Auditing Algorithmic Communication Flows

Bandy, Jack

doi:https://doi.org/10.21985/n2-ef0p-ez52

Work

Auditing Algorithmic Communication Flows

Public

Download PDF

Download All Files (.zip)

Human communication has become increasingly reliant on systems made and managed by large technology companies like Google, Apple, Twitter, and Meta (formerly Facebook). These systems offer people many benefits, but they also present new challenges for society. In recent years, researchers, lawmakers, and journalists have suggested that large technology companies and their systems can have a wide range of perverse effects, such as exacerbating political dysfunction, infringing on personal privacy, reinforcing racism and inequality, harming teen mental health, and more. Scientific understanding around these effects is still developing, and companies often complicate this understanding by presenting different narratives in marketing, lobbying, and even their own research publications. Public understanding of large technology companies is thus still progressing. This dissertation represents one effort to make sense of large technology companies, and specifically their algorithmic systems which permeate today's communication environment. I describe Google Ads, Apple News, Twitter Timelines, and Facebook Feeds as algorithmic communication channels, presenting four respective studies that empirically characterize these systems using algorithm auditing methods. Some empirical findings from these audits suggest potentially problematic tendencies, such as ambiguous audiences, high source concentration, slight echo chamber effects, amplification of low-quality sources, and more. Overall, the four studies demonstrate algorithms as infused, co-constituent components of digital communication, with at least three common characteristics described in the final chapter: (1) distributed agency, shared with people who operate them and people who use them, (2) ambiguous audiences for people who send messages through algorithmic channels, and (3) exchanges and connections with larger communication networks, including other types of actors and channels. Taken as a whole, this dissertation makes incremental contributions to understanding algorithmic systems, especially what they are and what they do. I conclude by discussing potential next steps for responding to the asymmetries of knowledge and power in the digital communication environment. Within the broader body of research focused on large technology companies, potential problems identified in the four audit studies are best understood as symptoms of a more fundamental issue regarding political economy. Even if technology companies turned to non-profit business models, strong privacy policies, improved algorithmic (or non-algorithmic) content ranking, and other technical ``fixes,'' we would still confront the fundamental political issue regarding who has power in the algorithmic communication environment. Currently, billions of people rely on communication systems created and governed by a small group of technology executives, and the public has little to no say in how these systems operate. As we work on the communication environments of the future, we would do well to confront this underlying issue of governance, in addition to analyzing its symptoms.

Creator