[Paper Review] Collaborative Storytelling with Human Actors and AI Narrators

Boyd Branch, Piotr Mirowski+1 more|arXiv (Cornell University)|Sep 29, 2021

Artificial Intelligence in Games16 references4 citations

TL;DR

This paper presents the first live, in-person performance of an AI narrator co-creating improvised theatre with human actors, using GPT-3 to generate narrative text in real time. The AI successfully guided story arcs and character development, with audiences and performers reporting strong preference for AI narration over AI characters, and the system reduced cognitive load on performers by handling narrative structure while enabling dynamic, emotionally resonant storytelling.

ABSTRACT

Large language models can be used for collaborative storytelling. In this work we report on using GPT-3 \cite{brown2020language} to co-narrate stories. The AI system must track plot progression and character arcs while the human actors perform scenes. This event report details how a novel conversational agent was employed as creative partner with a team of professional improvisers to explore long-form spontaneous story narration in front of a live public audience. We introduced novel constraints on our language model to produce longer narrative text and tested the model in rehearsals with a team of professional improvisers. We then field tested the model with two live performances for public audiences as part of a live theatre festival in Europe. We surveyed audience members after each performance as well as performers to evaluate how well the AI performed in its role as narrator. Audiences and performers responded positively to AI narration and indicated preference for AI narration over AI characters within a scene. Performers also responded positively to AI narration and expressed enthusiasm for the creative and meaningful novel narrative directions introduced to the scenes. Our findings support improvisational theatre as a useful test-bed to explore how different language models can collaborate with humans in a variety of social contexts.

Motivation & Objective

• To explore the collaborative potential of large language models as narrative partners in live improvisational theatre.
• To evaluate how AI narration affects narrative coherence, emotional engagement, and performer workload in real-time storytelling.
• To investigate whether AI can guide plot progression and character arcs without disrupting improvisational spontaneity.
• To assess audience and performer perceptions of AI-generated narrative quality and creative contribution.
• To develop a hybrid human-AI workflow that mitigates bias and ensures socially responsible storytelling.

Proposed method

• The researchers employed GPT-3 as a narrative agent, prompting it to generate long-form story segments in real time during live performances.
• A human operator curated and filtered AI-generated text using automated tools (e.g., Perspective API) and real-time judgment to remove offensive or inappropriate content.
• The AI narrator was integrated into a live theatre festival, where it provided ongoing narrative commentary and plot development during improvised scenes performed by professional improvisers.
• The system used constraints to generate longer, more coherent narrative outputs suitable for theatrical storytelling, distinct from dialogue-focused models.
• Performers were trained to accept and justify AI-generated narrative statements, treating them as 'offers' in the improvisational tradition.
• Audience and performer feedback was collected via surveys after each performance to evaluate narrative quality, engagement, and perceived AI contribution.

Experimental results

Research questions

• RQ1 How does an AI narrator influence narrative coherence and emotional depth in live improvised theatre?
• RQ2 Can large language models like GPT-3 serve as effective creative partners in long-form, spontaneous storytelling?
• RQ3 How do performers and audiences perceive the role of an AI narrator compared to an AI character within a scene?
• RQ4 To what extent does AI narration reduce the cognitive load on human improvisers?
• RQ5 How can bias and offensive content in AI-generated narratives be mitigated in real-time performance contexts?

Key findings

• Audiences and performers expressed strong preference for AI narration over AI characters, indicating that the narrator role was perceived as more natural and effective in guiding the story.
• Performers reported that the AI reduced their cognitive load by handling narrative structure, allowing them to focus more on emotional relationships and character dynamics.
• The AI introduced unexpected but meaningful narrative reversals—such as status shifts—often leading to more compelling story arcs than human-only improvisation.
• Despite occasional errors, such as misnaming characters, performers creatively adapted, transforming glitches into narrative opportunities like comic relief or fourth-wall-breaking moments.
• The combination of automated filtering (e.g., Perspective API) and human curation effectively mitigated offensive content while preserving narrative creativity.
• The study demonstrates that AI-as-narrator can function as a meaningful creative partner, enhancing rather than replacing human improvisational artistry.