Design and Validation of a LangChain-Based Instructional Agent for Real-Time Support in Computer Science Laboratories

Abstract

The paper describes the structure and testing of an instructional agent powered by LangChain on a real time basis and used in Computer Science laboratory lessons. The suggested agent incorporates timely templates, chat memory, a layer of academic-integrity policies in order to provide scaffolded advice (by giving hints, reasoning checks, and validation processes) instead of directly dumping answers. Responsiveness, explanation clarity, debugging support, usability, perceived accuracy and overall satisfaction were measured using a quantitative instrument (5-point Likert). The results provided are furnished (N=120; illustrative) with high means of construct (4.19-4.40/5), high internal consistency (overall Cronbach a=0.89), and high rate of agreement in satisfaction and recommendation (78 percent). In addition to descriptive result, the presented items of the study provide item-total correlations, exploratory factor structure, inter-construct correlations, and a predictive regression model of satisfaction, which is consistent with the typical psychometric and technology acceptance validation procedures [3-6,9-13].