Menu

Research Experience

    Master's Thesis in in Data Science

    Generating Text-based Adventure Games

  • My thesis in Data Science for my Master in Science in Engineering degree is on the topic of Generating Text-based Adventure Games.

    The thesis presents a method and analysis to generating tools for text-adventure game authors using OpenAI’s GPT-3 API.

    If the reader is familiar with Zork, Enchanter, Anchorhead, or even Colossal Cave Adventure, text-based adventure games might already sound familiar. To be more specific, text-based adventure games are one of the oldest video game genres, and often considered to be their origin.

    Interactive Fiction games are fully text-based simulation environments where a player issues text commands to effect change in the environment and progress through the story.

    These games typically feature a text parser, a user interface that allows the player to interact with the game solely using typed commands. They also feature a storyline which is mostly linear, although there are multiple possible ways to reach a given ending, and deliberate puzzles within the games.

    Text-based games are a form of interactive fiction, the term used to describe a text-only computer game. The term refers to a story-based game that features only text, as opposed to a graphical interface. Text-based adventures were more popular before the introduction of the point-and-click interface, in part because they required fewer resources to implement.

    30 years ago, on the subject of fictional text generation, the following exchange occurred in 1992 between Morpheus Nosferatu and Phil Goetz on Usenet, the 1990s version of Reddit:

    From: goetz@acsu.buffalo.edu (Phil Goetz)
    Subject: Re: Adventure generators (skippable)
    Newsgroups: rec.arts.int-fiction
    Date: 29 Oct 92 04:40:05 GMT
    Sender: nntp@acsu.buffalo.edu
    Organization: State University of New York at Buffalo/Comp Sci morpheus@sage.cc.purdue.edu (Morpheus Nosferatu) writes:
    > Has anyone ever worked on, or even heard of, an adventure generator?
    >
    > I’m not talking about an adventure design language like TADS or Alan,
    > but rather a stand-alone adventure generator that produces complete
    > adventures, where the user need only give a minimal degree of input,
    > such as the level of complexity, type of adventure (mystery, treasure
    > hunt, etc.), size of adventure, and so forth?
    > ...
    > But as anyone ever heard of someone trying to come up with a generator
    > whigh would produce infocom-style text adventures? I can just imagine
    > what kind of limitations it would have, but I’m curious to know if
    > anyone has tried this, and if so what degree of success they’ve had.
    No. ... The generator you speak of is not written, not being written,
    and not anywhere on the horizon. In 50 years, maybe. In 20,
    definitely not. The problem of writing interesting stories, which
    adhere to someone’s definition of a plot (with goal explanations,
    conflict, resolution, comlication, climax, etc., all occuring at
    appropriate intervals) is very hard, and I don’t expect a solution
    soon. But the problem of writing clever puzzles involves much greater
    creativity, and I have seen NO evidence that ANYBODY has a clue in
    these creativity issues; the most you will find in the field are a
    few vague theories of creativity.
    This problem is what Stuart Shapiro calls "AI-complete": Solving it
    would be equivalent to solving all the other problems of AI.
    Phil

    This thesis shows that Phil Goetz’s time estimate was surprisingly accurate, in that it illustrates how 30 years after that discussion, language models in 2021 are in fact capable of generating creative, human-like fictional texts.

    The research was conducted under the mentorship of Postdoc Lara Martin as well as the advisement of Professor Christopher Callison-Burch and my thesis supervisor Professor Clayton Greenberg.