In the mid-1960s Joseph Weizenbaum created a stunning piece of software. Years before HAL 9000’s screen debut in 2001: A Space Odyssey, this software, Eliza, made it possible to have a conversation with a computer.
Eliza ran on a computer less powerful than the one in my mobile phone, though at the time it was one of the most advanced at the Massachusetts Institute of Technology. The interface was textual. Rather than being hooked up to a monitor and a mouse — or HAL’s microphone, speaker, and camera — Eliza was experienced through something like a typewriter, allowing the computer to type to the user and the user to type back. Yet, in the midst of all this now-primitive technology, Eliza was able to produce remarkable results. People could type anything they wanted to Eliza, and the software would respond in different ways depending on the currently-active script. Here is an example using Eliza’s most famous script, Doctor, which caused the software to parody the conversational patterns of non-directive therapist during an initial visit:
You are like my father in some ways.
WHAT RESEMBLANCE DO YOU SEE
You are not very aggressive but I think you don’t want me to notice that.
WHAT MAKE YOU THINK I AM NOT VERY AGGRESSIVE
You don’t argue with me.
WHY DO YOU THINK I DON’T ARGUE WITH YOU
You are afraid of me.
DOES IT PLEASE YOU TO BELIEVE I AM AFRAID OF YOU
My father is afraid of everybody.
WHAT ELSE COMES TO MIND WHEN YOU THINK OF YOUR FATHER
DOES THAT HAVE ANYTHING TO DO WITH THE FACT THAT YOUR BOYFRIEND MADE YOU COME HERE
(Weizenbaum 1966, 37)
Eliza/Doctor’s final line, above, is a reference to an earlier part of the conversation in which the user wrote “my boyfriend made me come here.” The resemblance between Eliza/Doctor’s “YOUR BOYFRIEND MADE YOU COME HERE” and the user’s phrasing is not coincidental. Each script for Eliza is actually just a set of linguistic tricks, one of the simplest of which is to take one of the user’s statements about herself and turn it back with appropriate word substitutions (in this case, “my” becomes “your” and “me” becomes “you”). Most of these tricks involve looking for key words (or small groups of words) in the user’s responses, such as the “are like” in “you are like my father,” above, that leads Eliza/Doctor to ask what resemblance the user sees.
But when we encounter the interface of a piece of software we don’t necessarily get a clear picture of how it actually operates internally. And many users of Eliza/Doctor initially developed very mistaken ideas about its internals. They assumed that, since the interface appearance of the program could resemble something like a coherent dialogue, internally the software must be very complex. Some thought it must be something close to the fictional HAL: a computer program intelligent enough to understand and produce arbitrary human language. This happened so often, and was so striking, that computer science circles developed a specific term for this kind of misunderstanding: “the Eliza effect.”
Play and the Eliza Effect
This paper is a brief look at the Eliza effect, and at two previously unnamed effects that can arise as we experience the interface of a digital system and build an idea of its internal operations. More specifically, this paper looks where others haven’t when exploring versions of this relationship: the area of play.
Weizenbaum may have originally thought of his system as a plaything — he certainly characterized the Doctor script as a parody — but his attention was soon drawn to another aspect of users’ interactions with Eliza. He came to focus on the conceptual mismatch that gives the Eliza effect its name, and specifically on how it could “induce powerful delusional thinking in quite normal people” (1976, 7). He wrote a book dedicated to demonstrating that the internals of computers aren’t magical and that we do ourselves a disservice when we assume that human beings are so mechanical that we could, or should, have our intelligence matched by computational machines.
Weizenbaum wasn’t the only one who saw the Eliza effect as important to address in understanding our relationship with computers. A decade after his book, Lucy Suchman published Plans and Situated Actions (1987), in which she sees Eliza/Doctor as an iconic example in human-computer interaction of what ethnomethodologist Harold Garfinkel (citing Karl Mannheim) has called the documentary method of interpretation. Specifically, that people tend to “take appearances as evidence for, or the document of, an ascribed underlying reality, while taking the reality so ascribed as a resource for the interpretation of the appearance” (23).
