Designing conversational Agents for adaptive instructional support in business simulation gaming

Texto completo

Adaptive instructional support that addresses individual learner differences in learning strengths, challenges, and interests is essential for engaging learners in meaningful cognitive processes and promoting learning. Although business simulation games (BSGs) are widely used for experiential entrepreneurial learning, they often provide limited formative feedback and structured reflection, which can overwhelm novice learners and exacerbate existing educational disparities. While AI-supported conversational agents (CAs) offer potential for scalable adaptive support, theory-driven guidance for their design in digital game-based learning (DGBL) remains scarce, particularly from an ethical and equity-by-design perspective. We adopted an Action Design Research approach to generate design knowledge and iteratively develop and evaluate an AI-enhanced CA, Lara, across two cycles. Grounded in educational psychology and information systems research, we derived theory-driven meta-requirements targeting cognitive, motivational, affective, and socio-cultural engagement, informed by a needs analysis conducted within a collaborative BSG context. Guided by an equity-by-design stance, we specified four design principles and fifteen design features, which together form the CAIS-GBL framework, and instantiated Lara accordingly. Evaluations with student teachers and within a field study with BSG participants indicated generally positive perceptions of Lara’s cognitive and social presence, support for self-regulated learning, perceived ease of use, and behavioral intention. Notably, perceived social presence and support for self-regulated learning increased significantly over the course of the five-week BSG. Furthermore, students’ gender, age, and prior knowledge did not significantly predict CA perceptions, whereas AI attitude and social player type did. These findings suggest that dispositions, rather than demographic or background characteristics, shape CA acceptance, indicating that Lara’s adaptive support aligns with diverse learner needs. Our study contributes 1) a functioning artifact, 2) a roadmap for equitable and adaptive instructional support in BSGs, and 3) actionable, theory-driven design knowledge for CAs in DGBL.