Situating Knowledge

Gahegan and Pike's representation of seen and unseen issues that situate data

Gahegan and Pike's representation of the visible and typically expressed issues (as ovals) along with the implied--and typically difficult to express outside of narrative--issues that also serve to situate a piece of data.

The more I’ve become embedded in creating digital tools and objects, the more my scholarship languishes. I’d much prefer writing code or manipulating data to writing a paper about it or, so it seems, even reading a paper about it. But I’ve begun to catch up on all that reading, and just finished the excellent “A Situated Knowledge Representation of Geographical Information” by Mark Gahegan and William Pike. While directed at geospatial data, they highlight an issue growing ever more critical as scholars deal more and more with processed data: how to understand and express the effects of hidden productive and managerial factors on data. Or, as the authors put it:

In GIS and most other computational systems, the comprising datasets, methods, conceptual models and other resources are usually treated as objective entities. However, they were created for – and are applied to – specific situations that affect their objectivity in many ways. These situations are generally not captured or represented at all, except perhaps in the minds of the people involved in their creation and use, or after the fact in journal articles far removed from the actual processes of inquiry. And when data producers and data consumers do not share such situational knowledge, the chances of misunderstanding and misapplying resources increase.

If anyone is aware of other scholarship that explores this issue outside of the geospatial realm, I’d very much welcome a link.

This entry was posted in Algorithmic Literacy. Bookmark the permalink.

2 Responses to Situating Knowledge

  1. jb says:

    I didn’t read the Gahegan and Pike work, but looking at the chart, it seems similar to the Reliability/Validity divide in social science. Psychometrics has the most developed theory, especially Cronbach’s work and Messick’s (1995) model of construct validity. That model is standardized in the APA/AERA/NCME standards.

    The basic idea is that on one side you have the technical, objective aspects of data (measures or scores in the case of psychometrics), which are analyzed through reliability. On the other side, you have the *interpretive* issues of making use of the data, which is validity. (As a note, you cannot have validity without reliability. That is, if the explicit components don’t line up, the implicit components won’t either.)

    Reliability is operationalize as various forms of consistency: The data may be more or less consistent across situations/context or consistent in and of themselves. Data are never “valid,” but the interpretations (and uses) we make of them are more or less valid depending on the degree to which evidence and theory support them.

    Establishing validity is usually a matter of making the implicit aspects explicit, then establishing that explicitness with more data. In psychometrics, the data are usually normative, which may or may not transfer into the digital humanities.

  2. Elijah Meeks says:

    Thanks, jb, that gives me a few leads to follow up.