Research

Research Focus

My research examines how brands and CEOs communicate during societal crises—and how audiences respond emotionally, politically, and behaviorally. I explore corporate messaging in polarized environments using AI-powered methods, with the goal of understanding what drives engagement, support, or backlash.

In today’s sociopolitical climate, corporate silence can be seen as complicity, while speaking up may trigger division. My work investigates how message source (e.g., CEO vs. brand), style (e.g., emotional tone, symbolism, emoji use), and context (e.g., political polarization, social crises) shape audience reactions on digital platforms.

My dissertation, Voices in Action: AI-Powered Insights from Corporate Messaging During Societal Crises, bridges marketing, political psychology, and computational social science. It includes three studies that apply advanced AI and NLP techniques to analyze more than 380,000 real-world social media interactions:

Article 1 – Emojis & SDG Messaging

Draws on Media Richness Theory and uses fine-tuned deep learning transformer models to classify emoji types in brand communication about the UN Sustainable Development Goals (SDGs), revealing that concrete visuals increase engagement. → Accepted for EMAC 2025 and AMS 2025

Article 2 – CEO Voice During War

Applies machine learning-based emotion classification and framing analysis to show that CEO messages during the Russia–Ukraine war outperform brand messages in driving public engagement—especially when emotionally expressive (Executive Symbolism Theory). → Published in Journal of Public Policy & Marketing

Article 3 – DEI Reversal & Political Polarization

Uses Difference-in-Differences models, moral and emotional language classification, and AI agent-based labeling pipelines to study how partisan audiences react to corporate reversals on Diversity, Equity, and Inclusion (DEI), guided by Political Sectarianism Theory. → Targeting Journal of Marketing

AI Techniques & Analytics Methods I Use

Zero-shot and fine-tuned transformer models (e.g., BERT, RoBERTa)
Sentiment, emotion, stance, and moral language classifiers
Human-in-the-loop AI agents for classification and content generation
Multimodal analysis across text, emojis, and platform-specific engagement
Causal inference via Difference-in-Differences, Bayesian Structural Time Series (BSTS), and OLS regression
Experimental design using AI-generated stimuli for field or lab testing
Multi-agent chatbot framework to simulate persuasive brand-consumer interactions in controlled experiments
Role-specific agents (e.g., sentiment adjuster, content rewriter, ethical watchdog) for dynamic message generation

Purpose