This paper presents a Retrieval-Augmented Generation (RAG) pipeline designed to generate high-quality, emotionally intelligent responses to student reviews. Online reviews function as a critical digital interface through which students express expectations, frustrations, and satisfaction. Large Language Models (LLMs), while widely adopted for automated communication and support, often produce generic replies that lack emotional resonance. This study investigates whether emotion-aware and intent-aware prompting (e.g., emotions such as anger, anxiety, or satisfaction, and intents such as complaint, inquiry, suggestion, or praise) improves the quality, empathy, accuracy, and alignment of AI-generated responses in higher-education communication. Using a controlled experimental design, three prompting strategies — Simple (baseline), Emotion-aware, and Emotion+Intent — were tested across three GPT-5 model variants (GPT-5, GPT-5-mini, and GPT-5-nano), yielding nine experimental conditions. One hundred real student Google Reviews of University Canada West served as input data, evenly split between low-rating (1–2 stars) and high-rating (4–5 stars) reviews. All generated responses were evaluated using a deterministic rubric-based scorer across five dimensions: relevance, completeness, emotional responsiveness, clarity, and helpfulness (scored 20–100). Results show that prompting strategy is a stronger predictor of response quality than model size. Emotion-aware prompting (mean: 66.15) substantially outperformed the simple baseline (mean: 56.44), while Emotion+Intent prompting (mean: 65.02) produced the most contextually stable responses. Importantly, structured prompting enabled smaller, lower-cost models to achieve competitive performance, with GPT-5-mini emerging as the optimal configuration for institutional deployment. These findings align with the Computers Are Social Actors (CASA) framework and carry practical implications for universities seeking scalable, cost-effective AI tools that maintain empathy and accuracy in student-facing communication.

Authors:
Vahideh Baradaran Rafiee, University Canada West, Canada
Amirhossein Zaji, University Canada West, Canada
Amirhossein Kaviani, University Canada West, Canada

About the Presenter(s)
Dr. Vahideh Baradaran-Rafiee is an associate professor at the Department of Marketing, Strategy, and Entrepreneurship at University Canada West.

See this presentation on the full schedule – Sunday Schedule

Bookmark