Post date: Feb 11, 2014 5:52:33 PM
Anne Wright from the CREATE lab came to class today to give a guest lecture on the Quantified Self work that she has been doing. Her story showed a personal approach to managing data that had many parallels with the overall pipeline we've been studying in class.
She talked about some of the devices that already exist to help people track their data as well as other sources of data such as smart phone apps, regional data, air quality, and so on and so forth. The BodyTrack project supports bringing together a combination of context and physiological data.
One of the insights of the talk was the difficulty of predicting what will matter, and the importance of supporting reflection for that. 'Constructing a personal narrative is an interactive, individual process' -- and the process is crucial to the outcome. Equally important is how you design the protocol -- what do you track automatically, what do you self record, can you see relative timing of events at a scale that matters?
Then Anne introduced fluxstream.org, the website that she created to support self tracking. The site can draw data from many sources and visualize it. The site does not impose goals or judgements on the user, and does not support statistics. The discussion around this highlighted the fact that things that are statistically 'significant' to the point that they have an effect on the world tend to also be visible in the data (or the effect would not be big enough to matter, probably). We also talked about the potential negative of setting goals and judgements based on studies of other people. Instead, judgements by the person themselves around their own data was highlighted as the approach supported by fluxstream.
Anne then talked about key criteria for things that fluxstream can pull in: Useful for reflection; convenient cloud push; and a good API. Other issues: good timestamps are also important (e.g. is there timezone information?). An ability to ask for only what has changed could make things much easier (many don't support this). The differences in the data also lead to the need to create custom visualizations.