STAI CDT PhD student, Luke Thorburn (2021 cohort), has collaborated with Jonathan Stray (Center for Human-Compatible AI at UC Berkeley) and Pri Bengani (Tow Center for Digital Journalism at Colmbia University) to create the blog, Understanding Recommenders. We spoke to Luke about the blog and the opportunities that have stemmed from his work and being part of the CDT.
What led you to start this blog?
We started the Understanding Recommenders blog to communicate ideas related to the impact of recommender systems on society in a way that respects technical nuance but is accessible to people working in policy or civil society roles who may not have a technical background.
What is a recommender system?
A recommender system is an algorithm that takes a large set of items and determines which of those to display to a user — for example, the Facebook Feed, the Twitter timeline, Google News, or the YouTube homepage.
What is the main focus of the blog?
Each article focuses on a different theme related to recommenders and their societal impact. The aim is for the articles to serve as useful reference pieces for some years to come. So far, we have written about how recommenders work, preferences, engagement, amplification, measurement, transparency, and influence.
What have been some outcomes of writing the blog?
It’s been very rewarding to see that the writing is reaching its intended audience. We’ve received positive feedback from regulators, from employees at large social media platforms, and from researchers in industry and academia. On a personal level, this project has been an excellent means to familiarise myself more extensively with the literature. The project has also led to discussions with journalists, platforms and other academics.
How do you find being part of the CDT?
The CDT has provided a very supportive environment from which to pursue collaborations such as this. I’m grateful for the community we have and the flexibility we’re allowed when crafting a PhD experience.