• information visualization/data mining: Releasing the structure hidden in data in visual form, being able to explore it and learn things from it that I would not otherwise have known. Recombining data in new ways.
• art and design: making beautiful, breathtaking, meaningful visual things
• sense-making: making sense of chaotic or muddy information/knowledge, understanding it and making it available back to the community
• organizing/efficiency: organizing information so that I and others can more easily find it
• knowledge sharing, mentoring, learning: sharing what I have learned with other people to save them time, and help them organize things the way I want them; also learning from them. Getting people to work together to create and learn as a community.
• web publishing: writing and publishing things on the web, and helping more people do that more easily. This ties in with art and design.
• emergent phenomena: complex behavior/phenomena from simple systems. Generative art, sensor motes.

Share This

One thing I need to remember about mailing list, USENET and web forum analysis is that those are all public spaces, and e-mail corpora are data from a private space. The techniques used to analyze the data will be different, and the structure of the data will be different, as well. For example, there is no private individual-to-individual communication in mailing lists that would be recorded in the archives — such communication is out of band. So too for USENET. For web forums, participants typically don’t have ways to contact each other privately. These media are always one-to-all communications. One-to-one communications are always performed in front of an audience, and one has to use special techniques in order to tease out these communications, which can be lossy, since replies to other messages may not be delimited by a quoted section, and the header information is not as rich as in e-mail. We can only get the To: line for messages in these media, and pretty much we can get: to one, to a group or to all.

The knowledge that all messages are visible to the entire community (or perhaps the entire world) should cause people to behave differently than they would if they knew that their audience was restricted, in theory, only to those people to whom they explicitly send their messages.

E-mail has more options: one-to-one, one-to-some. Bcc: complicates things by preventing the full distribution of a message from being known to the recipients. So the audience of a message is smaller and purposely selected, although there is always the danger of forwarding of messages (see [1]). Cc: has a different relational implication than To:, and Bcc: has yet another implication. Neither of these are present in mailing lists, USENET or web forum analysis. On the other hand, for large communities, the “to all” communication in e-mail will be rare, because it’s typically not trivial or acceptable to send to such audiences. I know that at Caltech, very few people have the right to send a communication to even everyone in our department, let alone everyone at Caltech.

1. Brown, S. D. and G. Lightfoot (2002). Presence, Absence, and Accountability: Email and the Mediation of Organizational Memory. Virtual Society , 209-229.

Share This

Great and timely post at FlowingData: 21 Ways to Visualize and Explore Your Email Inbox. It has a lot of the papers that I’d already seen: Themail, MyMap, Enron Explorer, Perer & Smith. some new stuff, too, and stuff that i’d forgotten. I’m typing with one hand, and my arm hurts now.

I need to remember to look at USENET, mailing list and web forum analysis and visualization papers. Those three things are similar enough to e-mail corpora that there may be some useful crossover ideas.

Share This

Design science suffers more from diffuse boundaries with practice than I think behavioral science does, and this makes it hard for me to wrap my mind around what exactly differentiates design science from innovative design performed outside the research community. I am very interested in using design research to propel my own research, and so knowing what characteristics it has so that I can increase the probability of that whatever I do will be accepted, insofar as the form of the research is concerned.

Hevner et. al. say that the goal of design science is to provide utility; to solve a real problem, but not to show why the solution works (p. 80). They differentiate design science from “routine design” like so: routine design applies existing IS knowledge to solve current organizational problems (e.g. implementing yet another web server using best design principles is “routine design”), while design science “addresses important unsolved problems in unique or innovative ways or solved problems in more effective or efficient ways” (p. 81). Design science is frequently applied to problems for which the necessary IS knowledge may not yet exist. Its purpose is to address what have been called wicked problems (all of the below from p. 81):

• unstable requirements and constraints based on poorly defined or understood contexts

• complex interactions among subcomponents of the problem and its solution

• malleable processes and artifacts

• a critical dependence on creativity ( intuitive knowledge processes) to provide good solutions

• a critical dependence on social abilities (teamwork) to provide good solutions

I think that are a lot of innovative solutions being proposed and implemented outside the IS research community (Yahoo Pipes and various google products, for example), so I feel like the emphasis on innovation is not an effective distinguishing characteristic.

It is, I believe, the focus on the IS research community it is Hevner et. al.’s guidelines 4, 5 and 7 (research contributions, research rigor and communication of research, p. 83) that distinguishes design science most from innovative design. Those guidelines are really saying that if we innovate with the intention of furthering the goals of science, and contextualize it within the IS knowledge base and community by writing it up and communicating it back to the IS research community, then we are doing design science.

