<![CDATA[Geometry learning does not only require cognitive ability but is also influenced by students’ Math Anxiety. The integration of deep learning based on moral messages as a learning strategy that combines cognitive, affective, and character aspects is still rarely explored and has the potential to become a novelty in minimizing mathematics anxiety. This study involved 32 students of the Mathematics Education Study Program, FKIP Untan, using the Design and Development Research (DDR) method simplified into the stages of needs analysis, design, development, and limited trials. The study aimed to develop and test a deep learning model based on moral messages to minimize students’ Math Anxiety through Paired Sample T-test and SEM-PLS analysis. The results showed that deep learning based on moral messages was effective in minimizing students’ Math Anxiety through strengthening emotional regulation and creating a safe learning environment. After the treatment, students’ anxiety was no longer general in nature, but became milder, situational, and more focused on cognitive aspects, enabling students to better control their emotions and understand geometry concepts. Furthermore, this approach contributed 61.6% to changes in students’ conditions by reducing social anxiety, increasing academic courage, and fostering more positive learning engagement. These findings are important because they indicate that mathematics learning based on deep learning and moral messages can improve conceptual understanding while also enhancing students’ psychological conditions, making learning more humanistic, safe, and supportive of students’ academic confidence in mathematics.Keywords: Deep learning; Math Anxiety; Moral Messages.]]>
