I study the phenomena of reference and predicativity in linguistic systems: how words and phrases come to stand for things in the world, the physical and social grounding of that “standing for”, and how it is we can talk about those things. Reference can be designed or engineered, and it is the special problems of reference in artificial systems that draw me. Codified systems of reference are critically important in science and engineering. They support, among other things, software engineering (e.g. references to computational objects and software packages), scientific research (e.g. taxonomic, chemical, and astronomical names), Internet communication (e.g. web addresses and document identifiers), and data interoperability (controlled vocabularies and formal ontologies).
In programming language design, one aims at preventing surprises when reference is attempted in programming contexts. Design problems I’ve examined include the control of reference and visibility between program modules, reference in uses of syntactic extensions (‘macros’), and the use of scope control as a security mechanism. In addition to within-address-space references there are references to external entities in databases, repositories, registries, and knowledge bases. Data interoperability requires a theory (‘ontology’) of the entities that data is about and of the ways one refers them. The need for data interoperability led me to get involved in technical Web standards (the “semantic web”).
My recent work on reference has been focused on the 250-year-old system of nomenclature for biological taxonomy. This system has challenging twists and turns since often multiple names apply to the same entity and a single name may apply to multiple entities. These confusions present threats to the useability of biodiversity data. Detecting synonymies and polysemies automatically is prerequisite to quality control for data aggregators.
Work in these applied areas has made me wonder how reference reduces to social and physical processes. Philosophers who work on this question provide insight (Ruth Millikan has the best account I’ve seen) but don’t give a well grounded theory of reference that might be implemented on a robot (i.e. a computer with sensors and actuators). Computer software does not usually go beyond mere template-filling and preprogrammed rules. Ultimately we would like robots that can genuinely “know what they are talking about” and can refer not just to the highly controlled fiat “objects” in their own address spaces, but to genuine objects (physical and otherwise) in the real world.
Learn more on Google Scholar