TigerGPT: A New AI Chatbot for Adaptive Campus Climate Surveys

TL;DR

Traditional campus climate surveys often suffer from low participation and shallow data. The paper introduces TigerGPT, a GenAI chatbot that uses role-based initialization, real-time adaptive dialogue, empathetic cues, bold prompts, and user-driven topic selection to collect richer feedback. In a pilot with undergraduates, TigerGPT achieved high satisfaction, strong perceived understanding, and a substantial portion of participants preferring it over conventional surveys, demonstrating improved engagement and data quality. The findings suggest TigerGPT as a scalable approach to obtaining deeper campus insights, with future work focusing on broader topic coverage, system stability, and deployment across diverse campuses.

Abstract

Campus climate surveys play a pivotal role in capturing how students, faculty, and staff experience university life, yet traditional methods frequently suffer from low participation and minimal follow-up. We present TigerGPT, a new AI chatbot that generates adaptive, context-aware dialogues enriched with visual elements. Through real-time follow-up prompts, empathetic messaging, and flexible topic selection, TigerGPT elicits more in-depth feedback compared to traditional static survey forms. Based on established principles of conversational design, the chatbot employs empathetic cues, bolded questions, and user-driven topic selection. It retains some role-based efficiency (e.g., collecting user role through quick clicks) but goes beyond static scripts by employing GenAI adaptiveness. In a pilot study with undergraduate students, we collected both quantitative metrics (e.g., satisfaction ratings) and qualitative insights (e.g., written comments). Most participants described TigerGPT as engaging and user-friendly; about half preferred it over conventional surveys, attributing this preference to its personalized conversation flow and supportive tone. The findings indicate that an AI survey chatbot is promising in gaining deeper insight into campus climate.

Figures (9)

• Figure 1: Overview of TigerGPT's initial role selection process. This flowchart illustrates TigerGPT’s selection logic for campus climate surveys at the University of Missouri. Each user’s role (student, faculty, staff, or alumni) plus details such as a student’s type, a faculty member’s rank, or a staff member’s working area guides them to the most suitable survey template and topics.
• Figure 2: A screen shot of TigerGPT’s initial interface. A welcome screen prompting users to select their role at MU, as student, faculty, staff or alumni, to tailor subsequent interactions accordingly.
• Figure 3: TigerGPT offers users a variety of survey topics and an option of random selection.
• Figure 4: Sample Conversation with TigerGPT. This screenshot demonstrates five features: (A) the chatbot’s awareness of user information, (B) a prompt for the user’s preferred name, (C) a bold question supported by examples, (D) a follow-up request for more details, and (E) a “Switch Topic” button allowing the user to change the conversation’s direction.
• Figure 5: Sample Conversation for a Sensitive Topic. This example shows TigerGPT’s empathetic approach to user experiences of unfair treatment. In (A), the chatbot uses active listening and supportive language (with emojis) to validate the user’s feelings. In (B), it invites the user to share more details only at their comfort level, illustrating a respectful framework for discussing sensitive issues.