Reviewed by
Howard Soloman
Northern Illinois University
November 1997

Murray, J. H. (1997). Hamlet on the Holodeck, The Future of Narrative in Cyberspace. New York: The Free Press.

 

Introduction
A review of a book is best understood if the context within which it is written is made clear. This context is that of an assignment for a graduate level course in Interface Design for Educational Software. Had the book been read in another context, areas that are reviewed would have changed. This is particularly true when the book has not been specifically recommended for the course, and when the reader's expectations of its content have not been met. Under these circumstances, the reader can take one of two approaches: 1) dig into the book for material that, while not specifically oriented toward the subject matter, can be seen to apply to the overall field; or 2) give up and try another book. The first of these approaches is utilized here.

Murray's subtitle, The Future of Narrative in Cyberspace, indicates that the book is about narrative more than it is about software interface. However, a story will be successful when people choose to stay in touch with it. The book that can't be put down has drawn from its reader a large commitment of time and energy. Successfully designed software can similarly measure its success in the time and energy that the user is willing to commit to it. Insofar as Murray's analysis of narrative can turn up features that help cause the users to commit this time and energy, these same features may well be the ones that should be invoked in successful software design.

This review will use Hamlet on the Holodeck to point to Murray's research and conclusions about what has been done. It will state her conclusions about what was learned from having done it. It will try to analyze how these conclusions should be used while considering various interface choices. And it will try to show some places where the interface considerations can be used to achieve educational goals.

Section 1: What's been done?
Since language was invented, people have been trying to use words to tell things to each other. Regardless of which culture one is in, before it had writing as a tool, it had a long tradition of spoken information transfer. And from the beginning of that spoken tradition, it was split between language that directly passes the information and that which does it relating stories about people. The fables of Aesop are part of this long storytelling tradition, as are the epic poems of Homer. The content of "The Fox and the Grapes" carries a message to not be too greedy. But it also carries a joy associated with the act of storytelling. There is clearly a difference between commanding and telling a tale. The listener will far more often enjoy the tale. This is the advantage of the narrative form of language. A listener to a narration is willing to stick with the listening for as long as the narration is interesting.

The written word, which came to replace the spoken word as the dominant media form, continued the tradition of separating language into direct information presentations and language presented for its entertainment value. By the time of Shakespeare, oral presentations of written language were a popular entertainment form.

As radio became a more widespread media device, entertainment became more a part of its staples. A famous presentation of H.G. Wells, The War of the Worlds set a standard for believability. The strengths and weaknesses of the medium were exploited to the effect of causing people to flee from northern New Jersey. The credibility of the medium that had reported the sinking of the Titanic was not called into question.

Television has produced believable stories about the peasants of Italy reaching into the Semolina trees to harvest the annual crop of spaghetti. Films like This is Spinal Tap have chronicled the rise from obscurity of a rock band that never existed.

People are not stupid or easily fooled. They are just ready to believe what the media invents for them. Murray's position at MIT gave her ample opportunity to know of some of the major research projects that create people by using a computer. Her most prominent example is of Joseph Weisenbaum's 1966 computerized psychotherapist program, ELIZA. Hamlet on the Holodeck provides several examples of what a text-based conversation between a patient and ELIZA look like. Many people were anxious to suspend their disbelief in ELIZA's existence and treat her as a real character. Some people requested appointments. Some received bills for her services that they actually paid. (p. 71)

Following in the tradition of ELIZA, Michael Mauldin of Carnegie Mellon University invented a "chatterbot" named Julia who frequently appears in MUDs (Cyber role playing stages used simultaneously by multiple participants). Julia plays the role of being a gossip and a flirt so well that other participants have spent long stretches trying to get her to leave the mud and converse exclusively with them.

Murray has conducted classes at MIT in which the students are given an opportunity to write a Julia-like character. These have included very convincing samples of a girlfriend, a "folksy gruel eating grandpa, a smug pretentious artist, and an overly aggressive salesman." (p. 219-221)

All of them interact convincingly with live interactors, if only for a short while.

Section 2: What Was Learned from Doing It?
The key to a successful narrative is its ability to create for the interactor a state of immersion. "When we enter a fictional world," Murray claims, "we do not mere 'suspend' a critical faculty; we also exercise a creative faculty. We do not suspend disbelief as much as we create belief. Because of our desire to experience immersion, we focus our attention on the enveloping world and we use our intelligence to reinforce rather than to question the reality of the experience." (p. 110)

Murray sides with the "reader response" theory of literary criticism that holds that the act of reading is active rather than passive, that we "cast people we know into the roles of the characters, perform the voices in our heads," and "assemble the story into the cognitive schemata that make up our own system of knowledge and beliefs." (p. 110) Murray goes on to analyze the participant in a narrative experience as desiring not only immersion, but also agency. This she defines as "the satisfying power to take meaningful actions and see the results of our decisions and choices." (p.126) She does note that, however desirable it may be to experience agency, we usually do not experience it in a narrative environment. A film will reach the same ending without regard to the feelings we project about the characters in us. It's possible, in the time that follows a particularly good film, to induce a "Walter Mitty" effect and carry our invented roles with us outside the environment of the film.

