The User-Friendly Fallacy
Fred Kemp
(Originally printed in College Composition and Communication, Vol. 38, No. 1, February 1987. 32-39.)
[used by permission of the NCTE]
Why is it, after more than seven years of imaginative and innovative work by Hugh Burns, Helen Schwartz, Valerie Arms, Cynthia Selfe, William Wresch, and Dawn and Raymond Rodrigues, computer-supported writing instruction in labs and classrooms across the country remains dominated by grammar drill, text analysis, and word processing? Whatever the merits of such computer support, if they continue to signify in the popular imagination all that a computer can do, then the computer in writing instruction is doomed to being labeled superficial, remedial, and reductive. It will be relegated to the "catch-up" curriculum, or paraded in showcase research labs which produce data and case studies but do little of what the computer should be best at, providing powerful, self-paced writing instruction across the educational spectrum and at every ability level.
There is, however, a type of software pioneered by Hugh Burns at the University of Texas in 1979 which can provide the student writer a powerful stimulation to creativity and originality, and if properly developed and applied, may demonstrate even to the hard-core skeptic that microcomputers can do much more than simply cosmeticize text or drill students in the avoidance of comma splices. But this type of software has remained in a prolonged "infancy," largely as a result of an attitude which I have termed the "User Friendly Fallacy." The User-Friendly Fallacy stipulates that for a computer to provide truly significant help to the student writer, it must (1) produce the illusion of actual person-to-person interaction, and (2) evaluate the content and quality of student input. Without these capabilities, according to critics like Robert Hertz, computer-assisted instruction's role "seems limited to those facets of composition that are clearly delineated, such as sentence structure or the mechanics of grammar usage" (62-64).
The day will come, of course, when computers will be able to employ natural language ability in order to replicate human behavior and judge to some degree the meaning and quality of inputted text. That day will not be soon, however, and even then such powerful computers and software will be avail able only to those institutions which can afford them. Roger Shank, who supervises Artificial Intelligence research at Yale, is one among many who hold out the promise that computers will someday be able to truly understand and evaluate what we say to them, and William Wresch, in The Computer in Composition Instruction, sees this as one hope for the future of CAI in writing ( 10). But Schank, in The Cognitive Computer, calls this capability "a considerable way off" (226), and makes clear that it will not appear on the microcomputer or minisystems for possibly much longer than that. What is needed now are programs that operate on commonly available microcomputers and that provide guidance in the generation and interrogation of ideas. Can computers, which do not understand in any real sense the student's input, actually do this?
Hugh Burns' TOPOI
The answer is "yes," as Lt. Col. Burns, currently director of the Air Force's Intelligent Systems Branch in San Antonio, has shown. In the late 1970's, while working on his doctorate, Burns produced an application for computer support of writing instruction quite unlike anything that had preceded it. His software and procedures are thoroughly described in his dissertation, "Stimulating Rhetorical Invention in English Composition Through Computer Assisted Instruction" (1979), and in a summary article under the same title written with George Culp. Burns' intention was to place the heuristic strategies of Aristotle, Kenneth Burke (A Grammar of Motives), and the tagmemic matrix of Young, Becker, and Pike (Rhetoric, Discovery and Change) on computers for the purpose of studying their relative effectiveness in stimulating thought in student writers. The actual comparison of the three heuristic strategies proved less significant than the method Burns developed for programming them.
TOPOI, which has become the most widely distributed of his three invention programs (the other two being BURKE and TAGI), provides a typical example of the method. In this program, Aristotle's twenty-eight enthymeme topics for generating a persuasive thesis (see his Rhetoric) are reformulated into thirty-seven questions or thought prompts which challenge the student's understanding of his intended thesis in terms of definition, causes and effects, opposites and associations, consequences, and matters of fact and opinion. If the student's initial topic was "computers in education," for example, the computer would prompt, "What is the definition of 'computers in education'?" Or, "Who supports 'computers in education'?" The student's responses produce one of two results: a gradually expanding base of explicitly expressed relationships and perceptions, or a gathering realization that he does not have the information or insight to proceed with his thesis as formulated. The former result provides a transcribed self-interrogation from which a tightly developed thesis may be drawn. The latter informs the student early on that he needs to redirect his attention or research.
