<![CDATA[This study examines the effectiveness of the integration of a Large Language Model (LLM)-based AI chatbot named ‘GURU DIGITAL’ in improving the numeracy literacy of elementary school students in the Frontier, Outermost, and Disadvantaged (3T) areas. Addressing the challenges of educational disparities and low numeracy achievement in the 3T areas, ‘GURU DIGITAL’ is designed with an adaptive learning module that integrates local Sasak content, creating a personalized and culturally relevant learning experience. The research method uses a mixed method approach that combines Research and Development (R&D) with Participatory Action Research. The results show that ‘GURU DIGITAL’ was successfully developed with high NLP accuracy and an engaging adaptive module. Implementation in the field showed enthusiasm from students and teachers. A quantitative impact evaluation revealed a significant increase in students’ numeracy literacy in the experimental group (p < 0.001) compared to the control group, as measured by the Numeracy AKM. The proportion of students in the experimental group achieving the “Proficient” and “Advanced” categories increased drastically from 18% to 65%. Qualitatively, students and teachers reported high levels of satisfaction with ‘GURU DIGITAL’ due to its ability to provide easy-to-understand explanations, a fun learning environment, and personalized learning support. This study concludes that the integration of the ‘GURU DIGITAL’ LLM-based AI chatbot is a highly effective and potentially transformative innovation in improving numeracy literacy among elementary school students in the 3T (third-to-third) regions, while simultaneously leveraging local cultural contexts in the learning process. Recommendations include the replication and scalability of this model to ensure equitable access to quality education across the 3T (third-to-third) regions.]]>
