Ali Kaazempur-Mofrad

PhD candidate Expected June 2027

UCLA Statistics & Data Science

amofrad@ucla.edu

Statistical methods for decision-making under uncertainty

My research connects statistics and mechanism design to understand how preferences and beliefs can be elicited and used for inference and decision-making, with applications to academic markets, fairness in resource allocation, and uncertainty reporting by large language models (LLMs).

Advised by Xiaowu Dai · SCALE Lab

Curriculum vitae PDF
Google Scholar GitHub LinkedIn
Ali Kaazempur-Mofrad
amofrad@ucla.edu

Research

Papers

Information, incentives, and decisions.

Fairness
& resource allocation

Published · 2026

A Data Envelopment Analysis Approach for Assessing Fairness in Resource Allocation: Application to Kidney Exchange Programs

Ali Kaazempur-Mofrad and Xiaowu Dai

The Annals of Applied Statistics, 20(1):385–407, 2026

Fairness in kidney exchange depends on several outcomes, including waiting time, kidney quality, and graft survival. We develop a conditional data envelopment analysis framework that assesses fairness across these outcomes together, with conformal prediction intervals that provide finite-sample, group-conditional coverage. Using U.S. transplant data, we find that disparities across ethnic groups differ by outcome.

Market design
& matching

Preprint · 2026

A Statistical Market-Design Framework for Academic Job Markets

Ali Kaazempur-Mofrad, Xiaowu Dai, and Xuming He

Departments allocate interviews with limited information about candidates’ willingness to accept an offer. We introduce a preference-signaling mechanism in which candidates share a standardized questionnaire with the departments they choose, and construct simultaneous confidence bounds on candidate ranks to guide interview selection. We characterize incentives for truthful participation, and simulations informed by U.S. statistics departments show higher placement rates and fewer unfilled positions.

Mechanism design
& LLM uncertainty

Submitted · 2026

Beyond a Single Answer: Proper Scoring of LLM Uncertainty

Ali Kaazempur-Mofrad, Xiaowu Dai, and Xuming He

A single answer can hide a language model’s uncertainty over alternatives. We design a mechanism based on a proper scoring rule that elicits truthful uncertainty reports, in which the model assigns probabilities to its candidate answers and to “I don’t know.” A single report then supports decisions at any tolerance for error, including a threshold rule that controls the error rate among returned answers without requiring calibrated reports.

Teaching & mentoring

From statistical foundations to practice

Teaching at UCLA

Teaching Assistant · 2023–present

I have served as a TA for undergraduate and graduate courses in statistical theory, regression, computing, Monte Carlo methods, and LLMs. For a graduate course on LLMs, I designed a sequence of Python labs that build from next-token prediction to reinforcement learning and uncertainty quantification.

Mentoring

I currently mentor one master’s student in Statistics and two undergraduates in Computer Science and Statistics and Data Science on research projects in LLM uncertainty and matching markets. I also mentored undergraduate teams at ASA DataFest at UCLA from 2023 to 2026.

Courses, materials & mentoring →

Academic activity

Talks, posters & service

Full CV

Presentations

  1. Upcoming

    IMS International Conference on Statistics and Data Science (ICSDS)

    Talk Beyond a Single Answer: Proper Scoring of LLM Uncertainty

  2. Upcoming

    Conference on Statistical Foundations of AI

    Poster Beyond a Single Answer: Proper Scoring of LLM Uncertainty

  3. Shapiro Statistics Student Seminar & Social, UCLA

    Talk Research overview: fairness in resource allocation, statistical market design, and proper scoring of LLM uncertainty

  4. StatsFest, UCLA

    Poster Beyond a Single Answer: Proper Scoring of LLM Uncertainty

  5. American Statistical Association

    Talk Research presentation to leadership · A Statistical Market-Design Framework for Academic Job Markets

  6. Statistics Student Seminar & Social, UCLA

    Talk A Statistical Market-Design Framework for Academic Job Markets