LeeIverson

• I met for a couple of hours with Lee about the work that he and his graduate students have been doing in the PIM area and how we might work together particularly in the area of user testing. This is what I understand to be the gist of their work:

• Lee's group is doing some interesting work in the area of using personal relationships to enrich PIM experience. One particular research project has focused on organizing email by person (along the lines of Ducky's Outlook approach)
• Some factors they looked at to determine priority were:
  • Speed of reply to emails from person X
  • If the email is a reply to an email you sent
  • If the reply email was sent quickly
  • # of messages exchanged in a particular thread
  • Whether you're in the To: field or CC: field

• They're also interested in versioning attachments and treating them as persisent resources in the PIM where users can access version histories. This goes very well with our approach to treating attachments as first-class items that can be versioned.

• They're also looking to how relationships between people interact with security and sharing. We're struggling with many of the same issues and this is potentially another complementary area.

• We talked about possibly having Lee's research group do some formal user-testing combining some of their work with Email with some of our Chandler affordances (ie. Triage and Stamping workflows)
  • One example might be that they could include Stamping and Triage interaction into their email prioritization algorithm

• We agreed that we would like to separate "conceptual" usability testing from testing interaction design

• I also mentioned the possibility of instrumenting Chandler to collect click-path data from users over prolonged periods of time for analysis.

• Lee expressed some concern about modeling stamped items as a single item. However, after a long discussion, I think we came to the conclusion that there was a misunderstanding about how collections worked and how heavyweight they were.

-- MimiYin - 10 Jul 2004

This is a pretty good summary of our discussion. I'd like to highlight a few things.

• To this point, we've done our prototyping and analysis in Mozilla. It turned out that this was much harder to manage than we had supposed going in. The work that we've done on "relationship modelling" and message prioritization (triage) was pretty limited by difficulties in actually mining the information in this context. We are now doing some qualitative user studies of email reading and managing behavior and mining static email archives using Python mail tools. The hope is that NelsonSiu^? (the student working on this) will be able to incorporate these insights into Chandler blocks and parcels and then test the resulting system with users.

• We are definitely interested in user testing Chandler interfaces and conceptual models against alternatives. Perhaps a better way of understanding the separation of "conceptual" user testing from

interaction testing is to focus on situating the user testing in terms of evaluation of "design theories". These can certainly be either theories of conceptual issues or interactions, but at this point, we agreed that working out the conceptual issues is certainly more important and relevant.

• We did talk a lot about stamping vs. collection and email as task vs. email supporting a task. There are certainly both conceptual and implementation issues here (e.g. inheritance vs. composition) and we did decide that it was probably more useful to hash this out in a different context. Besides, I want to see how this plays out in the UI before I make too many judgements.

• In general a very interesting and useful discussion with lots of possibilities for moving forward.

-- LeeIverson - 14 July 2004