The AKSARA Learning Model: Developing an Instructional Model Through the Differentiated-Based Deep Learning (DBDL) Approach to Enhance the Literacy Competence of Junior High School Students

Abstract

This study developed and evaluated an instructional model that integrates differentiated instruction and deep learning to strengthen the literacy competence of junior high school students. The study employed a Research and Development design using the ADDIE model. The Differentiated-Based Deep Learning (DBDL) approach served as the conceptual foundation, and the development process produced the AKSARA learning model, whose five phases are Learning Readiness Analysis, Differentiated Knowledge Construction, Deep Learning Synergy, Actualization of Differentiated Meaningful Work, and Reflection and Evaluative Action. The model was supported by a model book, a teaching module, and student worksheets. Three experts validated the products and the research instruments using Aiken's V. A limited trial was conducted in one school with two Grade IX classes (n = 64), and an expanded trial was conducted in three junior high schools in Wajo Regency, Indonesia, with six classes (n = 160). Practicality was assessed through implementation-fidelity observation, teacher classroom-management observation, student-behaviour observation, and teacher and student response questionnaires. Change in literacy scores was examined with a one-group pretest-posttest design, Shapiro-Wilk tests, paired-sample t-tests, normalized gain, and Cohen's dz. The model book obtained an Aiken's V of 0.961, the teaching module 0.912, and the student worksheets 0.895; the research instruments ranged from 0.897 to 0.944, and all values met the critical value of 0.889 for three raters. Practicality increased from 90.91% in the limited trial to 92.05% in the expanded trial, both in the highly practical category, and implementation fidelity rose from 91.39% in the first meeting to 94.10% in the sixth. In the expanded trial, the mean literacy score increased from 47.563 to 84.500, a mean difference of 36.938 points, 95% CI [34.730, 39.145], t(159) = 33.050, p < .001, with an N-Gain of 0.704; 95.00% of the students achieved moderate or high individual gains. Because the design included no comparison group, because students were nested in classes and schools, and because the outcome test was developed for this study and administered in identical form on both occasions, the pre-post change should be read as evidence consistent with the model rather than as an estimate of its causal effect. The findings show that differentiation and deep learning can be combined into a single syntax that experts judge valid and that teachers can execute with high fidelity in ordinary classrooms.