I study how culture and human values are represented in artificial intelligence, how these systems shape human judgment and behavior, and how we can evaluate them across cultural and linguistic contexts. My work combines behavioral experiments, large-scale observational data, computational methods, and cross-cultural research.
Research on how AI systems represent culture, values, and human judgment.
Evaluates whether frontier LLMs capture not only where cultures are today, but how they have changed over time. The study compares more than two decades of World Values Survey data across 40 countries with cultural trajectories produced by four frontier LLMs.
Examines how LLMs may contribute to cultural homogenization through their representations, communication patterns, and effects on human behavior, and develops policy recommendations for preserving cultural diversity in AI systems.
Introduces a multilingual benchmark containing 3,000 tweets in English, Persian, Portuguese, and Italian, annotated with hate speech labels, moral categories, and text-span rationales. The benchmark evaluates whether LLM outputs align with human moral reasoning.
Research on how trustworthiness can be measured and demonstrated in AI systems.
Examines the gap between responsible AI processes and independently verifiable trustworthy outcomes. The paper proposes independent, outcome-oriented certification as a mechanism for making trustworthiness measurable, comparable, and externally recognizable.
Research on culture, cognition, moral judgment, and social evaluation across diverse populations.
Combining large-scale computational analysis with three behavioral experiments, including an intervention, we show that misperceptions of how political opponents view our morality shape trust and perceived threat and can be corrected.
A large-scale Registered Report measuring semantic priming across 19 languages and more than 25,000 participants, examining both the robustness of the effect and variation across languages.
A large-scale culturally diverse study examining how situational factors influence moral judgments across Eastern, Southern, and Western countries.
A large-scale intervention contest testing strategies for reducing attractiveness-based discrimination in social judgment.
For papers, data, code, presentations, and project-specific materials, see the individual project links or get in touch.