Information Technology


Adaptive Storytelling Through User Understanding

AAAI Conferences

Storytelling, when happening face to face, is a highly interactive process. A good storyteller attracts the audience by continuously observing their responses and adjusting the storytelling accordingly. The goal of this project is to simulate this process in digital storytellers. We created an automatic storytelling system that periodically estimates the user’s preferences and adapts the content in the subsequent storytelling by balancing novelty and topic consistency. We have performed an empirical evaluation on the effectiveness of our approach. The results indicate that gender and age play important roles in affecting one’s subjective experience. For younger subjects, stories with mixed amount of novelty and topic consistency are more preferred while for older subjects, larger amounts of variation are preferred. Additionally, in general, women enjoyed the stories more than men.


Automated Generation of Diverse NPC-Controlling FSMs Using Nondeterministic Planning Techniques

AAAI Conferences

We study the problem of generating a set of Finite State Machines (FSMs) modeling the behavior of multiple, distinct NPCs. We observe that nondeterministic planning techniques can be used to generate FSMs by following conventions typically used when manually creating FSMs modeling NPC behavior. We implement our ideas in DivNDP, the first algorithm for automated diverse FSM generation.


Learning Interrogation Strategies while Considering Deceptions in Detective Interactive Stories

AAAI Conferences

The strategies for interactive characters to select appropriate dialogues remain as an open issue in related research areas. In this paper we propose an approach based on reinforcement learning to learn the strategy of interrogation dialogue from one virtual agent toward another. The emotion variation of the suspect agent is modeled with a hazard function, and the detective agent must learn its interrogation strategies based on the emotion state of the suspect agent. The reinforcement learning reward schemes are evaluated to choose the proper reward in the dialogue. Our contribution is twofold. Firstly, we proposed a new framework of reinforcement learning to model dialogue strategies. Secondly, background knowledge and emotion states of agents are brought into the dialogue strategies. The resulted dialogue strategy in our experiment is sensitive in detecting lies from the suspect, and with it the interrogator may receive more correct answer.


A Nondeterministic Temporal Planning Model for Generating Narratives with Continuous Change in Interactive Storytelling

AAAI Conferences

In this paper, we propose a temporal planning model for real-time generation of narratives in Interactive Storytelling systems. The model takes into account continuous branched time and the specification of constraints defined as temporal formulae over dramatic properties of the narrative (e.g. joy or tension). In order to address real-time generation, dramatic properties are modeled as varying linearly and events go through a preprocessing stage. As proof of concept, the model is incorporated into an existing storytelling system, LOGTELL, which provides a way to logically specify genres; allows user interaction to influence events in the unrolling narrative; and dramatizes the story in a 3D computer graphics world. To illustrate the generation of narratives, we present a simple narrative example in the Swords and Dragons genre.


Integrating Monte Carlo Tree Search with Knowledge-Based Methods to Create Engaging Play in a Commercial Mobile Game

AAAI Conferences

Monte Carlo Tree Search (MCTS) has produced many recent breakthroughs in game AI research, particularly in computer Go. In this paper we consider how MCTS can be applied to create engaging AI for a popular commercial mobile phone game: Spades by AI Factory, which has been downloaded more than 2.5 million times. In particular, we show how MCTS can be integrated with knowledge-based methods to create an interesting, fun and strong player which makes far fewer plays that could be perceived by human observers as blunders than MCTS without the injection of knowledge. These blunders are particularly noticeable for Spades, where a human player must co-operate with an AI partner. MCTS gives objectively stronger play than the knowledge-based approach used in previous versions of the game and offers the flexibility to customise behaviour whilst maintaining a reusable core, with a reduced development cycle compared to purely knowledge-based techniques.


Mimicking Humanlike Movement in Open World Games with Path-Relative Recursive Splines

AAAI Conferences

In this paper we explore the use of recursive cubic Hermite splines to mimic human movement in open world games. Human-like movement in an open world environment has many characteristics that are not optimal or directed towards clear, discrete goals. Using data collected from a simple MMORPG-like game, we use our spline representation to model human player movements relative to corresponding optimal paths. Using this representation, we show that simple distributions can be used to estimate control parameters to generate human-like movement across a population of agents in a novel environment.


Creating Playable Social Experiences through Whole-Body Interaction with Virtual Characters

AAAI Conferences

This paper describes work towards the goal of enabling unscripted interaction with non-player characters in virtual environments.  We hypothesize that we can define a layer of social affordances, based on physical and non-verbal signals exchanged between individuals and groups, which can be reused across games.  We have implemented a first version of that substrate that employs whole body interaction with virtual characters and generates nuanced, real-time character performance in response.  We describe the playable experience produced by the system, the implementation architecture (based on the behavior specification technology used in Façade, the social model employed in Prom Week, and gesture recognition technology), and illustrate the key behaviors and programming idioms that enable character performance.  These idioms include orthogonal coding of attitudes and activities, use of relational rules to nominate social behavior, use of volition rules to rank options, and priority based interleaving of character animations.


Predicting Army Combat Outcomes in StarCraft

AAAI Conferences

Smart decision making at the tactical level is important for Artificial Intelligence (AI) agents to perform well in the domain of real-time strategy (RTS) games.  This paper presents a Bayesian model that can be used to predict the outcomes of isolated battles, as well as predict what units are needed to defeat a given army.  Model parameters are learned from simulated battles, in order to minimize the dependency on player skill.  We apply our model to the game of StarCraft,  with the end-goal of using the predictor as a module for making high-level combat decisions, and show that the model is capable of making accurate predictions.


Evolving Playable Content for Cut the Rope through a Simulation-Based Approach

AAAI Conferences

In order to automatically generate high-quality game levels, one needs to be able to automatically verify that the levels are playable. The simulation-based approach to playability testing uses an artificial agent to play through the level, but building such an agent is not always an easy task and such an agent is not always readily available. We discuss this prob- lem in the context of the physics-based puzzle game Cut the Rope, which features continuous time and state space, mak- ing several approaches such as exhaustive search and reactive agents inefficient. We show that a deliberative Prolog-based agent can be used to suggest all sensible moves at each state, which allows us to restrict the search space so that depth-first search for solutions become viable. This agent is successfully used to test playability in Ropossum, a level generator based on grammatical evolution. The method proposed in this paper is likely to be useful for a large variety of games with similar characteristics.


Evaluating Planning-Based Experience Managers for Agency and Fun in Text-Based Interactive Narrative

AAAI Conferences

Artificial intelligence (AI) techniques have been applied to video games to make the overall experience more enjoyable. In games with interactive storytelling (IS), player actions can substantially affect plot events and plot characters. Therefore, AI planning techniques have been used to shape the plot inresponse to player actions that conflict with authorial goals. While such methods are poised to increase player fun andagency, two recent implementations (ASD and PAST) have not been formally evaluated to date. In this paper we do so via a series of user studies for the first time. We show that ASD significantly enhances fun and agency, whereas PAST gets mixed results with an interaction between effects of the experience manager and player prior gaming experience in one user study, and marginally significant results for increased agency in a study with a constrained story domain.