Kobi is a DPhil candidate (Social Data Science) and Clarendon Scholar at the OII. Kobi’s research focuses on evaluating the capabilities of AI systems to influence the political public sphere, with a particular emphasis on personalized large language models (LLMs). His research has used experimental methods to quantify the persuasiveness of LLMs engaged in political microtargeting and partisan role-play, and has integrated natural language processing and network science to map partisan discourse during elections. Kobi is also a part-time Research Assistant in the Online Safety Team at the Alan Turing Institute working on the development of a research program in applied AI model safety.
He holds an MSc (Politics and Communication) from the London School of Economics and Political Science, where he was awarded the departmental prize for Best Overall MSc Performance, and an MSc (Social Data Science) from the OII. He also holds a BA (Political Science) from the University of Wisconsin – Madison.
Artificial intelligence, machine learning, NLP, large language models, political communication, persuasion, human-subjects experiments.