Research

I develop explainable AI methods to predict consumer behavior in digital environments. My work combines large-scale social media analytics (analyzing 13+ million posts) with econometric rigor to understand how marketing communication drives engagement, emotion, and action.


Research Streams

1. AI-Powered Marketing Communication Analytics

I create hybrid AI classification systems that balance predictive power with interpretability—critical for marketing applications where managers need actionable insights, not black boxes. My methodological contributions include:

Domain-specialized transformer models combining large language model embeddings with rule-based validation layers Multimodal analysis of text, emojis, and visual content across digital platforms Scalable NLP pipelines for sentiment, emotion, stance, and moral language detection

Key findings: Symbol emojis (🏳️‍🌈 for inclusivity) increase SDG message engagement by 281% while icon emojis show no effect. CEO communication generates stronger public engagement than brand accounts during geopolitical crises. When brands maintain their DEI commitments while supporters are vocal on social media, they gain more consumer engagement and purchase intentions than when they abandon DEI while critics are vocal—standing firm with supporters beats retreating from critics.

2. Consumer Responses to AI Technologies

Collaborating with researchers at iaelyon and Emlyon Business School, I analyze shifting human-AI relationships using change-point detection across millions of social media conversations. This line of research examines:

How consumers perceive and respond to AI agents in marketing contexts The role of humor in shaping attitudes toward AI automation Evolving assemblages between humans and artificial intelligence in digital ecosystems


Methods & Capabilities

Machine Learning: Supervised learning, ensemble methods, neural networks, predictive behavioral modeling

Explainable AI: Interpretable hybrid systems, mechanistic interpretability, XAI techniques for marketing applications

NLP & Text Analytics: GPT, BERT, RoBERTa, multimodal transformers, sentiment/stance detection, retrieval-augmented generation (RAG)

Causal Inference: Panel data analysis, difference-in-differences, negative binomial regression, text-as-data econometrics

Tools: Python, R, OpenAI API, Hugging Face, PyTorch, TensorFlow, Scikit-learn

Data Scale: 13+ million social media posts 500,000+ YouTube videos Controlled experiments with 250+ participants

Impact & Collaboration

My research bridges computational social science and marketing strategy, offering data-driven insights for brands navigating polarized digital environments. I’m open to collaborations on projects involving AI, predictive consumer analytics, or large-scale social media analysis.