research with future studies to suggest some directions which could go in the future.

Chapter 5 Conclusions

According to the research results and problems we faced, this chapter synthesizes

all the studies to propose several conclusions. The first section describes the

conclusions and suggestions to the audiences. The second section concludes the

research with future studies to suggest some directions which could go in the future.

5.1 Conclusions and Suggestions

On one hand, the research explicitly aims to provide a web-based tool, Forum

Visualizer, for users to visualize an online discussion forum using social network

analysis, and tries to discover and explain users’ general and particular sociograms of

interaction and behavior about their discussing atmosphere. On the other hand, the

research also implicitly provide intuitive observation and understanding of the online

discussion forum for users at a variety of scales, helping them explore themselves,

locate others with similar behavior and preference, and search for appropriate topics to

join in. According to the research results, we found that visualizing an online

discussion forum is practicable. First, the large dataset of discussing relations can be

explored and reviewed easily, rapidly, and clearly. Users can simply watch their

relations to others and their growth in the forum. An administrator can observe all

users’ relations to discover the hubs, isolates, and overall discussing conditions to

decide when and to what extent he should intervene, mediate, or coordinate to balance

or control the entire discussing moods.

Second, the design of an online discussion forum should be considered

importantly before running it. As far as the database is concerned, all kinds of

information should be stored to produce thorough sociograms and analyses in detail. In

this case, the absence of the real writer to whom a user wants to reply in a thread leads

to the apparently larger number of self-to-self relations than other relations. If the

relations among users in a thread could be recorded in an extra column of the database

table, the sociograms will be more elaborate and significant focusing on a series of

replies between two users.

Last, a social network focusing on discussing relations, i.e. a socio-centric

network in an online discussion forum, is a scale-free network because users have

interests and preferences. But under certain conditions such as limited discussing

topics and time, it could become a random network. Under a hypothesis that users

would post articles freely, it is an interesting phenomenon that the discussing social

network is still variable, flexible, and controllable. Knowing the characteristics and

expanding it to the online communities, more strategies could be applied to manage the

communities and to study users’ behavior and interactive relations for further

knowledge sharing research.

We hope that our work will stimulate further studies of visualizing online

discussion forums, make contributions to social network analysis and knowledge

sharing in online communities, and be applied to general online discussion forums.

Although the approach has not received much attention, we suggest that it will

probably turn out to be widespread in online discussion forums, often with important

dynamic consequences.

5.2 Future Studies

In addition to several conclusions, there are indeed many other study issues

related to this research that could be done in the future. Many directions can be further

investigated such as (with no priority):

First, the famous shortest path algorithms, e.g. Dijkstra and Floyd-Warshall, could

be used to calculate the centrality of closeness and betweenness in the sociograms,

helping analyze the relations and dependency in more detail. The more information we

collect, the more understanding to the sociograms we will have.

Second, the sub-graph concepts, e.g. components, cores, cliques, and clusters,

could be applied to help classify the users in similar interests or preferences,

suggesting recommendable persons to interact with in recommendable topics. This

could be a useful function for a user to know who are actually on his side with

common opinions. Several types of users could be categorized to build a stereotype in

the online discussion forum.

Third, Forum Visualizer could be tested by users to collect feedbacks to see what

their reactions will be. All views of sociograms will be opened for users to take a look.

By this strategy, the views they watched could be recorded to analyze if they will

watch others’ sociograms besides ego’s ones. This action could help build users’

graphs of growth in different periods in the online discussion forum. According these

graphs, we can confirm if the discussing atmosphere is stimulated to make the number

of discussing articles rise quickly with the tool by comparing to other forums without it

applied. Administrators could also identify levels of users’ participation in the online

discussion forum by the growth graphs. Monitoring users’ actions in this way will be

effective for managers to take controls in time, not until users has left mostly.

Fourth, the small-world network theory could be validated if the social networks

in the online discussion forum match the characteristic. An interesting issue is how

many degrees of separation a user will have on average in the whole network. Will it

still be six degrees? Plus the scale-free network theory, will the hubs increase the

connections to others easily and the isolates remain isolated? This derives two opposite

strategies. One is to control the hubs and leave the isolates alone in order to promote

the discussing atmosphere. The other is to focus more on the isolates, attracting their

attention to participate in the conversations more positively. Both strategies could be

adopted to be as contrast to inspect their advantages and disadvantages. This could be a

reference to understand how to well manage an online discussion forum or even an

online community.

Fifth, what a user really wants is also an attractive issue to do more research. We

can start by the self-to-self relations that could be represented in other ways such as

loops or avatars, specifying users’ self histories and changes. This information will be

valuable to help analyze a person’s inner world and discover advantageous thinking

paths about his behavior or action. The graph layout will be a great object to start with

too. Different kinds of graph drawing layout algorithms could be implemented to

delineate the same sociograms. These graph layouts will be chosen by users and their

reactions and favorites could be logged to help build humans’ mental models in

knowledge thinking. All the above methods are trying to induce nice knowledge

sharing approaches, contributing to advanced knowledge management.

Last, the quality of the discussing articles could be analyzed to make the research

more complete. That is to say, computer-assisted network text analysis or semantic

analysis methods could be utilized into the social networks in the online discussion

forum. The content and length of the articles should be concerned to add to the factor

of the weight value. An article with no positive discussion but only nonsense will be

judged less valuable than one with meaningful discussion. This is to avoid the unfair

situations happening that replying 10 word weighs 500 and replying 100 words still

weighs 500 because of the same one reply. With this function, the sociograms will

become more significant to express just like an epitome or implication from the online

discussion forum.

We hope to arouse attention to this field by the above-mentioned future studies.

There are many directions and objectives waiting for us to study hard. We definitely

hope that this thesis can be a good contribution to other researchers who are interested

in combining online discussion forums with social networks as well.