Share This

Puade, O. A. and T. G. Wyeld (2006). Visualising collaboration via email: Finding the key players. In Information Visualization, 2006. IV 2006. Tenth International Conference on, pp. 124-129.

In this conference paper presented at the 2006 Information Visualization Conference, the authors present a proof-of-concept visualization tool which attempts to depict the relative importance of e-mails and participants in a collaboration with the goal of finding key players. Their goal for the visualization is that they hope that the groups can use it for self-reflection: to learn from what the visualization shows them, and to change to then collaborate better. 24 e-mails from a group of game developers working on a software development project provided the data for the visualization. The authors recruited each of these ten people for the purposes of this study.

Data were prepared for the visualization as follows:

1. First, categorize e-mails, using Divitini and Farshchian’s [2] e-mail roles: “Awareness”, “Decision making”, “Accessing expert”, “Feedback”, “Resolving issues.” A single e-mail can be in more than one of those roles.
2. Ask each of the ten people to rate each of the 24 e-mails in terms of importance on a scale of 0-3, where 0 = not applicable, 1 = not important, 2 = important and 3 = very important.
3. Call the average rating for across messages for each person the “loudness” of that person, and make the “impact” of a person be the “loudness” times the number of e-mails.

They used two visualizations based on cocentric circles divided radially into pie sections. The first visualization impact by participant, and the second by message type (which included, in this case, subtypes). Each figure plotted in the visualization combined number of e-mails, loudness, and impact. They plotted on the same visualization (I think) an undirected graph of e-mails (?).

Analysis

The visualization showed something that they didn’t expect: “there is more variation between how the importance of participant’s emails is perceived compared to the importance of the types of email sent. However, emails of type ‘awareness’ clearly have a greater impact on the collaboration because of their dominant use” (p. 6).

They used Divitini and Farshchian’s [2] work to create an expert system to analyze and categorize this archive, and made some kind of comparison that I have to go back and understand better.

Issues

They mention that “the collaboration involved various activities in different location and continents” (p. 2). I couldn’t tell whether the group was typically co-located or was virtual. They do mention virtual teams on p. 1, in section 1.2.

They mention two other studies which will be very useful to me: Perer et. al. [2] and Divitini and Farshchian [1].

Critique

• Although the authors say that their goal for the visualization is that they hope that the groups can use it for self-reflection, as far as I can tell from the paper, the subjects never see the visualization.
• They analyzed only 24 e-mails — it seems like this is far too small a sample size to truly illuminate social structure.
• Where did this visualization come from? Why choose something like this? No explanation is given.
• Self-evaluation bias? People are rating their own e-mails.
• The introduction of the paper is great — I should use something like it for my own paper.
• The visualization will not scale well to many more people — for small groups only.
• Do all participants have the same definition of “importance”? Does this matter?

Argument

It’s not easy for me to see where they’re going with this work. The argument is not very strong:

They want to help collaboration participants create better collaboration outcomes via self-reflection (p. 2).

1. “Visual representations aid and enable the user to understand the different kinds and forms of data” (p. 1) especially of large amounts of data.
2. Creating an overview allows the user to see the big picture (p. 1).
3. We can create an overview visualization of the structural and behavioral aspects of a collaboration (p. 1)
4. We can use e-mail data to create the overview, because e-mail is a key collaboration medium (p. 1).
5. We can use this overview to identify various user roles and their grand purpose as to the collaboration as a whole (p. 1).
6. Node and edge graphs are misleading in identifying key players (p. 3), so something better is needed.
7. If we classify e-mails by type, importance and impact, we can better identify key roles (pp. 3-4). Do it by hand now; we’ll eventually do it by computer (p. 3). Importance and impact are defined as above, and meaningfully map to social roles.
8. Make a visualization of this data.
9. This visualization will enable collaboration participants to identify key players, and thus to do self-reflection.

The argument gets weaker and weaker after step 5.

References

1. Divitini, M. and B. A. Farshchian (1999). Collaboration and coordination through basic internet tools: A case study. In World MultiConference SCI/ISA, Orlando, Florida.

2. Perer, A., B. Shneiderman, and D. W. Oard (2006, December). Using rhythms of relationships to understand e-mail archives. J. Am. Soc. Inf. Sci. Technol. 57 (14), 1936-1948.

Share This