MA Thesis Defense: Toko Dougherty

‘Unifying epistemic opacity’

Toko Dougherty (Cal State Long Beach)

08 Jan 2026 •  2:00pm–4:00pm  •  MHB–915

Abstract: The widespread deployment of complex computational models, particularly deep neural networks, has generated substantial multidisciplinary interest in epistemic opacity—the phenomenon whereby computational systems resist explanation and understanding. However, the literature remains highly disunified, with researchers developing purpose-specific forms of opacity that resist conceptual comparison. This thesis introduces a general schema that defines any form of opacity by fixing two parameters: (1) modal strength (x), which specifies the type and degree of impossibility (epistemic, technological, etc.), and (2) an abstraction level (L), which identifies the inaccessible aspect of the process (computational, algorithmic, etc.). The schema subsumes major existing frameworks and taxonomies while introducing computational-level opacity as a novel form. By reducing diverse epistemic relations to possibilities of knowledge at different abstraction levels, the schema enables systematic comparison across philosophical and explainable AI approaches, providing a unified foundation for future research on computational opacity.

Oral defenses are open to the public.

For a draft copy of the thesis in advance, please contact Toko <toko.dougherty01@student.csulb.edu> directly.