Living with Ideas #3

September 23, 2003

A metaphor is a powerful device for making and sharing meaning. I love stories, and a metaphor is often a very narrative device: it tells a story about an idea, and uses pressures and tension to express the nuances. I read Patricia Miller's introduction to Information-Processing Theory with great interest, because this theory illustrates the power and risks of a compelling metaphor: the mind as an information-processing system, as a computer.

Throughout my childhood and adulthood I have developed a series of metaphors for my thinking and learning. These metaphors are increasingly sophisticated, as a direct result of my exposure and mastery of increasingly sophisticated technologies. My metaphors have included a coloring book, a film strip, a private library, a Macintosh desktop environment, a Web site, a relational database, and (most recently) an expert system. By expert system, I mean a self-organizing, problem-solving set of programs and databases, with an "executive level" of control. So while I've had little exposure to formal information-processing theory, it clearly comports with my intuitive and experiential beliefs about the nature of thinking and knowing.

(When my warrant for an argument is intuition or experience, it's partly a self-scolding. I expect more objective warrants in others' models, so I should demand the same from myself!)

Information-processing theory appeals to my behaviorist foundation, with its metaphors of input and output. This is just a metaphor, as Miller reminds us: "The circuitry of a computer is quite unlike the anatomy of the brain." (p. 234) But my fascination with human minds and computer technology is based on the same paradigm: a complex, often inscrutable system that nevertheless discriminately responds to specific input. My background in system dynamics gives me an appreciation and literacy with the modeling approaches of this theory. And in many respects, this theory is not far removed from behaviorism, although it clearly allows for more narrative about learners monitoring and influencing the development of associations. My background in logic and programming increases my appreciation and literacy with information-processing theory, especially the "boxes and arrows" mode of analysis. (p. 243) All this leads to a potential happy triumvirate in my developing framework for using computers games in education:

Such a framework combines a behaviorist's focus on observable outcomes (e.g. player behaviors in a game), attributed to metaphorical, internal input-output "programs." Each step of theory, methodology, and reasoning should be based on rigorous logical induction and deduction. Notably, I separate logic from i-p theory, because a theory based too much on the state-of-the-art in logic (and computer programming) might unnecessarily exclude certain theoretical explanations of thinking and learning. Likewise, as with any theory, I judge i-p and behaviorist metaphors partially on their logical adherence to objective observation.

In my Living with Ideas #2, I expressed great affection for Chomsky's quote: "We understand a new sentence, in part, because we are somehow capable of determining the process by which this sentence is derived in this grammar." Miller elaborates on this intriguing idea: "These rules are unobservable and must be inferred from the relations between language input and output." (p. 238) For me, this is the most exciting and compelling idea in game studies. As a player in a multiplayer role-playing game, I can't precisely know the informal model of motivations and beliefs another player is using to animate her character. So I make inferences and build a surrogate model in my mind. Building such models helps develop empathy, which is a critical part of my long-term inquiry into how to use games to teach "emotional intelligence," group dynamics, and leadership. It's similar to traditional puppet therapy (which was satirized in the film What About Bob?), in which subjects are forced to distance themselves from an interpersonal relationship, while simultaneously striving to intimately understand the needs and desires of the other person. Ironically, the extra layer of play and projective identity sometimes seems to clarify the interpersonal dynamics. In other words, by playfully interacting through imaginary characters, I have learned a lot about my own and others' "real" selves.

(My ironic "real" is another form of scolding. It's directed at the cultural mindset that games aren't real, especially in the sense that an interpersonal relationship between players in a game is somehow intrinsically inferior to a non-game relationship.)

Jodi Buchanan (my wife) has developed a similar theory, based on her experiences and observations as a player and leader in online games. She believes that two people (e.g. Joe and Samantha) may sometimes relate more intimately and empathically through their intermediaries: their characters, Thesius and Diana (see figure below). Their actual communication may be covert, encoded in the dialog between fictional characters. By building a surrogate model of Diana, Joe develops more empathy for Samantha. The process of building a model is the "somehow" that attracts me to the Chomsky quote.

In a similar fashion, by constructing theoretical and computer models of thinking and learning, information-processing psychologists are learning about real human capacity and functioning. While Miller criticizes the theory for being overly focused on individual cognitive activity, both the i-p researchers and their subjects extend their capacity using tools, a very situative paradigm. For example, a subject's capacity to seriate blocks by weight is dramatically increased with access to a two-pan balance. (p. 242) Also, many i-p researchers depend on computer models for understanding, since they're incapable of visualizing so many processes in their head. In much the same way, Joe's capacity for empathy is extended through having the character Thesius.

Another example of this situative perspective is the state abbreviations problem. (p. 270) "R.I." is declared to be the better answer, but this largely depends on familiarity with the United States, particularly New England. The "better answer" for those not familiar with the region might be N.D., since it includes an "N" and doesn't include a vowel. Here and elsewhere, some of the "results" of p-i are questionable, because the experimental methodology is mediocre.

I smiled at the problem of finding socks in the dark. (p. 271) Red herrings (e.g. "a ratio of 4 to 5") are a strong tradition in computer games.

I shall read Dodge (1986) for its account of information-processing in peer interaction.

A fundamental assumption of i-p theory is that the human mind is a self-organizing, self-correcting system. This is valid for me, intuitively and experientially. The rules assessment and measurements of external simulation fidelity are appealing methodologies, and somewhat similar to the behaviorist methodologies that appeal to me. On the other hand, the assertion that some acts of recall "just happen" is bizarre, as it seems to depend on my (nonexistent) consensus about the existence of innate motivation. (p. 251) I'd prefer a description derived from the arguments about strategies, that some strategies are eventually automated. Describing "recall as a by-product of a meaningful activity" is a semantic mess. Ironically, this illustrates one of Miller's most important criticisms: over-dependence on a metaphor can lead to unintended semantic and theoretical messes. Likewise, we shouldn't mistake the internal coherence of a model for objective knowledge of the real thing: a simulation with excellent external fidelity doesn't necessarily track truth. We may laugh at our ancestors' notion that the sun orbits the Earth, but based on the limited tools and data they had, it's a reasonable model.

Overall, information-processing theory is very appealing to me. A significant part of this appeal is the strong tradition of diagrams, a semantic tool which I clearly value (as I've demonstrated with my figures herein). I also like the cross-pollination with computer science and especially artificial intelligence. For example, "scripting" has similar meanings in i-p theory and game design (and theater): creating believable sequences of interactions between people/characters and their environment. This exploration has confirmed my fear that I'm undertaking some kind of Hegelian nightmare: trying to build a model of how a player builds a model of how another player builds a model of her character.