Bodies of Knowledge: Processes of Social Construction and the Epistemology of Empirical Measurement in Computational Social Science
| dc.contributor.advisor | McCormick, Tyler | |
| dc.contributor.advisor | Johfre, Sasha | |
| dc.contributor.author | Visokay, Adam | |
| dc.date.accessioned | 2026-09-16T18:35:14Z | |
| dc.date.issued | 2026-09-16 | |
| dc.date.submitted | 2026 | |
| dc.description | Thesis (Ph.D.)--University of Washington, 2026 | |
| dc.description.abstract | This dissertation argues that taking the relationship between conceptualization and measurement seriously as both a theoretical and empirical problem generates contributions that are methodological, substantive, and sociological. Across three empirical studies, I show that the categories through which social scientists observe neighborhoods, bodies, and causes of death are artifacts of measurement systems that carry their own histories and assumptions, and that choosing more conceptually aligned measures can unsettle conclusions that may be taken for granted. Using large language model annotation of Craigslist rental advertisements, I show that administrative neighborhood boundaries overlook the process by which listing agents claim and construct urban space, revealing patterns of spatial representation invisible to coordinate-based approaches. Comparing BMI-, waist circumference-, and total body fat-based obesity prevalence trends of the US adult population from 1999–2018, I find that while BMI-defined obesity rose 13 percentage points, obesity derived from directly observed body fat showed no statistically significant change – and that this aggregate estimate masks a pronounced gendered divergence. Addressing cause-of-death estimation where 40 percent of global deaths are not accompanied by a traditional death certificate, I use a validation dataset to demonstrate that language models applied to verbal autopsy text narratives can match or exceed structured-data classification performance, and introduce multiPPI++, extending prediction-powered inference to multinomial outcomes for valid population-level inference from predicted labels. Together, these studies advance an epistemology of empirical measurement: one in which how we observe and categorize the social world is treated not as simply preliminary to quantitative analysis, but as fundamentally constitutive of the knowledge it produces. | |
| dc.embargo.terms | Open Access | |
| dc.format.mimetype | application/pdf | |
| dc.identifier.other | Visokay_washington_0250E_30134.pdf | |
| dc.identifier.uri | https://hdl.handle.net/1773/57884 | |
| dc.language.iso | en_US | |
| dc.rights | CC BY | |
| dc.subject | artificial intelligence | |
| dc.subject | machine learning | |
| dc.subject | measurement | |
| dc.subject | obesity | |
| dc.subject | social construction | |
| dc.subject | statistical inference | |
| dc.subject | Sociology | |
| dc.subject | Statistics | |
| dc.subject | Artificial intelligence | |
| dc.subject.other | Sociology | |
| dc.title | Bodies of Knowledge: Processes of Social Construction and the Epistemology of Empirical Measurement in Computational Social Science | |
| dc.type | Thesis |
Files
Original bundle
1 - 1 of 1
Loading...
- Name:
- Visokay_washington_0250E_30134.pdf
- Size:
- 3.93 MB
- Format:
- Adobe Portable Document Format
