthermore, the goal is to develop a methodology to analyze the identified tasks further and get insights into how to improve the work.

The research implementation started by conducting a literature review. The first objective of the literature review is to gain background knowledge on key topics related to tasks performed in digital environments. The second objective is to understand how to execute the data analysis with the main focus on discovering potential methods and algorithms to identify frequent patterns from sequence data. For the empirical research part, action research is selected as the research strategy. Action research uses an iterative approach to solve organizational issues through action and reflection. Action research focuses both on addressing the organizational issue and creating actionable theory. The research is carried out by first developing an Apriori-based algorithm to identify frequent task patterns from the data collected by the case organization. Next, multiple analysis and result representation techniques are developed to draw insights from the identified tasks. As per action research guidelines, the algorithm and results were iteratively improved. The feedback for improvements was collected from evaluation and reflection sessions organized with personnel from the case organization.

The literature review and empirical research helped to identify key areas where the task analysis results can be utilized to improve work efficiency. These key areas are task simplification, standardization, and automation. An analytical framework was formulated as a result of the research to generate an actionable theory. As a cornerstone to the framework, an approach with two algorithms is proposed to discover tasks from the data, where the first algorithm focuses on matching tasks with predefined structures, and the second algorithm aims to identify the most frequently performed tasks. For the result representation, a key factor is to have support for grouping the tasks so that the tasks with similar features are grouped together. Another key factor for supporting the analysis of results is the option to filter the tasks based on their features. Lastly, visualizing task flows on the digital system window level, displaying task volumes, and highlighting user activity-related metrics were identified to be key visualizations to support the insights generation.