Moshe Mash, Ph.D.

My research centers on decision-making by self-interested and bounded agents, human and artificial, studied through game theory, mechanism design, multi-agent systems, and human-subjects experiments. I am developing a new direction that extends this to LLM-based agents: learning what an agent should communicate to a human, and how, from human preference, in mixed human-agent settings such as autonomous and human-driven vehicles.

A second, applied strand of my research uses large language models and modern NLP for text analysis in specialized domains: authorship attribution, automatic citation generation, and semantic cross-referencing of Hebrew and rabbinic texts (with DICTA), and computational analysis of legal and patent text, including textual studies of disparate outcomes in the patent system. Both involve fine-tuning LLMs for prediction, clustering, and semantic analysis of domain-specific corpora.