Unveiling Test Response Patterns: Cluster Analysis of University Students' Career Aptitude Assessments for AI-Enabled Coaching

Abstract

Career indecision is a major source of anxiety among undergraduate students. Though many students believe career counselling services provide essential aid for this anxiety, a lack of accessibility or availability is often attributed as the reasons why a significant proportion of students never avail of such services. AI poses as an alternative to traditional career guidance and shows great potential to address this lack of accessibility or availability. Sollicity is aiming to develop their own AI coach to assist students in a new and easily accessible manner. This project aims to help Sollicity better understand their student aptitude test data through implementing cluster analysis in an effort to develop generalized profiles to allow for improved career path recommendations by the AI coach. Minimisation of cost to Sollicity in the operation of the AI coach can be performed through dimension reduction techniques to reduce the necessary questions needing to be asked to guide a student to their desired career path. Another aim of this project is to validate the use of principal component analysis (PCA) techniques to infer if this is a viable method for Sollicity to employ in this effort of cost minimisation. The cluster analysis results reveal a consistent 2-cluster solution representing the students that completed the aptitude tests. Question minimization through cumulative variance PCA methods is found to preserve the cluster structure of the data well. The simplistic 2-cluster findings are unlikely to provide reliable generalized profiles for Sollicity’s AI coach. A significant data reduction due to many incomplete tests is likely the cause of such cluster findings. Future research should consider the implementation of alternative clustering algorithms in conjunction with or as alternatives to the ones implemented in this thesis to provide more comprehensive results.