<![CDATA[Affective assessment remains challenging in educational practice due to its subjective nature and the difficulty of systematically documenting students' attitudes and values. Although affective learning outcomes are essential for holistic education, assessment practices often rely on teacher observations that may be influenced by personal judgment. This study aimed to develop and conduct an initial validation of Affectra, a chatbot-based conversational assessment model for reflective affective assessment following learning activities. Unlike most educational chatbots designed for learning support or cognitive assessment, Affectra integrates conversational assessment, adaptive probing, Large Language Model (LLM)-based analysis, and Krathwohl's affective taxonomy into a unified assessment framework. A modified Research and Development (R&D) approach was adopted, consisting of needs analysis, model and instrument design, expert validation, alpha testing, and system refinement. The model was developed for secondary school learning, with validation limited to expert review and researcher-controlled alpha testing. Content validity was evaluated using the Index of Item–Objective Congruence (IOC) with six experts in educational assessment and educational technology, while system performance was assessed through functional and scenario-based testing. The instrument achieved an overall IOC score of 0.81, indicating satisfactory content validity, and demonstrated consistent classification across Krathwohl's five affective levels under predefined simulated scenarios. These findings provide preliminary evidence supporting the initial development and internal validation of Affectra. Further classroom validation involving actual students is required to evaluate its practical effectiveness.]]>
Copyrights © 2026
