Scott Turner: The Creative Process

[Readings] (01.23.09, 3:16 pm)

Scott Turner is most notable for his work on the Minstrel storytelling system. Minstrel is notable in terms of storytelling systems because it is one of the first, long with Liebowitz’s Universe, to employ a model of authorship and an author’s plans. Minstrel is also notable for its elaborate system for implementing creativity. The work is very tightly bound up in the rhetoric of traditional AI. The model of creativity works in terms of problem solving, and uses analogical reasoning to cast problem solving strategies from one domain into another.

While my work is not aimed at story generation, and my approach is very different, Turner is a useful perspective to keep in mind in terms of drama management, authorial goals, and creativity within story worlds.

Storytelling and Creativity

The problem as initially framed is an issue of story telling. Turner is interested in developing a computer program that can tell stories. Originally, this research began with finding a copy of Vladimir Propp’s Morphology of the Folktale, and then experimenting with the recombination of stories using Propp’s formal structure. Turner quickly found that Propp’s morphology, while it may be useful for understanding stories, is insufficient to instruct a computer to tell them. To tell or understand a story, it is necessary to have more than form, but a large network of information: author goals, reader expectations, and cultural knowledge. Turner turns his attention not to the makeup and content of stories themselves, but the process by which an author creates a story.

Developing stories involves thinking about the purpose and the message of the story, in addition to its form and content. This is one of the more innocuous claims that Turner makes, but it is arguably somewhat controversial. His goal is to rule out the idea of nonsense stories which do not employ causality or simply do not reach any kind of satisfying conclusion. However, many stories have messages, but those messages may not be clear or straightforward at all. They may be messages that require a great deal of interpretation, leading to diverse ranges of valid interpretations. Turner’s examples are simple, tending to employ very simple moral messages. In his defense, we have to start somewhere. It’s not possible to leap straight into a storytelling system that can write something along the lines of À la recherche du temps perdu without being able to write something simple first.

Storytelling requires a strong knowledge of not only the story, but the world in which the story takes place, the meanings of the terms and elements that occur within it. Minstrel uses an Arthurian world, where it is necessary for the storytelling system to understand what dragons and knights and princesses are, what it means for a knight to charge or be wounded, and what characters might be liable to do after some event. This does not mean understand in the sense of exact definition, but rather have a functional understanding of how these work within the story world.

Storytelling also requires creativity. This is the constraint that Turner issues which is the most remarkable. Creativity requires both the judgement of creativity, as well as an ability to be creative in the first place.

Minstrel’s architecture employs a problem solver, and treats the process of developing a story as a problem to be solved. Turner does not see this as a specialized process, used for scientific endeavors, but rather an everyday one. He claims that problem solving is invariant across problem domains, so the same problem solving method may be used in any domain, from astrophysics, to navigation, to grocery shopping. This argument is contestible, but through using it, he is able to make an interesting observation about how to use creativity. Creativity is the process of using knowledge or a method from one domain and applying it to another. In planning terminology, this involves an integrated process of search and adaptation. Experiencing a problem in one domain, the planner searches through other domains to find structurally similar problems, and then adapt them into the original domain. The interesting thing about this, for all the critiques of AI planning, is it is a method of thought that exists outside of conventional situational thinking, and is a reasonably effective means of producing creative solutions to problems.

The architecture that Turner uses involves extensively making use of author goals. There are four kinds of goals: thematic goals, consistency goals, drama goals, and presentation goals. In addition to employing creativity, the story must be able to satisfy these goals according to the author’s needs. Consistency goals are about producing consistent and causally sound stories, drama goals involve satisfying constraints to make the story dramatically interesting, such as having foreshadowing, suspense, and so on. Presentation goals aim to make the story clear and legible to the reader. Thematic goals are interesting to me, and I will address those in the next section.

A Model of Storytelling

Turner argues that storytelling is a matter of satisfying author goals. People write stories intentionally, for deliberate reasons. As such, the stories themselves have goals within them. To illustrate the importance of author goals, Turner looks at Meehan’s Talespin. Talespin does not make use of authorial intentions, and instead has only character plans and intentions. The Talespin stories are generally not too good, meandering and lacking purpose, and occasionally getting into infinite loops. Turner argues that storytelling requires more than mere simulation, and that authors are not merely simulators.

My goal in the adaptation project is not to tell stories, but make games. The challenge to simulation puts me on the defensive, but it is necessary to acknowledge that in order for there to be stories, there must be more than simulation alone. I would argue that what is missing from Talespin is some sense of values or meaning within the world. The example of the bear going to get berries is boring not because there is no authorial meaning, but because the actions and events are not meaningful to the reader.

One way of dealing with the matter of author goals I suppose is to challenge them in favor of reader goals. Authors may write bad stories that meander or go nowhere. The author may have goals which are uninteresting or nonsensical. The author may value these goals, but the readers may find them bewildering. This is not unusual, people write bad stories all the time. Merely having author goals is not enough to make a story interesting. The life of a story is not dependent on the author who writes it, but the community of readers who value it. This is a very important observation that should be made with literature of author goals in story planning.

The actual method of story generation in Minstrel is developed by planning and problem solving. The planner manages the author-level goals, and the problem solver works to find means of solving those goals. At the top level, Minstrel’s goal is to “tell a story”. As this proceeds, Minstrel will choose a moral, a theme, and then begins applying drama goals, performing consistency checks, and making the story presentable. It does this cycle for each scene in the story. This approach to story development is extremely top-down. The work is pioneering, but it is distant from the actual method by which a writer may actually compose a story. Writing involves iteration and revision of the entire work in cycles. I would argue that when the goals are established and some of the basic elements are introduced, the writer does perform simulation, to see what happens next. Minstrel achieves causality by looking backwards from an event and seeing what could have caused it. A real author will employ some of this approach, but will aslo use simulation to simply get causality by advancing time and playing it out.

The model presented is that stories have themes, where a theme is some sort of moral “lesson”. This is a contestible account of the reason behind storytelling, both within stories and of the reasons for writers to create stories. By employing diverse lessons to other storytelling domains, this suggests that the theme of a story and the world of a story can be separated unproblematically, which is false. Stories are culturally anchored, and themes may be conventionally bound to one set of story worlds. An example in the realm of Aesop’s Fables, which contain the kinds of lessons Turner is interested in, is the story of the And and the Grasshopper, which has a remarkably different ending as told in the West versus in Japan. Western culture is much more individually oriented, and focused on independence, whereas Japanese culture relies on interdependence and mutual support. Lifting themes from one domain to another may be creative, but it may also run the risk of cultural imperialism when done unawares.

Reading Info:
Author/EditorTurner, Scott
TitleThe Creative Process
Tagsspecials, digital media, narrative, simulation
LookupGoogle Scholar, Google Books, Amazon

1 Comment »

  1. Sounds like an interesting book :-)

    Comment by Scott Turner — February 3, 2009 @ 4:28 pm

RSS feed for comments on this post. TrackBack URI

Leave a comment

You must be logged in to post a comment.