Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Marianne Bossema, Rob Saunders, Vlad Glaveanu, Somaya Ben Allouch

TL;DR

The paper addresses the challenge of enabling human-centered co-creative AI that respects embodied creativity and context sensitivity, highlighting the limitations of disembodied data and traditional XAI. It proposes Pluri-perspectivism as an XAI framework to bridge human and machine epistemologies and introduces the Rashomon Machine to generate a spectrum of possibilities via a five-dimensional context map. Implementation is detailed through schema-guided prompting and few-shot conditioning of vision-language models to create a Rashomon Set of enactive explanations, enabling Perspective Taking and Offering to sustain exploratory creativity. The work identifies future research directions, including adaptive explanation strategies, temporal dynamics of interventions, and privacy-conscious deployments, aiming to validate and refine a context-aware, embodied creativity support system with non-expert users. The approach offers a principled path to reintroduce productive friction and human agency in human-machine creative collaborations, with potential impact on everyday creativity support for diverse user groups.

Abstract

While intelligent technologies offer unique opportunities for creativity support, there are fundamental challenges in designing human-centered co-creative systems. Explainable AI (XAI) can contribute when shifting its traditional role from justification (explaining decisions) to exploration (explaining possibilities). Contextual understanding is essential for supporting embodied creativity. Generative Artificial Intelligence (AI) models are fundamentally limited, however, by their reliance on disembodied data. We propose Pluri-perspectivism as a framework for XAI, to bridge the epistemological gap between human and machine, and promote creative exploration. It is a pragmatic, action-oriented solution to guide the system, repurposing XAI methods such as the Rashomon Technique. This facilitates exploring a spectrum of creative possibilities, and the exchange of 'perspectives' between human and machine. Using Pluri-perspectivism as a framework for XAI, we can reintroduce productive friction and support human agency in human-machine creative collaborations.

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

TL;DR

The paper addresses the challenge of enabling human-centered co-creative AI that respects embodied creativity and context sensitivity, highlighting the limitations of disembodied data and traditional XAI. It proposes Pluri-perspectivism as an XAI framework to bridge human and machine epistemologies and introduces the Rashomon Machine to generate a spectrum of possibilities via a five-dimensional context map. Implementation is detailed through schema-guided prompting and few-shot conditioning of vision-language models to create a Rashomon Set of enactive explanations, enabling Perspective Taking and Offering to sustain exploratory creativity. The work identifies future research directions, including adaptive explanation strategies, temporal dynamics of interventions, and privacy-conscious deployments, aiming to validate and refine a context-aware, embodied creativity support system with non-expert users. The approach offers a principled path to reintroduce productive friction and human agency in human-machine creative collaborations, with potential impact on everyday creativity support for diverse user groups.

Abstract

While intelligent technologies offer unique opportunities for creativity support, there are fundamental challenges in designing human-centered co-creative systems. Explainable AI (XAI) can contribute when shifting its traditional role from justification (explaining decisions) to exploration (explaining possibilities). Contextual understanding is essential for supporting embodied creativity. Generative Artificial Intelligence (AI) models are fundamentally limited, however, by their reliance on disembodied data. We propose Pluri-perspectivism as a framework for XAI, to bridge the epistemological gap between human and machine, and promote creative exploration. It is a pragmatic, action-oriented solution to guide the system, repurposing XAI methods such as the Rashomon Technique. This facilitates exploring a spectrum of creative possibilities, and the exchange of 'perspectives' between human and machine. Using Pluri-perspectivism as a framework for XAI, we can reintroduce productive friction and support human agency in human-machine creative collaborations.
Paper Structure (11 sections, 3 figures, 1 table)

This paper contains 11 sections, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Motivation
  3. Background
  4. Creative Experience
  5. Creative Context
  6. Pluri-perspectivism
  7. Pluri-perspectivism Scaffolding Creative XAI
  8. Integrating XAI Methods for Creativity Support
  9. Opportunities for Further Research
  10. Discussion
  11. Conclusion

Figures (3)

  • Figure 1: The Pluri-perspectivism Framework for Human-Machine Co-creativity. (Left) Attention & Action Orientations: A schematic map of the creative context, representing five different orientations towards the artwork in the center. This multidimensional schema serves as a reference system for grounding creative context-sensitivity. (Right) Exchange of Perspectives: The continuous feedback loop of exchanging perspectives, using the schema for mapping orientations and navigating the creative possibility space.
  • Figure 2: A sequence of snapshots of a drawing activity, showing artwork in progress, providing objects as templates for tracing, water-soluble markers, chalk and water brushes.
  • Figure 3: The Rashomon Machine feedback loop with XAI methods facilitating it. Bottom - Perspective Taking: The system assesses Feature Importance and maps user explanations on the Pluri-perspectivism schema. Right: The Rashomon Set functions as a collection of latent possibilities, continuously refined, serving as the system's mental model of the creative context. Top - Perspective Offering: The set is filtered based on the creative state, to select explanations that deepen or broaden exploration.