Upasana Dutta

Amir Ghasemian, Homa Hosseinmardi, Upasana Dutta et al.

Recommender systems have grown from content-organization tools into sophisticated systems that shape daily behavior. By controlling what we see, they shape what we perceive, raising concerns about filter bubbles, radicalization, polarization, and social inequality. Large language models (LLMs) enable more powerful personalization, intensifying these dynamics. Yet most recommenders are tuned for engagement or limited accuracy metrics, with little attention to broader social implications, e.g. how personalization reshapes exposure in socially consequential domains. We investigate whether LLM-assisted reranking, while improving personalization, inadvertently amplifies exposure to ideologically extreme or conspiratorial political content, a risk theorized but not empirically characterized in news recommendation. Using real news-consumption histories, we rerank YouTube's sidebar candidates through zero-shot, instruction-based prompting. We compare a baseline prompt with a constrained variant that preserves topical relevance and broadens ideological exposure while reducing conspiratorial or extreme content. Without constraints, reranking strengthened personalization but increased exposure to conspiratorial and extremist material for users whose histories contained such content. Lightweight prompt-level regularization reduced promotion of extreme content and increased ideological diversity, with modest relevance loss. Synthetic experiments suggest that LLMs rerank via statistical regularities in language rather than semantic understanding of ideology, clarifying why naive prompts amplify these patterns and why regularization can reshape them. Together, our results highlight the power of LLMs to operationalize contextual nuance in high-stakes recommendation, and the need to evaluate LLM-assisted personalization beyond accuracy and treat prompt design as a value-laden rather than neutral default.

Upasana Dutta, Alexander Ray, Aaron Clauset

How does the shape of a network change as its size increases? Although random graph models provide some expectations for such "scaling behaviors" in the structure of networks, relatively little is known about how empirical network structure scales with network size or how well random graphs explain those empirical patterns. Using a large, structurally diverse corpus of networks from four scientific domains, we first characterize the empirical scaling laws of real-world networks, considering how mean degree, transitivity, mean geodesic distance, and degree assortativity vary with network size. We show that networks from all four scientific domains exhibit a consistent set of scaling laws on these measures of network structure, but with differing scaling rates. We then assess the extent to which these empirical scaling laws are explained by three random graph models with different structural assumptions, showing that configuration model random graphs are a remarkably good model of network scaling behavior, although null models with modular structure are slightly better. These findings identify a new set of common patterns in the network structure of complex systems, provide new validation targets for models of network structure, and shed new light on the role of randomness in shaping the large-scale structure of networks.

Upasana Dutta, Rhett Hanscom, Jason Shuo Zhang et al.

Social media platforms have been exploited to conduct election interference in recent years. In particular, the Russian-backed Internet Research Agency (IRA) has been identified as a key source of misinformation spread on Twitter prior to the 2016 U.S. presidential election. The goal of this research is to understand whether general Twitter users changed their behavior in the year following first contact from an IRA account. We compare the before and after behavior of contacted users to determine whether there were differences in their mean tweet count, the sentiment of their tweets, and the frequency and sentiment of tweets mentioning @realDonaldTrump or @HillaryClinton. Our results indicate that users overall exhibited statistically significant changes in behavior across most of these metrics, and that those users that engaged with the IRA generally showed greater changes in behavior.