Quantifying Assemblage Theory: A Conceptual, Empirical, and Data-Driven Approach to Guide Discovery

Abstract

Inspired by the idea that digital text can support examination of the sociomaterial world, we present a novel computational approach, guided by assemblage theory, to understand the emergence of assemblages of personal automation practices. We overlay assemblage theory on machine learning models for text analysis, clustering, and dimensional reduction to turn these computational models into vehicles for operationalizing the possibility space, points of attraction, and other key assemblage theory constructs. We also use an inductive analysis to deepen the insights from our computational findings. We demonstrate our approach for visualization and discovery on a unique data set of conditional text-based rules of the form “if this, then that,” defining Internet-connected applets that consumers can run to automate events in their daily lives. We uncover 127 personal automation practices and interpret the material roles of their components, as well as the expressive roles that emerge in the realized possibility space. We also show how the full possibility space provides a dynamic dashboard that reifies both what is, as well as what could be. Our results can stimulate additional research and encourage practitioners to look beyond use cases to those underlying points of attraction driving consumers’ deeper needs to automate their lives.

Keywords: assemblage theory, automation practices, consumer experience, HDBSCAN, Internet of Things, natural language processing, possibility space, word2vec, word embeddings, UMAP

JEL Classification: C38, C49, M31, M39