Enacting the Semantic Web: Ontological Orderings, Negotiated Standards, and Human-machine Translations

Artificial intelligence (AI) that is based upon semantic search has become one of the dominant means for accessing information in recent years. This is particularly the case in mobile contexts, as search based AI are embedded in each of the major mobile operating systems. The implications are such that information is becoming less a matter of choosing between different sets of results, and more of a presentation of a single answer, limiting both the availability of, and exposure to, alternate sources of information. Thus, it is essential to understand how that information comes to be structured and how deterministic systems like search based AI come to understand the indeterminate worlds they are tasked with interrogating. The semantic web, one of the technologies underpinning these systems, creates machine-readable data from the existing web of text and formalizes those machine-readable understandings in ontologies. This study investigates the ways that those semantic assemblages structure, and thus define, the world. In accordance with assemblage theory, it is necessary to study the interactions between the components that make up such data assemblages. As yet, the social sciences have been slow to systematically investigate data assemblages, the semantic web, and the components of these important socio-technical systems. This study investigates one major ontology, Schema.org. It uses netnographic methods to study the construction and use of Schema.org to determine how ontological states are declared and how human-machine translations occur in those development and use processes. This study has two main findings that bear on the relevant literature. First, I find that development and use of the ontology is a product of negotiations with technical standards such that ontologists and users must work around, with, and through the affordances and constraints of standards. Second, these groups adopt a pragmatic and generalizable approach to data modeling and semantic markup that determines ontological context in local and global ways. This first finding is significant in that past work has largely focused on how people work around standards’ limitations, whereas this shows that practitioners also strategically engage with standards to achieve their aims. Second, the particular approach that these groups use in translating human knowledge to machines, differs from the formalized and positivistic approaches described in past work. At a larger level, this study fills a lacuna in the collective understanding of how data assemblages are constructed and operate