Reproduce “DuoDrama: Supporting Screenplay Refinement Through LLM-Assisted Human Reflection” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yuying Tang, Xinyi Chen, Haotian Li, Xing Xie, Xiaojuan Ma, Huamin Qu
DuoDrama is an LLM-based screenwriting assistant designed to prompt human reflection rather than automatically rewrite scripts. By first simulating a character’s inner experience and then evaluating the scene from the actor’s perspective, it generates context-grounded questions that improved feedback richness, specificity, and several measures of reflective depth compared with unguided evaluation and reviewer-style baselines.
Key insights
- Linking experiential simulation to later evaluation provides a stronger basis for specific feedback than asking an evaluator to critique the screenplay without personal-experience context.
- Role switching preserves character-level psychological detail while adding critical distance, which helped writers report deeper reflection and stronger motivation to revise than character-only feedback in selected comparisons.
- Feedback timing remains a design tension: post-hoc comments were often preferred for broader analysis, while frequent instant interventions could disrupt reading and creative flow.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.