Allowing third-parties to create interfaces onto OGD can lead to a diversity of interfaces. Within days of the release of COINS data at least four different platforms were available for navigating the dataset (C1-C4), and open-source code was available that actors could use to generate their own interpretation of the data via familiar tools (C5-C6). Each interface has a differing emphasis. The RA.Pid (C2) explorer focuses on presenting quantities of spending graphically, whereas the Guardian’s COINS explorer (C3) focuses on crowd-sourcing more information about the nature of specific transactions and categories in the dataset.
Creating interfaces onto data, and informational representations of datasets, involves making value judgments, both about what to present, and how to present it. Reflecting on the changing role of newspapers in relation to data, data-journalist Simon Rogers explained:
“I think our role as gatekeepers has changed to become roles as interpreters to help people interpret data and use it and get the most out of it, which is a tricky thing.”
Schooloscope (E4) is similarly aware of its role as an interpretation layer on top of raw-data. Rather than present statistics as UKSchoolMap (E1) does, Schooloscope provides a visual and textual narrative about the quality of each school, applying an algorithmic interpretation of the data on each school to make qualitative statements such as “Pupils at The Cherwell School are more or less content”. It is both a search interface and an information source: articulating a value-laden interpretation of each schools’ performance.
Economic forces (Schoolscope has had over £100,000 investment; the unfunded UKSchoolMap suffered downtime and accuracy issues during the data collection period), the link structure of the web (Hindman 2009; Guardian school appeals data appears high in Google results than data.gov.uks raw-data), and wider structures of media networks (Castells 2009), are likely to continue to impact on the practical availability of different interpretations of data to end-users.
Non-obvious interfaces onto data (i.e. doing more than putting geo-data on a map, or transactional data in a table) can be significantly more complex to create than straightforward representations that follow the existing structure of data.