A MOO is a text-based shared workspace: agents can access to a set of shared objects, but these objects and agents are viewed through verbal descriptions and handled through typed commands. Traditionally the word 'workspace' describes groupware in which shared objects are graphically represented and manipulated through direct manipulation. The shared workspace can be a set of pre-defined objects with constrained manipulations, such as in chess playing. Such workspaces are also used in human-computer collaborative systems such as HKE (Terveen, Wroblewski & Tighe, 1991) or MEMOLAB (Dillenbourg et al, 1994). In human-human computer-mediated work, the share space can also be an empty sheet (called whiteboard) where the user can draw with some graphic tools. This study is concerned with whiteboards.
-
Co-presence (can see the same things). When agent-A writes an information on the whiteboard, he does not know whether agent-B is looking at this part of the window. Agent-B may have scrolled the window. Some shared virtual environment such as Shared-ARK (Smith et al., 1989) provide a 'radar view' through which the user view which sub-areas of the virtual spaces are respectively viewed by himself and by her partner. In our experiments, we forced co-presence by giving to the whiteboard a fixed size, with almost no scrolling available.
-
Visibility and audibility (can see each other). These features are intrinsic to whiteboards: each partner can see what the other draws or write. In addition, some whiteboards show the partner's cursor. This feature is important in grounding. First, in enable the partners to track the other attention (Whittaker et al, 1993). Second, viewing each other cursor enable deictic gestures. The whiteboard we used did not display the partner's cursor. Subjects sometimes developed mechanisms to compensate this mechanisms, such as making a small mark or moving slightly the object being discussed. The degree of visibility and audibility changes from a workspace to another one. The whiteboards we used in experiments are fully shared: all agents have equal access to all objects. This is not the case for all whiteboards. Some whiteboards use private objects (one agent cannot change an object drawn by another one) or authorization procedures (one agent has to explicitly authorize modification of his objects or ask for the authorization of changing other's objects).
-
Cotemporality (messages received at the same time as sent). Whiteboards are synchronous. Sometimes a small delay may however occur, due to the client necessary for the client to refresh the screen display.
-
Simultaneity (can both parties send messages at the same time or do they have to take turns). Simultaneity may be reduced by the time necessary to edit objects. With direct manipulation whiteboards, most objects can be created and modified within one or two seconds. Simultaneity problems may occur when editing text boxes. Jermann's (1996) MOO-based workspace includes locking mechanisms through which both agents can edit the same object, but never simultaneously.
-
Sequentiality (can the turns get out of sequence). There is no sequentiality problem because objects are generally not answering to each other, there are no turn taking rules. However, users sometimes establish space occupation rules,
-
Reviewability (can they review messages, after they have been first received) is a key feature in share workspaces: objects are persistent (Smith, 1994), once created they stay as long as they are not deliberately erased. Whittaker, Geelhoed & Robinson (1993) emphasize that whiteboard help to 'retain the context': the whiteboard is a summary of what has been done (and hence of what remains to be done).
-
Reviseability (can the producer edit the message privately before sending). Graphical objects are generally not visible as long as the partner creates it (e.g. dragging the corner of a box). In the second whiteboard we use, text boxes where created through a small dialog box and where hence 'private' until the user clicked 'ok'. One has to discriminate here the whiteboards using vectorial objects versus those using bitmap. Vectorial objects can re-sized, re-colored, grouped, moved, etc., while bitmaps can generally only be erased. The repair costs are hence higher with bitmaps. However, these offer the advantage to include free drawing. The two whiteboards used in our experiments use the vectorial mode.
The costs of grounding vary between different whiteboard systems. For instance, the cost of formulation is much lower in Whittaker & al (1993) experiments, where subjects draw objects with a stylus on a tablet, than in our experiments, where the subjects used a graphic tools, with standard functions (boxes, circles, lines and arrows, text), but which was not very easy to use (the tool selection procedure was too rudimentary). Conversely, the change costs is lower in our experiments than in Whittaker's et al (where subjects have to erase objects), but it was still higher than it is in other whiteboards (text boxes could not be edited, objects could not be re-colored, ...).
Finally, let us mention that the presence of a whiteboard does not by itself solve all coordination problems. Whiteboards suffer from different problems inherent to direct manipulation (e.g. lack of abstract operations). Moreover, the whiteboard requires an additional effort to update and maintain the shared representation. If a participant fails to update the shared representation, coordination accidents may occur such as those reported by Rodgers (1993). She emphasized that sometimes "individuals that work together are required to coordinate their work even more in order to use the new collaborative systems" (p.295).
Grounding across different modalities
The VCE used in our experiments include a MOO environment and a whiteboard system. This configuration is derived from our research question: how schemata contribute to grounding in written dialogues. In the sections above, we have discussed grounding respectively in the MOO and in the whiteboard without taking it account the fact that both software are used simultaneously. However, the VCE makes a whole: grounding processes will be influenced by the complementarities and incompatibilities between the MOO and the whiteboard. The whiteboard role with respect to communication turn around three mechanisms: deictic gestures, retain context and enable graphical representations.
-
Deictic gestures play a role important in grounding. For instance, Frohlich (1993) emphasized the complementarity between conversational interfaces and direct manipulations interfaces: the latter reduce the 'referential distance' inherent to language interaction, by pointing to objects referred to in verbal utterances. Actually, complementarity exists in Whittaker et al (1996) system because it combines a computerized whiteboard with a free-hands audio system. When agent-A says "look there" while clicking to some point in the workspace, agent-B can associate "there" with the referred location because (1) agent-A produced the utterance and the gesture are almost simultaneous and (2) agent-B can simultaneously hear this utterance and look at the whiteboard. At the opposite, in our VCE the MOO and the whiteboard compete for both input and output resources: (1) both components mobilize agent-A's hands, hence input is exclusively done in the MOO or in the whiteboard and (2) both components mobilize agent-B's visual attention, when her focus is on the whiteboard, she may fail to notice what's occurring in the MOO. The frequency of deictic gestures should hence be lower in a 'MOO + Whiteboard' than in a 'Audio+ Whiteboard' setting.
-
Whittaker et al (1996) emphasize another form of whiteboard - communication complementarity. Since whiteboard information is persistent, it retains conversational context. This is especially true in their experiments with audio communication, since voice is non-persistent. In our configuration, this complementarity exists as well, but to a lower extent, since the MOO communication is semi-persistent (previous utterances remain displayed, but they scroll slowly upwards until they disappear from the screen). Hence, in a 'MOO+Whiteboard' environment, we may expect to have a more symmetrical relationship, where both the whiteboard provides the common grounds for conversation and conversation help to make sense of whiteboard elements.
-
Finally, the whiteboard may contribute to represent spatial relationship which are not easy to express in audio communication. Whittaker et al (1993) observe that the whiteboard is most useful for tasks which are inherently graphical, like placing different pieces of furniture of a given "floorplan".
|
Yüklə 1.41 Mb.