Modeling computational dynamics of job interview candidate's mental states using cognitive agent based approach

• Main Author: Sanni, Ajoge Naseer
• Format: Thesis
• Language: English; English; English
• Published: 2019
• Subjects: HF5549-5549.5 Personnel Management. Employment
• Online Access: https://etd.uum.edu.my/8629/1/s95227_01.pdf; https://etd.uum.edu.my/8629/2/s95227_02.pdf; https://etd.uum.edu.my/8629/3/s95227_references.docx; https://etd.uum.edu.my/8629/

Support for job interview is a domain that can benefit from the research on human-aware AI systems. A developed cognitive model provides the awareness of interviewee behaviours as a mechanism for intelligent support processes. The interplaying constructs of self-efficacy, motivation and anxiety has been hypothesized to define the mental states of an interviewee. However, these constructs have not been integrated, formalized and evaluated for their dynamic intricacies in previous studies hence cannot be implemented as the reasoning component in human-aware system. This study has developed a cognitive agent model as a basic intelligent mechanism for interview coaching systems. The model integrates three constructs; self-efficacy, motivation and anxiety. Each of the constructs is formalized as an entity agent model and then integrated. Design Science Research Processes framework and Agent Based Modelling methodology were used to conduct this study. Factors interaction and overlapping relationship approach was adopted to integrate the proposed constructs. The model is formalized using Ordinary Differential Equation technique and later being simulated. Generated cases were verified with stability analysis and automatic logical verifications techniques. For model validation, 36 undergraduate students were studied in a mock interview experiment. The results generated from the model simulation were then compared against human experiment. The evaluation was based on a statistical technique namely Hotelling’s T2. The simulation results have confirmed a number of patterns identified in the domain literature. The behavioural patterns of the agent models conform to the expected behavioural dynamics of candidate in interview situation. Results from the validation showed that there is no significant difference (i.e. ρ values: anxiety = 0.391, self-efficacy = 0.128 and motivation = 0.466) between the simulation and human experiments. Theoretically, by integration of the three constructs, the model could better represent the mental state of candidates in interviews. In general, by formalizing the model, it can define the dynamic properties in details. The integrated cognitive model serves as a platform for designing a human-aware system that understands the behavioural intricacies of the user during job interview sessions.