[Paper Review] DuoDrama: Supporting Screenplay Refinement Through LLM-Assisted Human Reflection

Yuying Tang, Xinyi Chen+4 more|arXiv (Cornell University)|Feb 5, 2026

Artistic and Creative Research0 citations

TL;DR

DuoDrama presents an ExReflect workflow enabling AI to provide coordinated internal character experiences and external evaluative feedback to aid screenplay refinement, demonstrated with professional screenwriters.

ABSTRACT

AI has been increasingly integrated into screenwriting practice. In refinement, screenwriters expect AI to provide feedback that supports reflection across the internal perspective of characters and the external perspective of the overall story. However, existing AI tools cannot sufficiently coordinate the two perspectives to meet screenwriters' needs. To address this gap, we present DuoDrama, an AI system that generates feedback to assist screenwriters' reflection in refinement. To enable DuoDrama, based on performance theories and a formative study with nine professional screenwriters, we design the Experience-Grounded Feedback Generation Workflow for Human Reflection (ExReflect). In ExReflect, an AI agent adopts an experience role to generate experience and then shifts to an evaluation role to generate feedback based on the experience. A study with fourteen professional screenwriters shows that DuoDrama improves feedback quality and alignment and enhances the effectiveness, depth, and richness of reflection. We conclude by discussing broader implications and future directions.

Motivation & Objective

• Identify design goals and requirements for AI feedback that supports reflection in screenplay refinement.
• Develop a system that coordinates internal (character experience) and external (story-wide) perspectives in feedback.
• Ground feedback in performance-theory principles to balance immersion and critical distance.
• Evaluate whether ExReflect-based DuoDrama improves feedback quality, alignment, and depth of screenwriter reflection.

Proposed method

• Propose ExReflect, a two-stage workflow where an AI agent first generates inner experiences from a character’s perspective, then shifts to an external evaluation role to produce feedback.
• Implement ExReflect via a multi-agent architecture where each character has an agent that maintains long-term and short-term memory for experiential grounding.
• Use a two-stage prompt architecture linking experience enactment and evaluation feedback, guided by five dimensions of reflection (emotions, motivations, relationships, pacing, theme).
• Preprocess screenplays into scenes and manage a centralized orchestrator to coordinate agent actions during refinement.
• Incorporate a self-verification mechanism to ensure feedback is evidence-grounded, diverse in expression, tied to the five dimensions, and timely.

Experimental results

Research questions

• RQ1 How can AI provide feedback that coordinates both internal character experience and external narrative structure during screenplay refinement?
• RQ2 What design goals and requirements do screenwriters have for AI-supported reflection in refinement?
• RQ3 Can an experience-grounded feedback workflow improve feedback quality and alignment with screenplay goals?

Key findings

• A formative study with nine professional screenwriters identified four design goals shaping ExReflect and DuoDrama.
• A fourteen-participant study showed that DuoDrama produced higher-quality, more aligned feedback and enhanced screenwriter reflection.
• Feedback in DuoDrama is anchored to five dimensions: emotions, motivations, relationships, pacing, and thematic consistency, and is contextually grounded in the draft.