Yet another decade later, Janet Murray revisited Eliza/Doctor in Hamlet on the Holodeck (1997). Here she makes a crucial turn for our purposes, seeing the project in terms of media. She views Eliza/Doctor as a dramatic character, Weizenbaum as a kind of playwright, and the source of the Eliza effect in such cases as “the human propensity to suspend disbelief in the presence of a persuasive dramatic presence” (224). Here Murray makes visible what wasn’t clear in previous accounts: that people enjoy Eliza. Initial interactions are often quite fun and engaging, and create the sense of open-ended possibilities with an interesting character.
But Murray doesn’t go far into talking about ongoing interaction with Eliza/Doctor, just as Suchman and Weizenbaum don’t. If they did, it would complicate their discussions. This is because — while the initial experience of Eliza/Doctor can create the surface impression of an incredibly complex internal system — sustained interaction with the system, the verbal back-and-forth, invites play ... and linguistic play with Eliza/Doctor quickly begins to destroy the illusion. In other words, precisely the open-ended textual interaction that helped foster the illusion of internal complexity and intelligence enables play that draws attention to the system’s rote simplicity, its distance from human interaction. Before long, many exchanges begin to look like this:
You are going to repeat what I say in the form of a question
WHAT MAKES YOU THINK I AM GOING TO REPEAT WHAT YOU SAY IN THE FORM OF A QUESTION
For those of us interested in designing computationally-driven play experiences, especially those involving human characters, this presents a two-part problem. First, how can we represent something very complex — human beings — using tools that are inevitably quite simple by comparison: computational processes? Second, how can we engage audiences in playful interaction with these representations without the boom/bust of the Eliza effect?
Before addressing such questions, however, it is worth considering a different dilemma of digital authorship. While the Eliza effect produces a surface experience that can give the audience an initially-inflated view of system internals, many digital works present interfaces that fail to give audiences the opportunity to even appreciate the complexity of internal systems that are present. Rather than the boom/bust of play that the Eliza effect describes, these systems fail to produce an initial boom.
The Tale-Spin Effect
Like Eliza, Tale-Spin is a landmark of digital media. Created by James Meehan in 1976, Tale-Spin is the first major story generation program. It made the leap from assembling stories out of pre-defined bits (like the pages of a Choose Your Own Adventure book) to generating stories via carefully-crafted processes that operate at a fine level on story data. In Tale-Spin’s case, the processes simulate character reasoning and behavior, while the data defines a virtual world inhabited by the characters. As a result, while altering one page of a Choose Your Own Adventure leaves most of its story material unchanged, altering one behavior rule or fact about the world can lead to wildly different Tale-Spin fictions.
Tale-Spin can generate fictions with or without audience interaction. When generating with interaction, Tale-Spin begins by asking the audience some questions to determine the initial state of the world, especially the characters present in the story. Storytelling begins from these initially-established facts, with the audience consulted as new facts are needed to move the story forward. For example, once the characters are known and the world is established, Tale-Spin needs to know the identity of the main character:
THIS IS A STORY ABOUT ...
1: GEORGE BIRD 2: ARTHUR BEAR
After the audience chooses a character, Tale-Spin next needs to know the problem of this character that will serve as the impetus for the story:
HIS PROBLEM IS THAT HE IS ...
1: HUNGRY 2: TIRED 3: THIRSTY 4: HORNY
The opportunity for play, with Tale-Spin, lies in these audience choices — both in the telling of any one story and across multiple stories. Just as the audience builds up a mental model of Eliza/Doctor through unconstrained textual input and consideration of the software’s responses, the audience of Tale-Spin builds one through question answering and considering both further questions and resulting stories in the context of answers given.
When the audience makes its choices, Tale-Spin doesn’t simply record these facts about the world. In addition, internal Tale-Spin mechanisms draw “inferences” from the facts. For example, if it is asserted that a character is thirsty, then the inference mechanisms result in the character knowing she is thirsty, forming the goal of not being thirsty, forming a plan for reaching her goal, etc.
Some uses of inferences are relatively straightforward. It’s no surprise that a thirsty character will form a plan for not being thirsty. But other uses of Tale-Spin’s inference mechanisms can be quite surprising. For example, Tale-Spin characters can use its inference mechanisms to “speculate” about the results of different courses of action. Meehan’s The Metanovel (1976) describes a story involving such speculation, in which a hungry Arthur Bear asks George Bird to tell him the location of some honey. We learn that George believes that Arthur trusts him, and that Arthur will believe whatever he says. So George begins to use the Tale-Spin inference mechanisms to “imagine” other possible worlds in which Arthur believes there is honey somewhere. George draws four inferences from this, and then he follows the inferences from each of those inferences, but he doesn’t find what he’s after. In none of the possible worlds about which he’s speculated is he any happier or less happy than he is now. Seeing no advantage in the situation for himself, he decides, based on his fundamental personality, to answer. Specifically, he decides to lie.
