Built to tweet


The use of Twitter at conferences seems to be growing, and I think we are beginning to see some limitations of the current tool suite with respect to making use of tweets at a conference in real time. At CHI 2010 I was not able to participate much in live-tweeting because I did not want to carry my heavy Thinkpad T61 around all day, and my iPhone wasn’t up to the task. While the iPhone was adequate for checking e-mail and using the CHI 2010 schedule app, the battery would run down by the end of the day of intermittent use. Furthermore, the screen wasn’t large enough to take notes, type tweets in a timely manner, and to keep up with the stream of tweets from other attendees. In fact, in some cases it seemed that people who were following the conference remotely had a better grasp of the breadth of activity in the sessions than I did at the conference.


Based on this experience and on the experiences of some others with whom I talked, it seems that a capable live-tweeting solution should have the following characteristics:

  1. Light weight. The device has to be portable enough to carry it from session to session over multiple days without it getting in the way when it is not being used.
  2. Full-size keyboard. The ability to type a tweet or a note to yourself quickly is important when live-tweeting. Delays in typing can cause you to miss things that are said, although even then it may be difficult to type and listen at the same time.
  3. Long battery life. A device that can be used actively for 6-8 hours without a recharge may make it possible to dispense with a charger. This not only reduces the weight that has to be carried, but also makes it possible to sit away from the wall, away from one of a few power outlets typically found in large hotel meeting rooms.
  4. The device should be usable from one’s lap, as desks or other surfaces are rarely available.

One class of devices that seems to fit these constraints is a NetBook or similar small laptop. Battery life on these devices can be as high as 8-9 hours on an extended battery, they are relatively light-weight, but many are large enough for comfortable typing even for people with relatively large hands. And of course you can put it on your lap and still type.


In addition to the hardware-related factors described above, there is also the issue of software. The current one-application-at-a-time limitation of the iPhone and iPad makes it difficult to use a suite of applications for this task, but no existing application seems to provide all the required features. Efficient tweeting benefits from the ability to switch among the following functions: taking notes, typing tweets, and monitoring the relevant feed. Occasionally, it may be useful to switch to do web search or to switch to a browser to display a specific URL. Yet switching back should not reset the state of your application, and it should be possible to toggle quickly between the browser and this application (whether it runs in the browser or not). For some purposes, a camera or audio recording capability might also be useful.

The figure below is a sketch of some of these features formatted to fit on a small screen.

1. Person selection drop-down; 2. tweet text; 3: hashtag selection drop-down; 4: tweet preview; 5: notes; 6: tweet stream tabs; 7: documents tab

The tweeting view should allow a consistent set of tags to be applied automatically to tweets, reducing the load on typing. Predefined person (1) and tag set (3) templates should be available with a single button-click, and the interface should automatically constrain the length of the tweet being composed (2) to accommodate the tags. A preview of the tweet being composed (4) should be shown, and should be editable to depart from the template when necessary. I like to start tweets with the name or twitter screenname of the person talking, followed by my rendition of what was said, followed by general hash tags. Specifying this through a drop-down or other similar preset (but expandable) list should make it faster to compose certain kinds of tweets.

The note-taking area (5) should allow light-weight recording of text, and, hardware permitting, drawings. Features that support complex editing are probably best saved for an off-line tool.

It may be useful to monitor other people’s tweets related to the same event (6) to understand other people’s perspectives, to engage with commentary, to see what else is going on in parallel, etc. As tweet streams from an event get larger, it may be useful to sub-divide the streams into more focused subsets. For example, the CHI 2010 stream included hashtags for rooms (e.g., #c1, #c2, #hfg, etc.). Although their use may be less reliable than the over-arching #chi2010 tag, they may still serve a useful grouping function. Thus it should be possible to set up several ad hoc standing queries against the twitter stream similar to the way that TweetDeck search columns work. Unlike TweetDeck, however, the interface should show only one or two at a time, relegating the rest to tabs. Each tab should show how many matching unread or newly-arrived tweets there are to give a sense of overall activity.

Another tab should contain references to documents (7) that have been mentioned in the tweets along the lines of the prototype I described earlier. This will allow the user to defer link following without losing the references, and will implement a voting mechanism based on the number of mentions for assessing the relative importance of the document.

I am considering building something along these lines, and would be happy to collaborate with others with similar interests. It would be interesting to compare this tool to other configurations of hardware and software to see if the integration works. I expect there will lots of surprises about which features are useful, which are unneeded, and which I haven’t thought of at all.


  1. Yes, the current software and hardware set doesn’t scale well.

    Unless, you aren’t interested in what happen before and are focused only on what is happening now.

  2. I found myself swamped at CHI even with the “now” tweets. I can only imagine what it will be like next year if Twitter use permeates further into the community.

  3. As we discussed at the workshop, Twitter currently is a pretty plain communication platform, and much should be done to really make it a usable platform. What you propose here sounds pretty interesting. It’d be neat to really try it out!

    At a conference, some simple location-based service would also be useful. What’s going on where, who’s in which room, where the good food is, etc, Of course, once you start to add all the useful features, the software might get too heavy and lose its original beauty and simplicity.

  4. It should be possible to add a map-based view to organize tweets based on whatever geo-tagging is present. It is unclear, however, how reliable the information is, and at what scale it is useful. For example, I expect that many people who use a conference’s hash tags are not actually at the conference. For some tasks, their geo-tag information should not be used, while for others it would be useful. This is where having an empirically-determined set or taxonomy of tasks would be useful to guide specific design decisions.

  5. Sounds great, Gene!

    The thought that remote participants could sometimes have more perspective than attendees reminded me of the notion of “local virtual” vs. “remote virtual” which Andy Powell has written about:

    I’ve definitely used laptops, rather than netbooks or iPhone, at conferences because of the high penalty of context-switching on non-laptop devices.

Comments are closed.