Section 3: Interface Considerations
When dealing with the digital world there are more possibilities for interaction. The click of a button for the act of making a choice is the user being a significant agent in what the program accomplishes. Agency is more meaningful in open environments like MUDs where, although there is a basic story line and setting, individual characters get to act in the story and influence where it is going.

When she moves the discussion from the traditional to the digital world, Murray brings up a third area of event that users desire. She calls this "transformation." Of this she says that "the transformative power of the computer is particularly seductive in narrative environments. It makes us eager to pick up the joystick and become a cowboy or a space fighter, eager to log onto the MUD and become 'BlackDagger' or 'ElfGirl'. Because digital objects can have multiple instantiations, they call forth our delight in variety itself." (p. 154) When involved in this type of environment, participants can instantly be whatever they want to be.

Immersion in the digital fiction, agency in the way the plot plays out, and the instant ability to change all must be held within some bounds or the experience will deteriorate into chaos. If a participant is in the role of an elephant, somebody must establish that elephants cannot sprout wings without having first eaten magic plants. Complete lack of rules for one participant spoils the agency that other participants are enjoying. And the rules must be consistent. Familiarity with the environment and its conventions promotes the comfort of the participants.

In her discussion of why people treat ELIZA as a real character, Murray wants to know "What was the representational force that allowed the computer to bring her so compellingly to life?" (p. 71) Her answer is to explore four principle properties of the computer as a vehicle for literary creation. These are that "digital environments are procedural, participatory, spatial, and Encyclopedic." (p. 71) Murray goes on to elaborate the meaning of each of these properties while using appropriate examples from the MIT research.

The procedural aspect of digital environments consists of programmed algorithms that describe the way real things or people behave. The procedural aspect of ELIZA, for instance, was an engine built to respond in standard ways that resembled those used by Rogerian psychotherapists. Besides incorporating set responses to patients' "key words", the ELIZA engine was programmed to give greater weight to certain categories of key words. This feature made the ELIZA engine respond differently based on what the patient was saying. The patient perceived this to be sufficiently like somebody actually listening to believe that ELIZA was a real person. "The lesson of ELIZA is that the computer can be a compelling medium for storytelling if we can write rules (procedures) for it that are recognizable as an interpretation of the world." (p. 73)

To elaborate what is meant by the digital world being participatory, Murray turns to another research project that evolved into the game of "ZORK". Players involved in this game were amply provided with "opportunities to make decisions and to dramatically enact the results of those decisions." (p. 77)What Murray finds important about ZORK is that "the fantasy environment provided the interactor with a familiar role and made it possible for the programmers to anticipate the interactors' behaviors. By using these literary and gaming conventions to constrain the players' behaviors to dramatically appropriate but limited sets of commands, the designers could focus their inventive powers on making the virtual world as responsive as possible to every combination of those commands." (p. 77)

While Murray could have chosen spatial metaphors such as navigation into different parts of a graphical user interface or the different levels of MYST to illustrate the spatial property of the digital environment, she continues with another example from ZORK. This time she describes a scene in which the interactor has entered an area through a trap door, lit a lamp, and then is informed by the computer that she hears the trap door being slammed shut and barred. The words on the screen that describe this are sufficient to assure that the player knows that the trap door is no longer a possible direction in which to go. "The interactor's navigation of virtual space has been shaped into a dramatic enactment of the plot." (p. 83)

The fourth property of a digital environment is its encyclopedic nature. A CD-ROM, if it were holding only textual information, would have a capacity equal to approximately 650 books. (The upcoming DVD-ROM will have a capacity of around 10,000 books.) The Internet makes it seem as if "all the world's resources seem to be accessible, retrievable, and immediate. It is a realm in which we easily imagine ourselves to be omniscient." (p. 84) Here, references to the fan culture that surrounds some television series helps to illustrate the point. With multiple fan contributions and updates on a daily basis, complete universes of discussion about plot twists present a more real universe than the series can create on its own. "What ifs" are explored in more combinations than would be available if the series stayed on television for centuries. Fans can have instant access to such documents as birth certificates, marriage licenses, and income tax reports of characters who do not really exist. The presence of this encyclopedic volume of information about the characters makes it possible for new viewers to join in and catch up, as well as enhancing the overall impression that the characters actually exist.

