List All Projects List Graduate Projects List Senior Design Projects List REU Projects Other Undergraduate Projects
Project # 91
Teaming with AI: The Influence of Communication Style in Human-AI Collaboration
| Team Lead: | Tazim Ahmed |
| Other Team Members: | Eshwara Prasad Sridhar | Ansal Mohammed Abdul Khalam Azad | Oyin Ewedemi | Daksh Kuttan |
| Faculty Advisor(s): | Dr. Yuan Zhou |
| Department: | IMSE |
| Project Type: | REU |
| Title: | Teaming with AI: The Influence of Communication Style in Human-AI Collaboration |
| Abstract: | Trust in AI affects whether people follow recommendations safely, but trust changes with experience and outside information. We built an agent-based simulation to model how human trust in a shopping assistant chatbot evolves across repeated interactions under different AI configurations and “news shocks.” Eighteen chatbot variants were created by combining communication style (task-oriented vs. social-oriented), temperature (0.2, 0.5, 0.8), and top-p (0.3, 0.7, 1.0). To ground the model in real data, three users rated trust after completing shopping tasks for each configuration; these ratings were converted to 0–100 trust scores and used to initialize simulated users with cognitive trust (competence) and affective trust (comfort). Over 100 interaction steps, agents updated trust based on whether the AI’s answer was correct, the chatbot’s style, and individual learning speed. At random times, positive or negative news shocks caused sudden trust jumps for a portion of the population, mimicking media effects. Results show that social style with low randomness (low temperature and well-chosen top-p) produced the highest and most stable trust and recovered faster after negative shocks, while highly random task-style models showed lower trust and weaker recovery. This framework helps designers test AI settings and resilience to external events before costly user studies. |
| Poster: | View Poster #91 (PDF) |
| Judges Scoring: | Submit Evaluation (judges only) |