Pengembangan Media Chatbot Interaktif Berbasis Artificial Intelligence (AI) untuk Meningkatkan Self-Efficacy dan Hasil Belajar Peserta Didik Menggunakan Model ADDIE

Abstract

This development research aims to develop an interactive chatbot-based learning media called HistorySmartAI using the ADDIE model, evaluate its feasibility through expert validation, and determine its effectiveness in improving tenth grade students' self-efficacy and history learning outcomes. The research was conducted at SMAN 5 Jember from April to May 2026. This study adopted a quasi-experimental method with a pretest-posttest control group design, involving class X.3 as the experimental group and class X.2 as the control group, each consisting of 36 students. Data collection was conducted through observation, expert validation sheets, pre-tests, post-tests, and a self-efficacy questionnaire. The instructional materials focused on the cultural heritage of the ancient Mataram Kingdom, specifically Borobudur Temple, integrated with a Self Regulated Learning (SRL) approach. Data from the small-group trial were analyzed using the Paired Samples t-test and Eta Squared effect size analysis, while data from the field test were analyzed using the Independent Samples t-test and N-Gain score analysis. The formative evaluation results from experts showed that the media was highly feasible, with material experts awarded a score of 5.00 (Outstanding), media experts 4.26 (Above Average), and language experts 4.55 (Above Average). The small group trial involving 10 students showed an Eta Squared effect size value of 0.810 for self-efficacy with a large effect and a value of 0.847 for learning outcomes with a large effect. Furthermore, the summative evaluation in the field test indicated that the media was highly effective for both variables. For history learning outcomes, the average score of the experimental class increased from a mean pre-test score of 62.10 to a mean post-test score of 87.26 with an N-Gain score of 66.00 (Moderately Effective). Meanwhile, for students' self-efficacy, the average score increased from 59.10 to 86.58 with an N-Gain score of 67,84 (Moderately Effective), which significantly outperformed the control class. In conclusion, HistorySmartAI is highly feasible and effective as a learning medium for improving students' self-efficacy and history learning outcomes.

Keywords: Chatbot, HistorySmartAI, Self-Regulated Learning, Self-Efficacy, History Learning Outcomes