Atharva Mahindrakar

Generative AI can rapidly produce complex driving scenarios, but these outputs are often difficult for users to understand, control, or evaluate. Scenario descriptions appear as text or parameter lists that lack narrative structure, visual clarity, and transparent causal reasoning. This creates barriers for researchers, designers, and other stakeholders who must interpret scenario logic, detect unrealistic elements, or communicate scenario details to others. Treating scenarios as stories, e.g. visual, temporal, and causal sequences, offers a more intuitive way to generate, refine, and understand AI-driven simulations.

In this project, students will design interactive interfaces that bridge natural-language generation, storyboard visualization, and explainable AI. They will develop a prototype that integrates conversational scenario creation, storyboard-based scenario representation, and explanatory overlays or narratives. Students will evaluate the impact of these interfaces on scenario comprehensibility, user control, trust, and error detection.

Deliverables will include literature reviews, UX concept design and interaction flows, prototype of integrated xAI scenario creation, usability analysis, scientific paper, and a mini-conference poster or demo.

Research questions / Possible topics for exploration:

·        How does a conversational interface influence user creativity, control, and precision when generating AI-driven scenarios?

·        Do storyboard-based visualizations improve user comprehension, communication, and memorability of scenarios compared to text-only descriptions?

·        Which explanation modalities (causal graphs, annotated storyboards, narrative summaries) best support error detection and trust in AI-generated scenarios?

• How does integrating conversation, visualization, and explanation into a single workflow affect perceived usability and mental model clarity?