Designing a Rashomon Machine: Pluri-perspectivism and XAI for Creativity Support

XAI should shift from justification to exploration: Rather than simply explaining why an AI made a specific decision, explainable AI can be redesigned to help users explore multiple creative possibilities—a fundamental reframing that better supports creative work than traditional XAI approaches focused on decision justification.

AI systems lack embodied understanding that humans possess: The paper identifies a critical limitation: generative AI models rely on disembodied data and therefore cannot fully understand the contextual, physical, and experiential knowledge that human creators bring to creative work, creating an "epistemological gap" that must be bridged in co-creative systems.

"Pluri-perspectivism" enables human-machine creative partnership: By adopting a framework that allows both humans and machines to exchange multiple "perspectives" on a problem (using techniques like the Rashomon Method), AI systems can present diverse creative options rather than single solutions, preserving human agency and decision-making.

Productive friction supports better creativity: The paper argues that reintroducing deliberate friction and complexity in human-AI collaboration—rather than designing for seamless automation—actually enhances creative outcomes by keeping humans actively engaged in the creative process rather than passive recipients of AI suggestions.