Announcing CANSSI Ontario’s 2025 Cohort of Mdoc Trainees: Yahang Qi

Yahang Qi is a PhD student in the Department of Statistical Sciences (DoSS), working under the joint supervision of Professors Dehan Kong (DoSS) and Zhijing Jin (Department of Computer Science, U of T).

Together, they will work on the Mdoc project “Enhancing Causal Inference with Large Language Models.” Their research centers on enhancing causal inference using large language models (LLMs), with the goal of extending statistical methodologies to accommodate natural language data and support reliable, interpretable AI systems.

Yahang’s work addresses a fundamental challenge in modern data science: how to apply causal inference—traditionally designed for structured numerical data—to the unstructured, text-rich environments in which LLMs operate. His research develops theoretical guarantees for causal inference methods that incorporate natural language, enabling rigorous reasoning and decision-making.

The theoretical component of his research focuses on adapting structural causal models and classical estimators to handle the complexity and ambiguity of text data. He investigates how LLMs infer latent variables and causal relationships from language, and how these processes can be formalized to ensure fairness, transparency, and consistency.

On the applied side, Yahang designs pipelines that use LLMs to extract causal variables from real-world text and estimate treatment effects under realistic conditions like confounding and missing data. These methods aim to make causal analysis scalable and accessible, empowering researchers and practitioners to draw reliable conclusions from complex datasets.

Yahang Qi’s research integrates statistical theory, algorithmic development, and empirical validation to advance causal inference using large language models (LLMs). By extending traditional causal methods to incorporate natural language data, his work provides both theoretical foundations and practical tools for reliable and interpretable AI decision-making. Through interdisciplinary collaboration with experts, Yahang ensures that his methodologies are grounded in real-world relevance, contributing to the development of ethically responsible and socially aligned AI systems.

CANSSI Ontario launched the Multidisciplinary Doctoral (Mdoc) Training Program in the DoSS in 2019. Students admitted to the Mdoc Program can be domestic or international and will receive the same funding package as the other doctoral students entering the DoSS including Teaching Assistant assignments. Learn more.