Observations

Web-based services that facilitate online communications divide into two categories:

1. Services that augment a communication elsewhere, such as Google, Amazon or PhotoBucket (hosts pictures for posting to a weblog or other discussion system).
2. Services that host the communication itself, like Yahoo! Groups (web-based interface to mailing lists) or LiveJournal (weblogs).

A service’s user interface is strongly tied to which of these categories it falls into. People frequently pass around links to Google search results, but Google itself doesn’t transmit the communication. On the other hand, Google does ensure that users are not hindered from passing around links to search results. The URLs are short and human readable. The search results displayed are the same no matter who opens the page. Google intentionally does not customise results based on past searches because that would prevent people from passing links around. Amazon does the same with book information pages.

Both these examples are about user interface as presented in a URL. For services that augment communications, URLs are the most critical element of the user interface. Services that host communications need to design their entire interface very carefully.

Two of the most popular weblogging systems in use today are Movable Type and LiveJournal. I’ve previously compared them and expressed my dissatisfaction with how Movable Type discourages discussion on a post. (While MT is software and LJ is a service, the difference is immaterial for the purpose of my comparison. LJ’s software is also called LiveJournal and MT’s service is TypePad.) I believe now that this comparison was immature. The limitations of Movable Type’s commenting system, combined with the Trackback feature, creates a possibility that LiveJournal doesn’t encourage.

As the number of weblogs increases, there is a limit to how many I can track in my available time. Now here is the problem: what if someone I don’t read makes a very interesting post, something I would have wanted to read? I’ve argued for why LiveJournal’s commenting system is superior. If an interesting post was made on LiveJournal, a huge discussion could have followed, yet I wouldn’t even be aware because the post was off my radar, and the discussion was contained within the post.

I think that Movable Type’s limited commenting system and Trackback feature makes a difference here. Because MT’s commenting system lacks identity (anyone can use any name) and control over formatting, a person with serious commentary would rather post to his own blog (with a summary of the original), and then use Trackback to link his post with the original post. Further commentary happens on another blog, linked again with Trackbacks. The idea keeps hopping this way, until it lands in one of the weblogs in my radar, from where I can trace back to the source.

The hopping around unfortunately also creates the impression of an echo chamber, of people talking to each other oblivious of the world around them. The nature of a trackback-powered discussion makes it difficult to link to any one page as representative of the entire discussion, and I think the difficulty of tracing the conversation also turns off anyone with only marginal interest.

But the point I’m trying to make is not about the social aspects of weblogging. It is about how the user interface fashions the social aspects. Implementing threaded discussions or trackback is trivial for either Movable Type or LiveJournal. What is interesting is that the mere presence or absence of a small feature can create such clear-cut conventions.

In real life, we have real physical limits to the user interface. A magazine that is stapled in the middle has to have an even number of sheets because a single sheet can’t be stapled in. The fixed size of a sheet means that a story has to be edited to exactly fit the sheet, sometimes by cropping out a part that may have been important. Online, changing the user interface is relatively trivial and very powerful. Unfortunately, several sites fail to understand how form affects discussion, and end up with models that don’t provide expected results. This is most visible in the “Talkback” section on most news sites. Users have no identity, navigation is designed to maximise ad impressions, and moderators impose severe restrictions on the discussions. The intent is to restrict discussion to feedback for the author’s article, but the end result is that this rarely happens. People with serious commentary would rather post it to their own weblog, where they have greater control.

Work Plan

I intend to study cause-and-effect relationships in user interface design on a few leading social service sites. Specifically, I intend to show how community members use each component of the user interface, and how this in turn shapes the community. Current candidates are Ryze.com, LiveJournal.com, a few Movable Type and WordPress weblogs, Flickr.com and PhotoBucket.com. I will consider some more during the research process if they prove to be valuable case studies.

I realise that user interface is not the only thing that affects community, and that it is therefore impossible to establish an exact cause-and-effect relationship. For this reason, as the second stage of my research process, I intend to take my findings to a community that is willing to implement necessary changes, and observe their effects.

Timetable

January 1 to February 28, 2005: Record observations on user interface and effects.

March 1 to June 30, 2005: Take findings to a willing community and observe changes.

July 2005: Collate observations for final presentation.

Addendum

Commentary on this proposal is at my journal.