Advances in AI for web integrity, equity, and well-being.

My research develops data mining, AI, and applied machine learning methods to combat malicious actors (sockpuppets, ban evaders, etc.) and dangerous content (misinformation, hate, etc.) on web platforms. My vision is to create a trustworthy online ecosystem for everyone and the next generation of socially-aware methods that promote health, equity, and integrity of users, communities, and platforms online. Broadly, in my research, I create novel graph, content (NLP, multimodality), and adversarial machine learning methods leveraging terabytes of data to detect, predict, and mitigate online threats. My interdisciplinary research innovates socio-technical solutions that I achieve by amalgamating computer science with social science theories. My research seeks to start a paradigm shift from the current slow and reactive approach against online harms to agile, proactive, and whole-of-society solutions. In this article, I shall describe my research efforts along four thrusts to achieve my goals: (1) Detection of harmful content and malicious actors across platforms, languages, and modalities; (2) Robust detection models against adversarial actors by predicting future malicious activities; (3) Attribution of the impact of harmful content in online and real world; and (4) Mitigation techniques to counter misinformation by professionals and non-expert crowds. Together, these thrusts give a set of holistic solutions to combat cyberharms. I am also passionate about putting my research into practice-my lab's models have been deployed on Flipkart, influenced Twitter's Birdwatch, and now being deployed on Wikipedia.

Examining the impact of sharing COVID-19 misinformation online on mental health.

Misinformation about the COVID-19 pandemic proliferated widely on social media platforms during the course of the health crisis. Experts have speculated that consuming misinformation online can potentially worsen the mental health of individuals, by causing heightened anxiety, stress, and even suicidal ideation. The present study aims to quantify the causal relationship between sharing misinformation, a strong indicator of consuming misinformation, and experiencing exacerbated anxiety. We conduct a large-scale observational study spanning over 80 million Twitter posts made by 76,985 Twitter users during an 18.5 month period. The results from this study demonstrate that users who shared COVID-19 misinformation experienced approximately two times additional increase in anxiety when compared to similar users who did not share misinformation. Socio-demographic analysis reveals that women, racial minorities, and individuals with lower levels of education in the United States experienced a disproportionately higher increase in anxiety when compared to the other users. These findings shed light on the mental health costs of consuming online misinformation. The work bears practical implications for social media platforms in curbing the adverse psychological impacts of misinformation, while also upholding the ethos of an online public sphere.

Predicting Dynamic Embedding Trajectory in Temporal Interaction Networks.

Modeling sequential interactions between users and items/products is crucial in domains such as e-commerce, social networking, and education. Representation learning presents an attractive opportunity to model the dynamic evolution of users and items, where each user/item can be embedded in a Euclidean space and its evolution can be modeled by an embedding trajectory in this space. However, existing dynamic embedding methods generate embeddings only when users take actions and do not explicitly model the future trajectory of the user/item in the embedding space. Here we propose JODIE, a coupled recurrent neural network model that learns the embedding trajectories of users and items. JODIE employs two recurrent neural networks to update the embedding of a user and an item at every interaction. Crucially, JODIE also models the future embedding trajectory of a user/item. To this end, it introduces a novel projection operator that learns to estimate the embedding of the user at any time in the future. These estimated embeddings are then used to predict future user-item interactions. To make the method scalable, we develop a t-Batch algorithm that creates time-consistent batches and leads to 9× faster training. We conduct six experiments to validate JODIE on two prediction tasks- future interaction prediction and state change prediction-using four real-world datasets. We show that JODIE outperforms six state-of-the-art algorithms in these tasks by at least 20% in predicting future interactions and 12% in state change prediction.

Predicting human behavior: The next frontiers.

Machine learning has provided researchers with new tools for understanding human behavior. In this article, we briefly describe some successes in predicting behaviors and describe the challenges over the next few years.