Exploring Impacts of Arts Participation through Automated Inductive Annotation

Researchers have worked to identify the varied outcomes associated with participation in the arts. Historically, those efforts have tended to take a 'top down' approach in that they focus on specific types of outcomes, or existing theoretical orientations. In this paper, we demonstrate an approach to annotating transcribed interviews that blends inductive human coding with machine learning in order to support 'bottom-up' discovery of previously unrecognized outcomes. We utilize a recent collection of retrospective semi-structured interviews with 102 international participants in the arts. The annotation approach yields a system for clustering and cataloging the interviews according to a set of modeled topics that facilitates corpus-based research, and may promote identification of new categories of outcomes. The work is one part of a project to create a community-driven taxonomy of outcomes of arts participation that encompass a universe of reported outcomes not delimited by domain, type, or theoretical orientations of stakeholders.