I’ve been hesitant to post about research-related topics on my personal website but I’ve decided I might as well since it might help me get more inspiration as I share this information with nameless people on the Internet. This is a very quick overview of Interactive Storytelling and the main computer science research that it builds upon. I’ve only touched on a few of the Interactive Storytelling systems since the four mentioned are ones I’m most familiar with.
I. A Quick Introduction
Interactive Storytelling has been described by multiple names: Interactive Drama, Drama Management and Interactive Fantasy to name a few. The main idea however stays the same: place the user in a virtual world containing possibilities for dramatic situations. The user interacts with the world revealing the story through interactions and modifying the story through the interactions. Interactive Storytelling can trace its roots back to story creation research, then to Brenda Laurel’s PhD dissertation “Toward the Design of a Computer-Based Interactive Fantasy System” in 1986, followed by Weyhrauch’s dissertation “Search-Based Drama Management” in the early 90s. From this point the paths diverge based on what the researchers wish to exploit. Some wish to have the story run in real-time working to guide the user to the optimal place within the word to witness an event to guide the user through the story, such as Magerko’s Interactive Drama Architecture. Others allow the user to move at their own pace while modifying the story to suit the user – like Mateas’ Facade, Fairclough’s OPIATE and Thue’s PaSSAGE.
II. Story Creation
Story creation research attempts to have the computer create a coherent story based upon a set of logical methods. The initial research into the area was Mehan’s TALESPIN. TALESPIN used a logic solver to solve a set of logical statements then using these logical steps to author the story. For instance, it would define Bucky the squirrel as being hungry, a tree with a nut within it, a chair which Bucky could move and the fact eating a nut would solve Bucky’s hunger. The solver would then solve to make Bucky the squirrel not hungry by having Bucky see the nut within the tree, moving the chair to the tree, hopping on the chair to grab the nut and then eating the nut. These steps could then be turned into a simple story. TALESPIN was able to create procedural stories that described actions the characters were taking and complete it upon that, however, it was missing internal rationalization, metaphors and other common literary techniques.
Minstrel was the next notable entry into story creation. The researcher, Turner, believed that creativity could be approximated by using a case-based reasoning system to write the story so when a situation where no experience was recorded, it could create a response based on similar cases within the system. Minstrel used meta-level goals and plans that described what the author was trying to achieve then views the writing of the story as a problem solving exercise by invoking the case-based reasoner.
The latest research in story creation is BRUTUS. BRUTUS creates tales of betrayal and heartbreak using many knowledge base sources in order to change the setting, characters, betrayal type and tone of the story. It’s overall goal is to create a story that is sufficiently different from its initial logic definitions and knowledge representation.
III. Interactive Storytelling
In 1986, Brenda Laurel proposed a system that could handle a first-person interactive virtual world. Her dissertation listed out 13 functions that must be handled by any system in order to create a working interactive fantasy world; but, proposes no way to solve the problems and in the words of Chris Crawford “it is a wish list and not a plan.” This was, however enough to set the foundation for Search Based Drama Management (SBDM).
SBDM was the first attempt at controlling an interactive story such that the story is rearranged to better tell the story according to a set of author-defined metrics and an algorithm that can effectively guide the plot. A Drama Manager used an adversarial search algorithm (based off of a min-max game tree algorithm) to determine the best move for the drama manager to execute based on the user’s moves. A drawback to SBDM is that what the author considers a good story is not always what the user considers a good story.
Facade is widely considered to be the first fully realized interactive drama system. It offers a complete, real-time, first-person dramatic experience with a character-driven story. Facade’s story is broken down into beats. The beats tightly integrate the story and actions for the AI controlled characters and are the smallest unit of plot, such as a line of dialog and the associated reaction from an AI-controlled character. A director agent coordinates the behavior of the AI characters with the beat enacted and makes the decisions on which beat to enact next.
OPIATE is much different than SBDM and Facade since it uses no direct input from an author. The user’s actions are the sole motivator in determining how the emergent story flows. The virtual world contains non-player characters (NPCs), items and different settings. The NPCs have an array of likeness variables that describe how the NPCs perceive the user (sometimes called Player Character or PC) and the other characters. The NPCs also have a simple vocabulary that allows them to gossip with one another and affect each other’s perceptions of the other characters. During play, the case-based reasoning system chooses an encounter based upon the state of the characters and items within the world. The reasoner compares each character and item in the world for the given roles specifed in each encounter and determines a total suitability value for the encounter. If no suitable encounter is found, up to three encounters can be combined to create a new encounter to satisfy the threshold value for an acceptable encounter. Once an encounter is selected, the NPCs that best fit each role in the encounter are given their roles within the encounter and the user is then engaged by the appropriate characters to begin the quest.
The final system I’ll describe is PaSSAGE. PaSSAGE, like OPIATE has an emergent story. Users progress through the world interacting with characters and completing encounters in order to be given a new one. PaSSAGE uses a user model that tabulates the way in which the player is interacting with the world, as quests are completed the user model is used to determine the next quest the user will be given.
This quick overview of Interactive Storytelling will hopefully give you a little insight into what Interactive Storytelling is and some of the current research in the field.
Feel free to leave questions and comments (but no concerns!) and I’ll either update the page or answer the question in the comments section.