Admittedly, the operating concept of the software does not seem very pro found. Much the same thing occurs every time a student shows up during of office hours, and the instructor patiently runs down a well-internalized list of prompts, usually beginning with, "Well, Mark, why did you pick . . . ?" But it is the simplicity of the operating concept that makes for its great strength. The beginning of wisdom is the questioning of that which we have already accepted in some form or fashion. The questions that Burns' software puts to students challenge every aspect of their own understanding of their specifically stated subject. But, because his programs challenge relationships only and not content, his heuristic method empowers the student in ways which would be impossible for even those instructors who are acutely aware of the authority relationship between teacher and pupil. Burns' method, simple as it seems at first, is in effect a profound means of stimulating student self questioning without implicitly or explicitly devalorizing the student's experiences or ideas.
In order to test this software design and possibly improve it, Professor Jerome Bump, in the Fall of 1985, established Project Invention Heuristics at the English Department Microcomputer Research Lab at the University of Texas, and appointed me project director. We specifically targeted upper division and honors composition students for two reasons: we felt the abstract nature of some of the prompts would prove less threatening to them, and we hoped to demonstrate that microcomputer writing instruction is capable of imaginatively serving high-ability students.
Burns had obviously felt, as many CAI programmers did and still do feel, that an important element in the success of instructional software is the illusion of two-way conversation, the replication of human to human activity, what I have called the "User-Friendly Fallacy." Therefore, in its directions, the program suggests that the student not think of the computer as a ma chine, but ask questions normally. This required that a fairly complicated series of program steps be included to cull key words out of the student's responses in order to direct proper and corresponding reactions, to produce the illusion that the computer was actually understanding what was being said to it. This series of steps produced on the microcomputer a delay in execution following each student input, a delay which increased in proportion to the length of the input. Faced with this clearly debilitating machine procedure, I began questioning the entire notion that sophisticated software must pretend a human competence. I began testing a variety of formats using the Burns de sign. The students who used the programs, all relatively computer-literate, collaborated with me in my programming effort.
Programming in BASIC is itself a winding, involving, endearing, and maddening process. Working with someone else's code over a period of a year (each attempt criticized almost immediately) was an intense experience which seriously challenged some of my treasured assumptions about how students react to computers, and about how computers should react to students. I found myself being subjected to a heuristic procedure as thoroughly demanding as the programs themselves. In the process, I became convinced of two things: composition software should be developed in accordance with the pre dispositions and presuppositions of the user, not those of the programmer; and the ultimate success and acceptance of computer-assisted instruction in composition lies with the type of software that Burns pioneered.
Types of Computer Interaction
In essence, there are three ways a computer can interact with its operator: it can expect no response, a closed response, or an open response. No-response software simply displays text or simulation, asking for input only to maneuver through the text. Closed-response software asks a number of questions of the operator, and then compares the operator's response to a response previously anticipated and programmed, the so-called "correct" response. The computer reacts on the basis of whether a match is made or not. If the student's input exactly matches the programmed answer, the computer puts out positive reinforcement and moves on to a harder question. Otherwise, the machine shows disapproval in some fashion and branches to a tutorial or to easier questions. With objective answers, this is a fairly simple matter. But if it is to produce the illusion of conversation, the program must anticipate a wide variety of responses and perform a considerable number of comparisons. Keywords and key terms ("what?"; "explain"; "why?"; "I don't understand") are searched for in order to cue some version of intelligent machine response. If the student, for instance, has entered the word "why" in a certain syntactical environment, the program interprets this word as a request for an explanation of some sort, and branches to an explanation routine. Obviously, keyword search mechanisms are fine showcases for the cleverness of the programmer, but require the operator to respond as anticipated if the program is to work.
For this reason, closed-response programming empowers the programmer and, by extension, the computer. Even in so-called "dialogue" programs in which an effort is made to simulate conversation, students often feel the pressure to provide the "right" comments, and occasionally blame themselves if the computer throws out nonsense. Even unsophisticated operators learn quickly that the computer knows the responses that the operator had better produce or suffer the consequences of an admonishment or, worse, a crashed program (which almost inevitably causes the student to say, "I think I did something wrong," when in fact the student may have simply responded more imaginatively than the programmer anticipated). No matter how cleverly positive and negative reinforcement are used, the fact remains that no student will leave a closed-response program knowing more than the programmer wanted him to know.
