<![CDATA[Digital interface design increasingly uses machine learning to personalize content from individual behavioral data. However, many adaptive interfaces remain socially blind because they optimize click history, dwell time, and navigation patterns without considering peer comparison, collective influence, and visible social feedback. This study aims to evaluate whether Behavior-Aware Interface Design, which integrates machine learning and social computing, can improve user engagement more effectively than a standard interface. A quasi-experimental pretest-posttest control-group design was conducted with 100 Indonesian digital users over 10 weeks, comprising 55 participants in the experimental group and 45 in the control group. User engagement was measured using the User Engagement Scale, while behavioral metrics included click-through rate, dwell time, and session length. Data were analyzed using an independent-samples t-test, a paired t-test, Wilcoxon test, Mann-Whitney U test, and ANCOVA in SPSS. The results show that the experimental group achieved a higher gain in Total Engagement than the control group, with gains of 16.35 and 10.04, respectively. ANCOVA confirmed a significant group effect on posttest engagement (p = 0.000; Partial Eta Squared = 0.552), and the model explained 75.6% of the variance in posttest engagement. Although CTR, dwell time, and session length increased significantly in both groups, the gains in these behavioral metrics did not differ significantly between groups. These findings indicate that behavior-aware interfaces are effective at improving holistic engagement, but user engagement cannot be reduced to individual behavioral indicators alone. The study concludes that integrating social computing into adaptive interface design offers theoretical and practical value for culturally contextual digital platforms, especially in collectivistic environments such as Indonesia.]]>
