Interaction and Group-making in On-line Learning Communities

Marshal B Anderson, M.Ed. Workshop 2 Co-operative Project, Spring 1999

Results

Sociometric Data

Sociometric diagrams were generated from the data giving the following results. The charts are of the positive (score 2 and 3) choices made by group members.

Sociogram of Real World Group - Positive Choices

The pattern produced by the real world group's choices shows that a very complex system of working preferences has developed. There are several direct pairings (20-16, 5-8, 25-7, 10-24, 2-18, 3-4, 8-9, 15-6) which would suggest these students would feel very comfortable working together. The pattern identifies two distinct 'stars' (5 and 15) and three more possible stars, (8, 21 and 20) who, with the exception of 8, seem to be creating central points for the formation of three possible groups identified in three colours on this diagram. It should be noted that the above diagram is based on raw (un-weighted) choices; Appendix 2 contains information about weighting which shows that the identification of 'stars' is actually more problematic. Having said that, there are no 'cleavages' (clean breaks between groups) in the diagram which might suggest to us that the group mix well with each other. There are in fact only four 'isolates' (students chosen by no other student) in the group (17, 13, 14 and 22) and we might view this group as particularly well mixed.

Sociogram of On-Line Group - Positive Choices

The sociogram of the on-line group shows quite a different picture. Only 12 of the group members are actually shown in this diagram because the rest were isolates. Looking at this diagram we can identify two possible groupings amongst the members displayed (identified using colour) but these would seem to be of little value as they can not take account of the remaining 18 members. The additional arrows represent the number of positive choices relating to each displayed member. This shows a number of 'stars' emerging (28, 30, 7, 21, 11) round which groups could undoubtedly form, but there is nothing like the spread of interest we saw in the real world group.

Why might this be? One possible explanation might be that the real world group are much more likely to know each other in a more general sense and to have set up social as well as working relationships. We might certainly expect them to have a both deeper and wider knowledge of each other and to use many more factors in constructing their relationships that the on-line group is able to.

 

Conferencing System Data

Arrangements were made to access the conference server to examine any relationship between grouping choices and activity in the conference. Two figures were available for each group member, total number of contributions and total number of words. These data were taken early on in the study, shortly after group-making decisions were taken, but they were taken from the main conference, not from the sub conferences generated by the sub groups – i.e. they are an expression of contributions up until the point of the group deciding and splitting.

These figures generated a third figure which was the average length of each contribution. The reasoning behind this was to give some indication as to the basic nature of contributions before attempting to draw any conclusions from them. First we look at the basic distribution of each set of figures to see a normal distribution.

This first graph shows the number of contributions to the conference and seems to demonstrate a normal distribution - i.e. some students contributed many messages, some just a few, but most were in the central range.

In this second graph we look at the number of words contributed in total. Here the distribution is still largely normal, though it is skewed towards the lower end of the scale. The graph actually hides one maverick value that should be noted; student 26 submitted a total 154112 words - 50% above the next value down. This is a huge input over the time period. On investigating this it was discovered that student 26 was in the habit of contributing large amounts of text in the form of articles they thought were of interest to the group, therefore this entry is invalid and that should be noted in the following discussion.

This third graph shows us the average (arithmetic mean) length of postings. While this is a very blunt instrument it gives us some idea of the verbosity of contributions, though none whatsoever of their value (the content was not available). Again, we see a skewed but substantially normal distribution of activity. Thus with these three graphs we have demonstrated that the data collected follows basic statistical norms and we might draw some, albeit tentative, conclusions from the processed data.

The point of interest here is whether there is any connection between the number and length of postings and the sub-groups that are formed. That is, when choosing whom to collaborate with, are on-line group members drawn toward members who contribute a similar amount, an opposite amount, or is there no significant link. The basis of this notion comes from Sprott's (1958) ideas on prestige. Within this he says:

"It is not surprising therefore to find that in many cases sociometric choice is found to go with other accepted standards of excellence." (page 150)

What we are dealing with here is a point of perception. The fact that such text-based communication lacks may of the face to face cues we are used to is well covered in the literature, for instance:

"Electronic messages lack such social information as senders' and receivers' job titles, social importance, hierarchical position, race, age and appearance. The context is undifferentiated too, since formal and casual exchanges look about the same. People (might) have information about senders, receivers, and situations from other sources, but there are few cues in the computer interaction itself to remind people of that knowledge." (Kiesler S , 1991, Page 154)

and

"When technological change creates new social situations, traditional expectations and norms loose their power. People invent new ways of behaving." (Sproull and Kiesler 1991, page 39)

There seems to be some identifiable co-relation between the choosers choice and the contribution rate of the person they choose. There are many possible explanations for this and the data gathered is not able to inform us further in this study. One hypothesis might be that one of the few non-verbal 'cues' we have in asynchronous conferencing will be the overall quantity of postings, and that might spark the same sort of subconscious judgement that a person's clothing or accent would in a face to face group. If it does, then we might expect to find at least some correlation between a choosers rate of contribution and the person they choose. The identification of the cues available to one-line conference members, and the weight they are given, might enable us to re-design the conference environment in ways that encourage more 'telepresence'

On analysing the figures it became clear that there were few statistically significant connections to be found in the current data. However, two patterns did emerge that are worthy of discussion.

This graph plots the average length of a student's own postings against the average length of those they chose to work with as a +3 option. There might be a pattern emerging here that suggests some connection between size of postings and choice. The blue lines (drawn simply by sight) show that the co-relations, while very loose, might fall into a given area. Further statistical analysis of this data, and further sampling of similar groups would help confirm or refute this tendency.

Here we are looking at the data which describes the number of anti-choices made against students by other students compared with the number of postings made by the 'un-chosen' student. Here there seems to be a link in as much as those students that made more posting are overtly rejected by some of the other students. The blue lines here (drawn simply by sight; omitting three possible maverick plots) suggest a tendency that would warrant further study. The high number of zero values reflects the fact that, as in the positive choices, negative choices tended to concentrate on a small number of group members.

Taken together, do these graphs demonstrate that the simple quantity of postings and words has some effect on the choices made? Certainly it seems that several students really don't want to work with others in their group who make long/many contributions to the discussion, though there is no evidence that the opposite is true - that students would rather work with those who make less contributions.

One possible explanation for this might be that those who make large contributions are perceived as being either more demanding to work with, or perhaps tend to dominate discussion. Grint (1989) states;

"Some of the most frequently asserted advantages of CMC are the flattening of hierarchies, the consequential expansion of participation, and the channelling of attention away from the messenger to the message. The interviewees [12 Open University students] had some sympathy with this point, but generally seemed to have adopted systems for re-constructing status." (page 191)

It might be possible that what we are seeing here is exactly that re-construction of status/prestige; that although we might not have all the non-verbal cues to help (or hinder) our assessment of those we work with in an on-line environment, we still need them and therefore construct others. The sheer weight, or lack of it, in the contributions of others could well form such a cue. What can be said with some certainty is that there is enough evidence in this small study to suggest that research on a wider scale could prove worthwhile.