In this same way, text analysis programs along the lines of WRITER'S WORKBENCH and HOMER (both described in Wresch's The Computer in Composition Instruction, chapters 4 and 5 respectively) represent closed-response systems. The "facts" of grammar and usage which the programs identify are dependent upon the "facts" which have been loaded into them. The programs seem smart only because they employ a privileged set of facts which the computer is attempting to load into the student in much the way the programmer loads them into the program. Rules of usage are applied with a rigidity tar surpassing the worst "Mrs. Grundy" who ever assailed a split infinitive, and therefore the program is absolutely incapable of perceiving the essential link between subtle meaning and idiosyncratic usage. Closed-response software inherently targets a student clientele deficient in the "basics," and few understand this more quickly than the student operators themselves.
On the other hand, open-response programming, which constitutes the unique and valuable character of Burns' software, makes no attempt to "understand" the student's responses or evaluate them. The programming effort is directed instead toward provoking the student's awareness of relationships, of knowledge structures, of her own purposes and procedures. Open-response and open-ended programming place no limits on the size or form of the response, and therefore do not threaten the operator with constant censure. The computer guides the student's self-examination and provides a printout of the journey for the student himself to read, to reexamine, and to criticize.
Computerized guided heuristics have, we found, another advantage over closed-response programming. Students who use style analyzers to polish off their product develop a dependency upon them and feel uncomfortable handing in a paper without first running it through a spell and style check. Invention programs don't seem to create this dependency. Although the thorough "self-challenging" process of the sequenced thought prompts is at first quite new and often elicits a sort of "where has this program been all my life" response, by the third or fourth time a student has undergone a particular sequence of prompts (Aristotle's topics in TOPOI, for instance), the entire self questioning process has to some degree been internalized. The student has not memorized the prompts, of course, but the habit of examining a thesis through a wide variety of challenges, of looking for those points of view which don't manifest themselves immediately, becomes an implicit part of the discovery process. The programs still remain useful to the students, and as open response programming gains in sophistication and distribution, the variety of specific applications will continue to attract students who may have already had a great deal of experience with other heuristic programs. But the process of self-interrogation is "learned" rather quickly, indicating that open-response programming is not limited to simply being a writing tool, as are style analysis and word processing, but is also clearly instructional.
The simplicity of the template method is deceptive. And Burns, I think, did not recognize the strength in the operating theory of his original programs enough to completely abandon the User-Friendly Fallacy and explore the nature of the questioning (or guiding) process. For instance, his questions came up in random sequence, the ostensible purpose being to suggest spontaneity if a person used the program more than once. As we discovered almost immediately, however, and as Burns has said in retrospect (Wresch 31), the power of the software lies in its "directing" or "channeling" function, and this requires a specific sequence of prompts, ordered and phrased in such a way as to lead the student through ever deeper levels of his own reasoning. For in stance, one of Burns' questions in TOPOI is, with "x" being the topic, "Who supports x?" The student then types one word or a hundred, and proceeds to a completely different question. In one of our versions of the program, the student's response is displayed and a further series of questions involving political or personal "support" for x is asked, each allowing for a further open response. As a particular line of reasoning is pursued, the student's total response for that line of reasoning is displayed for revision or as reference to help in answering further thought prompts.
At any time the student can skip to the next prompt by entering a "c," for "continue," or jump out of the "support" line of reasoning altogether by entering a "j" for " jump." She soon discovers that not only is she thoroughly investigating relationships which seem relevant to her purpose, but the act of rejecting irrelevant lines of reasoning drives home the shape of that purpose and clearly maps what she is and what she is not to include in her paper. This gathering insight may encourage her to revise her thesis, and she can do so by means of the "r" or "revision" command. Students have reported, in an hour on the computer, up to nine revisions of their thesis statement.
Eight of the fourteen students who initially participated in the project considered the invention programs the most useful in the lab, in spite of the fact that all of the students made use of competent text analysis and word processing programs. Their written comments on CAI INVENTION are interesting: it was "most helpful in uncovering new meanings and thoughts," "helped me relate certain aspects to unrelated areas," "allowed me to expand my thoughts about the subject," and "it gave me new ways of thinking of my problem." These remarks are typical of the strong sense of enlightenment experienced by students operating the program for the first time. Such reactions suggest both the power of systematic heuristics and the general absence of successful emphasis upon prewriting strategies by writing instructors.
The Error of Expectation
Those who criticize open-response software do so in terms of its limitations as measured against ultimate hopes for computerized instruction. Andrea Herrmann's criticism in her article "Using the Computer as Writing Teacher: The Heart of the Great Debates" is typical. She points out that invention programs do not "evaluate the number and sophistication of ideas actually used by the students." Even those who have written open-response programs and generally support the design admit this fault, most notably, Burns himself. In the first chapter of Wresch's The Computer in Composition Instruction, Burns mentions that some students are disappointed that the computer cannot help them rate the content of their responses.