The successful digital experience will take advantage of the "procedural, participatory, spatial, and Encyclopedic" nature of the environment. If it consistently does anything less, the user may as well migrate to a book or movie. The users want the computer to behave as if it were a real human, but without the flaws of human nature. The success of a character like Julia is a product of "responsiveness and appropriateness to her surroundings," which she uses to "induce people to collaborate with her in little dramatic scenes." (p. 214) People will participate in scenes with Julia because of how real she seems to them within the environment where she's found.

Julia's success in conversation stems in part from her being a query system. This is "a form of artificial intelligence program that takes in users' questions and consults a database to generate appropriate answers. In order to make query systems succeed, one must limit their domain of expertise and then anticipate many ways in which questions are asked." (p. 217) Have the characters appear in places where people expect them to behave like they actually behave.

Another advantage that Julia has in creating a aura of believability is that "when she appears on a MUD (a kind of Cyber role-playing stage for multiple users), her interlocutors are also in character. They come to the conversation predisposed toward dramatic collaboration." Murray concludes from this that "a successful chatterbot author must script the interactor as well as the program." (p. 219) Anticipate the predispositions of the program users and it becomes possible to get away with less information in the database.

As a result of the classes where she has given students an opportunity to write chatterbot characters like Julia, she has consistently found that "the most successful characters have been those who are self-absorbed, evasive, or obsessive in familiar ways." (p. 219) These characteristics allow the author to effectively disguise the limits of the universe to which the character is appropriate.

Section 4: What Does This Have to Do With Education?
As we look at interface design for educational software, it is not unusual to fall into a trap of supposing that the education for which software will be designed will be part of what Benjamin S. Bloom, in his Taxonomy of Educational Objectives (New York: David McKay Co, 1956) calls the cognitive domain. If we do so, we ignore the potential of the computer as a device to encourage learning in the affective domain.

Consider that there are four main strategies historically incorporated into affective domain education. These are: precepts, exemplars, rituals, and environment examples. (Harry S. Broudy, B. Othanel Smith, Joe R. Burnett, Democracy and Excellence in American Education, (Chicago: Rand McNally, 1964), p. 121-123. The first two of these strategies can be significantly enhanced by taking advantage of the properties of digital environments.

Take, for instance, the simple precept "A penny saved is a penny earned." The learner of this can be put into a participatory environment in which she is provided with some money, opportunity to save, and temptation to spend. An engine can be built that develops consequences of her choices in various situations. She can be in a position to replay different scenarios and explore different consequences. She can come into contact with various role models who can give her advise about saving. She can replay again and see new consequences in light of the new advice. She can reach her own conclusion about the wisdom of following the precept. This experience doesn't exist yet, but it can be easily built. It uses the digital environment to allow the learner to explore and learn on her own that the precept is surprisingly more wise than she would have thought.

Exemplars can also be made more accessible through the use of digital media forms. If Abe Lincoln exemplifies honesty, enough information can be gathered about Lincoln's life to model a character, mostly fictional, but based on Lincoln history. The success of such a device will depend on creating an environment in which the interactor can meet, watch, speak with, and learn from a very accurately modeled Lincoln like character.

As a character from history, Lincoln's responses to various events in his world are a matter of historical record. These could be the core of the database that the Lincoln algorithm could use to assure that the computer responds to all conceivable inputs as if it were Lincoln. It may even be possible to build an engine powerful enough to extrapolate the database to predict how Lincoln would respond to events that he never encountered.

Often, when considering interface design, we are looking for prescriptive advice about how best to get the learners to absorb the facts and procedures under consideration. The attractiveness of a screen and the clarity of a spoken voice are definitely aids in this endeavor. If screen and voice improvements are the objects of our quest, Murray's Hamlet on the Holodeck is not of much use. But if we allow that storytelling has a place in educational processes, then we need to concern our quest with the general features of quality stories, as well as how computer users can best be involved with them.

Several analyses of media choices suggest that film, video, and novels are the best media to use when attempting to develop attitudes. These forms all share a relatively long term ability to hold attention. This is the traditional heart of narrative. Moral messages have been included in narrative since the form began. Its history include everything from bible stories, the epics of Homer, and the works of Shakespeare, to television shows ranging from "Leave It to Beaver" to "Star Trek". All of these entertain the participant while imparting, however well or badly, instruction in the moral sphere. Their success in imparting the message is a function of how well they entertain. And it is here that Murray's Hamlet on the Holodeck makes its contribution.

If we turn for a moment to the bible as an example of a media device designed to influence behavior in the moral sphere, we can compare the readability of the books of Genesis and Leviticus. Leviticus feeds its readers a steady diet of prescription after prescription of how to live. Most readers would rather hear it from the stories of Genesis. The message is present. But it has a human character to it. Whether that character is Abraham or Lincoln, learners will succeed better if the character invokes the qualities that make a person believable.