This is a relatively complex piece of psychological action, and certainly tells us something about George as a character. But the interface appearance of a Tale-Spin story never contains any information about this kind of action. For example, here is a quote provided by Meehan from a similar moment in a Tale-Spin story:
Tom asked Wilma whether Wilma would tell Tom where there were some berries if Tom gave Wilma a worm. Wilma was inclined to lie to Tom. (232)
As we know from the tale of Arthur and George, a complex set of speculations and character-driven decisions took place as Wilma considered Tom’s request. But all that — probably one of the most interesting parts of this story, as it is simulated inside Tale-Spin — is lost in the gap between the above two sentences.
No matter how creatively one plays with Tale-Spin, such hidden action cannot be deduced from its interface outputs. This is probably why, though Tale-Spin is seen as a landmark in computer science circles, it is often treated with near-ridicule in literary circles. Critics as astute as Janet Murray, Espen Aarseth, and Jay David Bolter have failed to see what makes Tale-Spin interesting, focusing instead on what its output looks like at the interface.
Of course, while we can call this a failure of these critics, it is probably more accurate to describe this as a failure of Tale-Spin itself. While Tale-Spin’s author created complex and interesting internal processes, he failed to make that apparent at the interface level. While playing with Tale-Spin actually involves setting an intricate world in motion, the audience experience is blunt and repetitive.
This situation is far from uncommon in digital media, perhaps particularly in the digital arts, where fascinating processes — drawing on inspirations ranging from John Cage to the cutting edge of computer science — are often encased in an opaque interface. In fact, this effect is at least as common as the Eliza effect, though I know of no term that describes it. Given this, I propose “the Tale-Spin effect” as a term for works that appear, at their interface, significantly less complex than they are internally.
The Tale-Spin effect, like the Eliza effect, is not only a description of audience experience — it is also a warning to authors of digital media. Just as play will unmask a simple process with more complex pretensions, so play with a fascinating system will lack all fascination if the system’s operations are too-well hidden from the audience.
Luckily, the third effect I will discuss here is not a warning of this sort.
The SimCity Effect
In the mid-1980s, Will Wright created a landscape editor for authoring his first game, an attack helicopter simulation. Working with the editor, he had a realization: “I was having more fun making the places than I was blowing them up” (Wright 2004). From this the idea for Wright’s genre-defining SimCity was born.
Wright realized that interacting with his terrain editor was more interesting than interacting with its outputs. In a way this is quite similar to the insight offered by the Tale-Spin effect: let the audience play with the most interesting parts of the system.
SimCity, of course, unlike a terrain editor, doesn’t simply wait for a user to do something. Time begins passing the moment a new city is founded. A status bar tells the player what’s needed next — starting with basic needs like a residential zone and a power plant and, if play succeeds for any period, ramping up to railroads, police stations, stadiums, and so on. A budding city planner can lay out spaces, but it’s up to the city’s virtual inhabitants to occupy them, build and rebuild, and pay the taxes that allow the city to continue to grow.
As cities grow, areas respond differently. Some may be bustling while others empty out, or never attract much interest at all. SimCity provides different map views that can help diagnose problems with abandoned areas. Is it too much pollution? Too much crime? Too much traffic? Players can try changing existing areas of the city (e.g., building additional roads) or create new areas with different characteristics. Observation and comparison offer insights. Why is this commercial area fully developed, while that one lies fallow? The answer is always found by trying something different and considering the results.
In other words, the process of play with SimCity is one of learning to understand the system’s operations. Conversely, the challenge of game design is to create an interface-level experience that will make it possible for audiences to build up an appropriate model of the system internals. As Wright puts it:
As a player, a lot of what you’re trying to do is reverse engineer the simulation… The more accurately you can model that simulation in your head, the better your strategies are going to be going forward. So what we’re trying to [do] as designers is build up these mental models in the player… You’ve got this elaborate system with thousands of variables, and you can’t just dump it on the user or else they’re totally lost. (Pearce 2002)
Here, again, we lack a term for an experience. I propose “the SimCity effect” for this important phenomenon: a system that, through play, brings the player to an accurate understanding of the system’s internal operations. Of course, the SimCity effect is most important to consider in cases where the system is complex, but it applies generally. Pong works as well as it works because it effectively communicates at the interface level its quite simple internal operations.
