Cheap and Easy Open-Ended Text Input for Interactive Emergent Narrative

TL;DR

The paper addresses player overwhelm in wide-open action spaces of interactive emergent narrative games by enabling natural-language intents to drive actions through a lightweight, on-device embedding-based ranking. It presents PWIM, which combines sentence embeddings (via all-mpnet-base-v2/SBERT) with a Praxish action ontology and a Versu-inspired state-tracking layer, plus a UI that emphasizes close matches while allowing manual selection. The implementation is fully client-side JavaScript, demonstrating feasibility without server dependencies. The authors show that diverse intents surface appropriate top actions and discuss misclassifications and potential fine-tuning as practical mitigations, highlighting PWIM's viability for open-ended IEN without heavy parser development.

Abstract

We present a demonstration of Play What I Mean (PWIM): a novel, AI-supported interaction technique for interactive emergent narrative (IEN) games and play experiences. By assisting players in translating high-level gameplay intents (expressed as short, unstructured text strings) into concrete game actions, PWIM aims to support open-ended player input while mitigating the overwhelm that players sometimes feel when confronting the large action spaces that characterize IEN gameplay. In matching player intents to game actions, PWIM makes use of an off-the-shelf sentence embedding model that is lightweight enough to run locally on a player's device, and wraps this model in a simple user interface that allows the player to work around occasional classification errors.