In social environments, autonomous systems must reconcile potentially conflicting norms, obligations and contextual constraints in order to make rational and explainable decisions. In this paper, we present an executable implementation of a norm-based argumentation framework for decision-making in socially regulated agent environments and analyze the scalability of preferred semantics in norm-rich scenarios such as RoboCup@Home Human-Robot Interaction tasks. We show that the number of generated arguments can grow rapidly, making naive preferred computation impractical in practice. Exploiting a structural regularity common in robotic normative reasoning—namely that multiple norms support identical action-level conclusions—we introduce a head-level collapse optimization. This transformation preserves preferred semantics while reducing the search space from \(O(2^{|A|})\) to \(O(2^{|H|})\) . Experimental results demonstrate that the optimized approach remains computationally stable under scaling.