Haleh Yazdi

AI Safety & UX Researcher

Haleh Yazdi,
Ph.D.

Bridging Behavioral Science & AI Alignment

I'm a researcher with a Ph.D. in Experimental Psychology and 12+ years of experience across Google, Meta, and academia — studying how humans develop moral reasoning, form biases, and interact with AI. As a Staff-level UX Researcher at Insulet (previously at Meta Reality Labs and Google), I translate psychology research into concrete safety decisions, from driving compliance with trust and safety standards to reducing user harm in immersive VR environments. My work translates developmental & social psychology into actionable safety policies, harm taxonomies, and evaluation criteria that help AI systems behave in alignment with human values.

View Research Get in Touch

Background

About Me

My path to AI safety research began with a fundamental question: how do humans develop moral minds? Through years of cross-cultural fieldwork in Canada, India, Iran, and Korea, I built a granular understanding of how cognition, social norms, and reputation shape ethical behavior from childhood onward.

That foundation turned out to be surprisingly well-suited to some of AI's hardest problems. How should models reason about harm? What does it mean for an AI to be trustworthy? How do real users — including children — actually interact with and moralize AI agents? These are questions I've been researching, in various forms, for over a decade.

At Google and Meta I worked at the frontier of human-AI interaction — designing safety features, developing novel evaluation metrics, and partnering with ML and policy teams to ensure products protect users at scale. I bring both empirical rigor and product fluency to every research question I take on.

  • 🎓
    Ph.D. in Experimental Psychology University of California, San Diego · 2022
  • 🎓
    M.Sc. in Experimental Psychology University of Oxford, UK · 2013
  • 📊
    Micro MBA UC San Diego Rady School of Management · 2016
12+
Years of Research Experience
$340K
in Fellowships & Grants, incl. DoD NDSEG Award & Fulbright Finalist
5
Published Manuscripts in High-Profile Journals
32
Conference Presentations across 6+ Countries
🏆
GradSlam Winner — Best Thesis, All UCSD Graduate Students

Notable Projects

Research Highlights

🤖

Children's Moral Beliefs Towards AI Systems

Investigated how children and adolescents perceive home AI agents (Alexa, Siri). Found that reputation concerns and empathic interactions lead children to attribute moral standing to AI — directly informing youth-specific safety policy design.

AI Safety · Child Development
Google Scholar →
🌍

Morality & Prejudice Formation Across Cultures

Field experiments in Canada, India, Iran, and Korea examining how laws, religion, and social norms shape children's moral development. Yielded a novel two-stage cross-cultural methodology directly applicable to global harm taxonomy design.

Cross-Cultural · Harm Taxonomy
Dev. Psychology 2025 → Child Dev. 2020 →
👁

Reputation, Transparency & Ethical Behavior

Using experiments and economic games, showed that non-anonymous settings reliably increase ethical behavior in both children and adults — with direct implications for accountability and reward-based safety mechanisms in model training.

RLHF Relevance · Accountability
J. Exp. Child Psych. 2020 →

Career

Experience

Staff UX Researcher
Insulet
Oct 2025 – Present

Lead end-to-end research on AI-driven insulin delivery systems, supporting users of all ages living with diabetes.

  • Survey research & user interviews drove 3% increase in conversion to Omnipod from competitor pumps in 3 months.
  • Partner with clinical and data science teams to mitigate ethical and emotional risks in algorithmic insulin delivery.
  • Apply developmental psychology frameworks to design caregiver-centered solutions.
Senior UX Researcher
Meta Reality Labs
Nov 2024 – Oct 2025

Quest Spatial Awareness team — applied developmental and social psychology to adolescent safety in immersive tech.

  • Drove cross-functional development of in-headset safety features: +6 min avg daily use, −15% user collisions.
  • Reduced time-to-fun by 1 min 15 sec in the safety setup flow without regressing on safety metrics.
  • Partnered with Legal and Health & Safety teams for trust & safety compliance.
Senior UX Researcher
Google
Nov 2024 – Oct 2025

Designed and led user studies for Google Beam, Google's AI-first 3D videoconferencing product.

  • Developed & validated new comfort and premiumness metrics adopted by 3 Google product teams.
  • Identified and resolved 13 UX pain points, reducing user failures by 85% and improving task completion by ~21%.
Senior Psychology Researcher (PhD & Postdoctoral Scholar)
UC San Diego
Sept 2014 – Dec 2023
  • Led a team of 16 research assistants across 6 countries (Canada, India, Iran, Korea, Mexico, US).
  • Won GradSlam — best 3-minute thesis presentation among all UCSD graduate students.
  • 5 published manuscripts in high-profile journals; 32 conference presentations.
  • Secured $340K in fellowships & grants including DoD NDSEG award and Fulbright Finalist.
Product & Marketing Consultant
H. Yazdi Research & Consulting LLC
Jan 2019 – Feb 2024
  • Boeing: Global survey research (1,500 participants, 8 countries) — insights determined final icon designs in Boeing 737 MAX.
  • Netflix: Tailored streaming features for users 65+ → 67% feature adoption in 3 weeks, 20% viewership increase over 6 months.
  • Pearson & Digital Promise: Research blog posts with 3,000+ shares.

Scholarship

Select Publications & Presentations

Expertise

Skills & Tools

🛡 AI Safety & Model Policy

Model-aligned safety policies Harm taxonomy design Safety-critical UX metrics Psychological risk research Behavioral pattern analysis Human-AI interaction

🔬 Research Methods

Experimental design Mixed methods Cross-cultural studies Longitudinal research Psychometrics Survey design Behavioral measurement

💻 Quantitative Tools

R SPSS SQL Qualtrics

🧪 Applied UX Research

IRB / COPPA / HIPAA Usability studies User interviews Diary studies Ethnography Critical user journeys Rapid prototyping Jira · Confluence · Miro

Get in Touch

Contact

I'm always interested in connecting with researchers and teams working at the intersection of behavioral science and AI. Feel free to reach out.

LinkedIn Google Scholar