Automatic Assessment and Identification of Leadership in College Students

As one of the 21st Century skills, fostering leadership competencies in college students has been a focus of research in the behavioral education field. While leaders are expected to have general leadership traits such as communication and resource management, leading teams in particular expertise areas requires additional skills specifically tailored to the tasks involved. For example, in information technology (IT), the roles of leaders are different from those in other non-IT fields due to the characteristics of IT projects. IT leaders, in particular, are required to have not just managerial and interpersonal skills but also necessary technical and practical capabilities. Understanding the key elements that contribute to the emergence and development of student leaders necessitates the ability to assess and recognize leadership in college students. However, current approaches for measuring leadership qualities entail administering questionnaires, which have coverage constraints and may mislabel future leaders if the survey is administered too early in a student's educational phase, when their leadership talents may not have emerged yet. Addressing this problem, this paper introduces LAIGA, a machine learning-based framework for automatically and passively assessing and identifying leaders in college students from their academic profiles and behavior in the learning management system (LMS). The proposed technique can capture students' actions that indicate general leadership qualities and technical competence, as evidenced by their academic performances. Specifically, we address two problems: leadership assessment and leader identification tasks. A case study of senior-year college students in the IT major is used to validate the proposed method, yielding 5.87% MAPE in the leadership assessment task and 83.2% F1 in the leadership identification task. Furthermore, the results also show that the proposed method can forecast leadership assessment and identification after graduation, knowing as little as students' information during their first year of study. The findings of this study not only shed light on the ability to use machine learning technologies to automatically assess and identify leadership in college students but also allow educational organizations to understand factors that help develop leadership skills and design mechanisms that promote the growth of student leaders. The proposed method is highly generalizable to other fields of study as long as students' grades and LMS activities are available.