Comments on MyLifeBits: A Personal Database for Everything

Summary


In MyLifeBits: A Personal Database for Everything, Gemmell, Bell and Lueder introduce the use of an SQL database to supplement and organize the collected data in their wide-ranging Personal Informatics project. In their project the authors seek to amalgamate and access data from the entirety of a person's life, ranging from their location to their work production to their internet usage.

The author visualize their system as a "Memex" which was a conceptual device that stored, indexed and retrieved different types of information (largely as microfilm) and acted as a type of extended personal memory, allowing the user to store a huge amount of information in a structured an indexed environment.

In the original use of the system the users organized the majority of their data and information using the directory structure of the underlaying file system. However this limited the searchability and retrievability of their data and made it difficult both to organize the information and to link together connected pieces of information.

To resolve this problem they introduced an sql database at the heart of their system that is responsible for tracking both the information they collected and the meta-data about the information. Using this approach they were able to achieve full text search over all textual information they collected and all meta-data for non textual information and they were able to tie together different pieces of information with special links in the database. In cases of location and time they were able to automate the process of storing and organizing this information.

Thoughts

Technically, I'm struck that this system lacks flexibility. In particular adding new types of information to the system seem to require manually adding not only the meta-data to the database but possibly even the meta-data and types. In addition while the database reduces some of the work necessary to organize the collected data there still seems to be a fair amount of manual (or at least scripted) work required to organize the data as it comes into the system.

While the authors are using this system to investigate personal informatics, I can't help but feel that the system is not of particular benefit to the users. Tracking all possible information a person experiences seems as though it would be both quite invasive at the point of collection and very difficult to organize.

Clearly in the years since the paper has been published changes in technology have made it easier to implement many of the things they would like to use their Memex, ideally even phasing out the concept of one location physical for all the given Memex into the cloud. Additionally, the flexibility of a no-sql database might also prove to be beneficial for their implementation as it allows different schemes to access information than either filesystems or the plain sql implementation. It seems that the amount of data which can be recorded will continue to increase for the foreseeable future, and in order for this type of system to be useful for people, the interface for it will need to be much more carefully considered.