TOGNOTTI S. & MENDELSOHN P.
(TECFA - Faculty of psychology and educational sciences - University of Geneva)
Revising text is a rather complex task that acts on each level of the writing process. It implies evaluative reading, diagnosis, and modification at least (Hayes & Al., 1986). Studies have shown that novice writers encounter several kinds of difficulties, one of which is to take their audience into account. Other problems are linked with error detection and text improvement strategies (Bereiter & Scardamalia, 1987; Hayes & Al.1987). Cooperation in revision can help error detection as author receive peer feedback and, interactions can stimulate qualitative criteria creation as well as problem solving procedure (Paoletti, 1989; Zammuner, 1995). Furthermore, a peer work can also enhance regulation and self-regulation skills (Dillenbourg, 1991).
We have conducted an experiment in which our main purpose was to enhance the relationship between authors and readers to help novice technical writers to adopt the audience point of view. We used negotiation and text classification to help them on error detection. We thus have created a situation that supports the impact of evaluative reading of texts written by other authors on the writing of one's own text. This situation implies one being audience for others and others being audience for one's text. To support this situation, we have built a computer supported cooperative writing tool taking charge of five subjects simultaneously. This tool has been implemented on a textual virtual reality server (MOO) and has been made accessible through WWW as a shared interface. This writing environment is composed of two parts: a MOO window in which subjects test verbs and objects as well as interact amongst themselves and an HTML form area in which they write their texts, read and classify them.
Four groups of five students each produced short technical texts that described MOO verb/object functionality and syntax. Audience was said to be novice MOO users. Two conditions were used in this experiment. For both conditions, subjects had to:
These operations have been repeated three times to observe text evolution and impact of evaluative reading on following descriptions. Furthermore, in the second condition, subjects had to negotiate the quality of each written text before ranking it.
The general trend is improvement from one text to the other. Our results show that negotiation makes subjects integrate more quality criteria in their following texts. Without negotiation, texts tend to keep their general characteristics, e.g. syntax presentation. Negotiation and, as a matter of fact, texts quality verbalisation, is thus a necessary condition to observe interesting improvement in text quality for each subject. Results also show that relevant technical text quality criteria quickly appear within the negotiation. As a consequence, strategies for text improvement are easily applied as other texts are considered as examples in which one had identified good or bad elements.