Data Justice Research Talk Series: Overexamined Algorithms and Overlooked Agency

Join us for a Dr. Hosseinmardi's research talk along with a question and answer session. A light lunch will be provided. 

Abstract:

In recent years, critics of online platforms have raised concerns about the ability of recommendation algorithms to amplify problematic content with potentially radicalizing consequences. Yet most attempts to evaluate these claims suffer from a core methodological gap: the absence of appropriate counterfactuals—what users would have encountered without algorithmic recommendations—making it difficult to disentangle the influence of the algorithm from users' own intentions.

To address this challenge, Dr. Hosseinmardi and her research team first examined the scale of the problem and possible explanations. While they identified several distinct communities of news consumers within YouTube, from moderate to more extreme, they found little evidence that the YouTube recommendation algorithm is actively driving attention to problematic content. Overall, the findings indicate that trends in video-based political news consumption are determined by a complicated combination of user preferences, platform features such as recommendation systems, as well as the supply-and-demand dynamics of the broader web.

Dr. Hosseinmardi and her team propose a novel method called "counterfactual bots," which enables them to disentangle the role of the user from platform features on the consumption of highly partisan content. By comparing bots that replicate real users' consumption patterns with counterfactual bots that follow rule-based trajectories, they show that, on average, relying exclusively on the recommender results in less partisan consumption, with the effect being most pronounced for heavy partisan consumers.

--

12 p.m. - Lunch

12:30 p.m. - Presentation 

(Audience Q & A and networking to follow)