What is exciting about the SimCity effect, and about Wright’s work generally, is that it helps us get at the new possibilities opened by working with computational media. Pong is very similar to games we play without computers, but SimCity is a more complex system than even the most die-hard Avalon Hill fan would want to play as a tabletop game. This ability to work with computational processes, to create complex computational systems, is the opportunity that digital media affords — and the SimCity effect points the way toward creating experiences of this sort that succeed for audiences.
Two questions were left dangling at the end of this paper’s discussion of Eliza. First, how can we represent human complexity using computational processes that are inevitably quite simple by comparison? Second, how can we structure play with these representations without the boom/bust of the Eliza effect?
A quick answer to these questions can also be drawn from the work of Will Wright, in the form of the best-selling computer game of all time: The Sims. This is not only the most successful game of all time, it is also a representation of human beings and their lives that successfully invites and structures play. It doesn’t attempt the freeform textual dialogue of Eliza/Doctor, but rather has its proto-characters speak gibberish while iconic representations of conversational topics appear above their heads. In this way, and in many others, it builds on the power of the SimCity effect: providing the audience with a surface representation and opportunities for interaction that are at the same level of complexity as the internal system operations. It doesn’t over-promise like Eliza/Doctor, and, unlike Tale-Spin, it translates the interesting complexity of its systems into audience experience.
Of course, many of us would like to play games which actually have linguistic content — in which characters actually say things in human language. Here it is useful to consider another aspect of systems: their appropriateness to what they represent. When playing an RPG such as Oblivion, we talk with other characters by activating them, hearing or reading lines they speak, and then choosing our own lines or topics of dialogue from a textual menu. This does a good job regarding the SimCity effect (the underlying system is just as simple as the surface representation) but this system is rather ill-suited to representing human characters. Each interaction with this system, each moment of play governed by it, is an abject failure compared with the smooth and compelling exploration of space provided by such games.
Given the mismatch between human complexity and most dialogue systems, how could we find a better solution? One answer, of course, is to follow Wright’s lead and keep pushing forward on the complexity of the systems. But this is not a viable solution for most game designers. Perhaps better guidance could come from another designer: Jordan Mechner. His Prince of Persia: The Sands of Time deals elegantly with the limitations of its dialogue system by never making them available for direct interaction. Instead, the audience plays with the systems that work well when governed by the SimCity effect, such as acrobatic movement through the game’s compelling visual spaces. Well-drawn characters exist, and speak dialogue, but their dialogue is driven by play with other systems. The main non-player character, Farah — and the player character himself — speak dialogue related to their current positions in space, progress on solving puzzles, and how these translate into forward movement in the story. The player can elicit responses from Farah by, for example, moving the player character in front of her and switching camera views to stare at her — but interaction remains firmly within the game’s systems that are well-suited to representing their subjects (movement, the gaze, and combat).
Of course, such approaches will never produce the excitement we can feel during the initial moments of play under the Eliza effect. But such experiences can remain compelling over long periods of play, and result in characters more engaging and well-drawn because they are not founded on quick-crumbling illusions.
Meanwhile, we can continue to explore more complex models for representing human lives in the territory opened up by The Sims. Or, as I do in my own practice, we can explore the potential for systems that enable textual play with literary language about human relationships, rather than restricting ourselves to play with iconic graphical representations. Or, like the notable recent independent game Façade, we can combine more complex internal models with greater attention to the crafting of audience expectations that Janet Murray calls “scripting the interactor.”
Aarseth, E. J. (1997). Cybertext: Perspectives on Ergodic Literature. Baltimore: Johns Hopkins University Press.
Bolter, J. D. (1991). Writing Space: The Computer, Hypertext, and the History of Writing. Mahwah, New Jersey: Lawrence Erlbaum Associates, Inc.
Murray, J. H. (1997). Hamlet on the Holodeck. New York: The Free Press.
Meehan, J. R. (1976). The Metanovel: Writing Stories by Computer. PhD thesis, Yale University.
Weizenbaum, J. (1976). Computer Power and Human Reason: From Judgment to Calculation. New York: W. H. Freeman