BerkeleyHaas

Peiyao Li
Ph.D. candidate in Marketing, UC Berkeley Haas School of Business

I am on the 2026-27 Marketing academic job market.
​
My research interest centers around applied machine learning and AI in marketing and broader business applications. My current projects focus on understanding and tracking brand perception, understanding human behavioral patterns using survey experiments and generative models, and measuring creativity and innovation.

Here is my google scholar profile and CV.


Job Market Paper:
No Spoilers: A Contamination-Free LLM for Brand Perception Analysis
with Zsolt Katona
Abstract: Brands are focal objects of marketing, and how consumers perceive them shapes positioning, consumer choice, and firm performance. These perceptions are often not directly observed but reflected in the large volume of unstructured text consumers produce about brands, making language models a scalable tool for measuring them. Tracking such measures over time, however, raises two challenges. First, model parameters may encode information from after the measurement date, producing lookahead bias. Second, changes across model versions undermine comparability. We address both with Data-Restricted Incremental Pretraining and Projection (DRIP^2), a workflow that builds a lightweight temporally bounded sentence encoder from pre-sample data, constructs quarterly brand representation vectors, removes platform-wide language shifts, and projects the vectors onto researcher-defined dimensions. Applied to more than 60 million tweets from 2016--2022, DRIP^2 generates interpretable maps of brands, sports teams, and actors. The measures respond to major real-world events, predict future profit-margin growth, and reveal short-lived shifts in the intended direction following marketing campaigns and movie releases.





​

Publications:

Determining the validity of large language models for automated perceptual analysis 
 with Noah Castelo, Zsolt Katona, and Miklos Sarvary
Marketing Science, 2024 
Keywords: Artificial Intelligence, perceptual maps, large language model, natural language processing, market research​.

AI and perception biases in investments: An experimental study
with Anastassia Fedyk, Ali Kakhbod, and Ulrike Malmendier
Journal of Financial Economics (JFE), forthcoming
Best Paper Award, DC Finance Week
Keywords: Large language models, Behavioral biases, Experimental economics, Investment preferences, Financial surveys, Generative AI.

​Are Patents with Female Inventors Under-Cited? Evidence from Text Estimation
 with Yael Hochberg, Ali Kakhbod, and Kunal Sachdeva
Journal of Financial Economics (JFE), 2026
Keywords: Innovation, Gender, Patent, Machine Learning, Big Data, Inference.

​NoLBERT: A No Lookahead (back) Foundational Language Model
with Ali Kakhbod
NeurIPS 2025
Keywords: Foundational model, language model, lookahead bias.
​


​Working papers:

​Measuring Creative Destruction
with Ali Kakhbod, Leonid Kogan, and Dimitris Papanikolaou
Review of Financial Studies (RFS), Revise and Resubmit 
First place, Crowell Memorial Prize, Panagora Asset Management
Keywords: Innovation,  Displacement, Patents, Machine learning.

Personality, Selective Recall, and Investor Heterogeneity: Experimental Evidence
with Anastassia Fedyk, Ali Kakhbod, and Ulrike Malmendier
Keywords: Economic expectations, Selective Recall, Experimental economics, Interpretable latent-variable discovery, Machine learning, AI clones.

How AI outperforms humans at creative idea generation
​with Noah Castelo, Zsolt Katona, and Miklos Sarvary
​​​Keywords: Artificial intelligence, Creativity, Product design, Advertising, Marketing research.