DuoDrama: Supporting Screenplay Refinement Through LLM-Assisted Human Reflection

Key findings from this study

This research indicates that:

• AI agents adopting experiential roles before evaluation roles generate more coordinated feedback across character and story perspectives.
• Screenwriters' refinement processes benefit from feedback that integrates character-level internal dynamics with narrative-level external structure.
• Designing AI feedback systems based on domain-specific performance theories and practitioner insights produces tools that enhance reflection depth and richness.

Overview

DuoDrama is an AI system designed to support screenwriters during screenplay refinement by generating feedback that coordinates character-level internal perspectives with story-level external perspectives. Current AI tools inadequately integrate these dual viewpoints. The system implements the Experience-Grounded Feedback Generation Workflow for Human Reflection, wherein an AI agent first adopts an experience role to simulate character and narrative dynamics, then shifts to an evaluation role to generate perspective-coordinated feedback.

Methods and approach

Formative research with nine professional screenwriters informed the design of ExReflect, which operationalizes the dual-perspective feedback approach. The system's effectiveness was evaluated through a study with fourteen professional screenwriters, assessing feedback quality, alignment with screenwriter needs, and the depth and richness of reflection supported.

Results

DuoDrama improved feedback quality and achieved stronger alignment with screenwriter expectations compared to baseline approaches. The system enhanced the effectiveness of reflection by supporting screenwriters in examining both character interiority and overarching narrative coherence simultaneously. Participants reported that the feedback supported deeper and richer reflection during refinement, moving beyond surface-level critique to enable more nuanced consideration of how character perspectives interact with story architecture.

Implications

The coordination of dual perspectives in AI-assisted screenwriting represents a broader principle for creative feedback systems: AI tools can more effectively support creative refinement when designed to model multiple interpretive frames relevant to the domain. This finding suggests that performance theories and role-based reasoning can structure AI feedback generation to align with how practitioners think about complex creative artifacts. Future systems might extend this approach to other narrative or performance domains where internal and external perspectives require simultaneous consideration.

Scope and limitations

This summary is based on the study abstract and available metadata. It does not include a full analysis of the complete paper, supplementary materials, or underlying datasets unless explicitly stated. Findings should be interpreted in the context of the original publication.