EmergingNarratives_LP

Neat and Scruffy traditions of AI bring different epistemic commitments to the development of Semantic Web infrastructure. Neat traditions (in lineage of McCarthy, Hayes) of AI tend to emphasize ontologies, controlled vocabularies and other "formal" approaches to semantification. Scruffy traditions (in lineage of Minsky and Schank) tend to emphasize the importance of open and linked data over formally organizing data. Further, those who are coming out of hypertext traditions tend to emphasize the value of linking data, above all.

Semantic people don't really like the way big data is characterized. They want to see //meaning// in data - broad data.

Semantic Web developers are already seeing limits to knowledge representation and responding to them with calls for "pragmatism" and shaping "middle ontologies."

Web ontologies are trading zones where diverse researchers with diverse vocabularies/commitments come together, and don't necessarily negotiate, but do get stuff done. Importantly though, the way this gets embedded in infrastructure means that different vocabularies/commitments manifest at different points in the infrastructure.

Environmental justice data is particularly difficult to visualize on a map because it requires data standardization amongst several diverse data sources. For instance, the data that the EPA leverages to document environmental justice is reliant on both U.S. census and the National Air Toxics Assessment, and to visualize this data, the EPA has to negotiate different data standards and dimensions. In their EJSCREEN tool, the air quality data that gets mapped is national in coverage, making it difficult to also map health data at the local scale. The time intervals of air quality data can be hourly, while the time intervals for census data can be yearly. These complex spatial and temporal dimensions of data already pose challenges to making environmental justice data knowable.

The EPA sees the Semantic Web as an opportunity to enable data mashups, which they believe could highlight potentially hidden data stories.

The only time the classifier ‘Race’ appears in any of the standard linked data vocabularies (aside from several references to ‘Race’ as a sporting event or a game) is in the ‘Appearance’ vocabulary – “an ontology about sex, gender, skin, eye, and hair color.” In this sense, the only way to account for the social dynamics around race using Semantic Web infrastructure is in describing the way an individual looks. Notably, there are no standardized vocabularies that allow a data practitioner to mark when a community is marked by race. Existing Semantic Web infrastructure thus cannot make it known that black communities tend to be disproportionately exposed to toxic chemicals and air pollution.

Edwards: hyper-text/hyper-tension, also not closed world. They say it's open but I think it's something else.