Assessing Leadership Competencies Through Social Network Analysis

Abstract

Leadership competencies are regularly identified as some of the most in demand workplace competencies. However, the development of these competencies requires appropriate assessments that are often either highly subjective (e.g. manager appraisals) or prohibitively expensive (e.g. roleplays with trained actors). The increasing usage of workplace social networks and increasing prevalence of digital collaboration tools presents a continuous stream of social interactions that can contain evidence of leadership occurring in situ. In this paper we present initial research on the feasibility of Social Network Analysis in the workplace to assess leadership competencies. We examine the assessment in terms of content, construct, and criterion validity. We then present our hypotheses on how the assessment can be conducted including the algorithms necessary to extract relevant features from a social network graph model. Our initial research, to our surprise, shows a weak correlation between an individual’s degree centrality and betweenness centrality and the leadership competency that is self-reported. However, experiments indicated a strong positive correlation between network structure based and social collaboration activities based features and the characteristics of the leadership competencies. Our initial machine learning experiments achieved an Area Under the Curve (AUC) score of 0.899 when social network and collaboration activity based features were leveraged to distinguish individuals with self-reported leadership competencies from others. Finally we discuss our findings on the practicality of the approach, and future work on validating and improving the results obtained using parallel conventional assessments for leadership competencies.