Most people expect that the highest function of a computer is to replicate human behavior, to respond and make judgments with the skill of a human being. This ability, or the illusion of it, has been considered paramount by CAI programmers who want to produce sophisticated and imaginative soft ware. Hence Burns believes that without a willing suspension of disbelief on the part of the student, "the programs fail" (Wresch 17). He describes the program mechanism which searches the student's response for keywords in order to cue a proper machine response as "the primitive soul of the program" (Wresch 27). Roger Schank, however, believes that keyword systems which provide the illusion of cognitive interaction merely demonstrate "people's gullibility" (136).
The real soul of Burns' programs and of subsequent programs developed by Schwartz, Arms, Selfe, Wresch, and the Rodrigues's, is actually the thought prompt, the question which provokes, as Burns calls it, the "tensions" which lead to greater and greater self-discovery (Wresch 27). The fact that the computer cannot evaluate student responses is in fact a strong argument for, not against, the software. As Helen Schwartz remarked in a College English article concerning her open-response program SEEN, "Students soon realize they can not get answers from the computer. They soon revel in the fact that they are doing the thinking not the machine. The computer doesn't really know-or care-what the user says. This can be liberating" (241).
This coincides exactly with my experience. The young, computer-literate student (and this population is rapidly expanding) is neither awed by the illusion of the human machine nor tolerant of the glitches which inevitably puncture that illusion. Accordingly, the programming effort should concentrate on the nature of the thought prompts, the sequence and structuring of the prompts, and the response review mechanism. The writing instructor must direct the programmer, and instructional effectiveness must direct the technology. Not vice versa.
But people continue to expect that sophisticated instructional programs should, in some way, be more human, more like computerized teaching assistants complete with lively patter and foolproof electronic grading. As long as instructional software is judged in terms of the power of the technology- that is, on the basis of how human it makes the machine or how clever the program mechanism is-open-response software will continue to seem modest and unexceptional, especially in light of the powerful technical achievements of text analyzers and word processing. CAI in writing instruction will be restricted to supporting the packaging of essays instead of contributing to their creativity and originality.
Works Cited
Aristotle. The 'Art' of Rhetoric. Trans. J. H. Freese. Ed. G. P. Gould. Cambridge: Harvard UP, 1982.
Arms, Valerie. "Creating and Recreating." College Composition and Communication 34(1983): 355-58.
Burke, Kenneth. A Grammar of Motives . Berkeley: U of California P, 1969.
Burns, Hugh. "Recollections of First-Generation Computer-Assisted Prewriting." Wresch, The Computer in Composition Instruction: A Writer's Tool. 15-33.
-----. "Stimulating Invention in English Composition through Computer-Assisted Instruction." Diss. U of Texas, 1979.
-----, and George H. Culp. "Stimulating Invention in English Composition through Computer-Assisted Instruction." Educational Technology 20.8(1980): 5-10.
Herrmann, Andrea W. "Using the Computer as Writing Teacher: The Heart of the Great De
bates." Proceedings of the Annual Summer Conference on "The Computer: Extension of the Human Mind 11" Eugene, OR: July 20-22, 1983. ED 260 406.
Hertz, Robert M. "Problems of Computer-Assisted Instruction in Composition." The Computing Teacher Sept. 1986: 62-G4.
Rodrigues, Dawn, and Raymond Rodrigues. "Computer-Based Problem Solving." Wresch, The Computer in Composition Instruction: A Writer's Tool. 34-46.
Schank, Roger, and Peter Childers. The Cognitive Computer. Reading, MA: Addison-Wesley, 1984.
Selfe, Cynthia, and Billie J. Wahlstrom. "The Benevolent Beast: Computer-Assisted Instruction for Teaching Writing." The Writing Instructor . 2(1983): 183-92.
Schwartz, Helen. "Teaching Writing with Computer Aids." Coliege English 46 (1984): 239-47.
Wresch, William, Ed. The Computer in Composition Instruction: A Writer's Tool. Urbana: NCTE,
1984.
Young, R. E., A. L. Becker, and K. L. Pike. Rhetoric: Discovery, and Change. New York: Harcourt, Brace, and World, 1970.