Programming for generalization: Confronting known challenges in the design of virtual reality interventions for autistic users

This study explored how to program for generalization using a fully immersive virtual reality (VR) intervention for teaching transportation skills to autistic adults related to using a university shuttle. Using multi-methods, this research sought to identify similarities and differences in behavior between the VR environment and the real-world, participants’ perceptions of telepresence in the VR environment, and how participants characterized similarities and differences between the VR environment and the real-world. Male adult participants (N ​= ​6) completed training and then engaged in two VR-based training sessions of increasing complexity, after which they enacted what was learned in the real-world. Fidelity of implementation was high across VR and real-world sessions and no significant differences were found in behaviors between the VR and real-world sessions, providing evidence for skills generalization from the contrived VR setting to the naturalistic real-world setting. Participants reported high perceptions of telepresence (e.g., being there) and social presence (e.g., being there with others), and qualitative evidence suggests they made connections between the virtual world and real-world. Implications and directions for future research are discussed. •Generalization heuristics can support generalization from contrived settings to naturalistic settings.•Visual realism alone is insufficient for promoting generalization.•Task realism, object/behavioral congruence, & implementation fidelity promote generalization.•Multi-modal virtual reality (i.e., 360 video, fully immersive VR) beneficial for generalization heuristics.