Mitchell Centre Seminar Series - Bernie Hogan (Oxford University) - Perspective Matters: Egocentric Networks and the Bundling/Grinding Problem in Sheaf Theory
| Dates: | 19 November 2025 |
|---|
| Times: | 16:00 - 17:30 |
|---|
| What is it: | Seminar |
|---|
| Organiser: | School of Social Sciences |
|---|
| Speaker: | Bernie Hogan |
|---|
|
|
Consider a whole network, such as a classroom. Each student can create an egocentric network visually and in metric space through traditional methods. Yet the aggregation of these networks remains a nontrivial problem. The conventional solutions, often treated under the rubric “Cognitive Social Structures” does not sufficiently accommodate information at the local level within the global projection. Like many aggregating approaches from linear / continuous analysis, traditional approaches ‘grind’ rather than ‘bundle’ the local data into a global space. By working through the axioms of sheaf theory from topological data analysis and more broadly from discrete math, we can conceptually accommodate all of the local information in an egocentric “presheaf” and consider the resulting whole network as a “sheaf”. This conversion not only opens up a host of potential avenues for novel analytical modelling from topological data analysis but also reveals analogies with the estimation process involved in transformer models (aka LLMs) while providing a strikingly aligned framework to White’s elusive ‘catnets’. Yet, the most mind-boggling implication of this work is the necessity of and capacity to include observer perspective (i.e. subjectivity) into our mathematical framework with analytical precision.
Speaker
Bernie Hogan
Role: Associate Professor
Organisation: Oxford University
Travel and Contact Information
Find event
2.016
Arthur Lewis Building
Manchester
