Grounding in computer-supported collaborative problem solving
|
|  Yüklə 1.41 Mb.
|
Symmetry in acknowledgmentWe compared the behavior of the 9 pairs with the highest acknowledgment rate versus the 9 pairs with lowest rates. It appeared two groups differ significantly with respect to the heterogeneity of talk between the pair members: when we count the number of MOO interactions ('say' + 'page') per minute, the average difference between Hercule and Sherlock is 0.33 for pairs with low acknowledgment rate versus 0.06 for high rate pairs (F=6.87; df=1; p=.05). This might be explained by the fact that, when one detective is interacting more frequently than the other, the slower cannot acknowledge all messages. The reverse explanation could also be true: heterogeneity in individual talk rates may be due to lack of acknowledgment by one partner. To verify the latter hypothesis, we counted the degree of symmetry in acknowledgment, i.e. whether Sherlock acknowledges Hercule as often Hercule acknowledges Sherlock. In general, acknowledgment is very symmetrical: in average, one pair member acknowledges 8% more often his partner than vice-versa. This low level of asymmetry seems compatible with a general concern in collaboration regarding the equilibrium of activity. The average asymmetry is equal to 8% both in the group with a low acknowledgment rate and in the group with a high acknowledgment rate. This leads us to think that symmetry in acknowledgment is more determined by a social contract than by individual features. This contract is rarely explicitly negotiated, except when it is broken. It was for instance the case in pair 7 (see example 10). It was also the case when one partner suddenly did not acknowledged a few utterances, because he was busy elsewhere (see example 5) or because messages were lost (e.g. using 'say' while the partner is out).
The MOO technology gives more flexibility regarding delayed acknowledgment than voice conversation: in voice dialogue, if the delay is too long, one may forget what is being acknowledged, while, in the MOO, one can always answer later on, the partner has just to look a few lines up to identify what is being acknowledged. We observe that the delay varies according to the modality of acknowledgment. In MOO verbal interactions ('say' + 'page') the average delay for the group is 48 seconds. It tends to be shorter (mean for the whole group: 35 seconds) when action is involved: acknowledging talk by action, action by talk or action by action. Utterances such as "Let's go to room 4" or "Ask Marie about the insurance" have either to be acknowledged almost immediately or to be ignored. Conversely, acknowledging a movement by an utterance such as "Hi, welcome here" has also to be immediate to make sense. On the contrary, the acknowledgment through the whiteboard seems to be slower. The average delay for acknowledgement whiteboard items by talk or acknowledging talk by whiteboard actions is 70 seconds. This longer delay can be explained by the persistence of information on whiteboard: there is no urgency to acknowledge information which will (probably) remain a long time on the whiteboard. If we compare three situations (voice -- MOO talk - whiteboard), the delay seems to increase with the persistence of information in the concerned medium. We do however not have enough data (except for the talk/talk acknowledgment) to draw statistical inferences. Let us now focus on delay talk/talk acknowledgment. We compared the 9 pairs with a short acknowledgment delay (mean = 31 seconds), hereafter referred to as 'fast acknowledgers' versus the others (mean = 64 seconds). Both groups spend roughly the same time on the task, respectively 128 and 118 minutes. Within this time, the fast acknowledgers not surprisingly acknowledge more messages: 173 messages for the fast acknowledgers versus 108 or the slower group (F = 8.36; df = 1; p = .01). However, the acknowledgment rate is identical in both groups (mean = 0.41). This contradiction can be explained by the fact that fast acknowledgers interact more frequently: 1.36 message ('say' +' page' per minute for fast acknowledgers versus 0.93 for the slow group (F = 9.10; df = 1; p= .01). Actually, the number of interactions does include the number of acknowledgment, hence the relationship between these two variables is not surprising. However, if we take off the number of acknowledgment from the global number of interactions, we still find out that fast acknowledgers communicate more frequently (mean = 0.8 message per minute for fast acknowledgers, mean = 0.54 for the other group, F= 5.6, df=1; p = .05). In other words, fast acknowledgers seem to simply fast talkers. In summary, we can say that the delay of acknowledgment is longer for more persistent media and that it does not seem to indicate a greater mutual attention, or some similar indicator of sharedness, but more simply some trend to interact more frequently (shorter messages, better typing skills, ...). The length of the delay is one aspect of the conversational rules that are implicitely established between participants. Despite the fact that users can type in the MOO or use the whiteboard independently from each other, they can only pay attention to one of the windows. Even in the MOO window, its is difficult to pay simultaneously attention to the pane where new messages are displayed and to text entry pane. The "average delay" rule determines when a subject stops waiting for an answer, how often he returns to his screen to see whether his partner said something or how many lines he must scroll up to find one's partner answer22. When this social rule is broken, a partner may become impatient as in example 10. Typing several times the same message23 creates a sudden scrolling in the partner's MOO window, potentially gaining his attention. Actually, in everyday use of the MOO, the characters present on the MOO but who cannot answer rapidly indicate it by setting their 'mood' (a small string which appear besides their name) to 'busy' or 'afk' ("away from keyboard").
Example 10: Long acknowledgment delay may cause impatience (from Pair 7, translated) | |||||||||||||||||||||||||||||||||||||