Believable Digital Humans: Toward a Mutable Design Framework for Customer Service Experiences

Digital humans are increasingly deployed in customer-facing service contexts, yet guidance for systematically designing and evaluating their believability remains limited. Although prior research identifies principles associated with believable agent behavior, these insights are often fragmented, under-operationalized and outcome focused in nature. As the state-of-the-art stands, little to no guidance is offered as to how designers can reason holistically about digital human experiences. Our work addresses this gap, proposing a design framework (as an artefact) that conceptualizes believability as a higher-order construct organized into dimensions, characteristics, and attributes—the aim being to support design reasoning and comparative evaluation under a realistic service scenario. Adopting a design science research approach, the framework is developed through a synthesis of prior literature, stakeholder engagement, and co-creation activities—the outcomes articulated as a nascent design theory. The framework is instantiated through two alternative digital human brand ambassadors, evaluated in the context of a higher education customer service scenario. Evaluation is explicitly structured around criterion, causal, and context validity, with conservative analytic controls applied to bound inference. Our findings show that the framework can be meaningfully applied in use, and that framework-guided design choices can produce perceptible, though selective, differentiation in user judgments of believability as stands. Differentiation is most robust for personality-related and selected behavioral characteristics, while other characteristics show weaker or context-sensitive effects. Notably, some appearance-related perceptions differ despite controlled visual features, suggesting that believability judgments are holistic and shaped by cross-dimensional interactions. Overall, the outcomes contribute to a mutable, design-oriented framework that supports structured comparison and diagnostic insight, providing a foundation for cumulative and context-aware research on believable digital humans.