Evaluated an AI Chatbot to Ensure Accurate Responses to Faith-Based Questions
The Challenge
The organization wanted to understand whether an AI chatbot could provide a helpful, approachable way for members and prospective visitors to ask questions about its beliefs and teachings.
The Process
I designed and moderated usability sessions in which participants asked the chatbot real questions about doctrine, scripture, and everyday faith.
Rather than evaluating whether participants simply liked the chatbot, I focused on where it produced misleading, incomplete, or overly confident responses. I documented patterns in the failures and identified the types of questions that consistently challenged the model.
The Impact
The research demonstrated that the chatbot performed well when answering straightforward factual questions but struggled to consistently provide responses that aligned with the organization's doctrine and mission.
One participant encountered a response that directly conflicted with the organization's beliefs. Although it was a single interaction, it quickly eroded that participant's trust in the chatbot and demonstrated how easily confidence in AI can be lost when responses extend beyond appropriate boundaries.
Based on these findings, I recommended implementing guardrails around the AI's responses, including limiting topics, providing source references, and establishing clear escalation paths when questions exceeded the chatbot's capabilities.
The findings helped the organization better understand where AI could provide value and where human expertise remained essential.
My Value
I looked beyond whether the chatbot worked and focused on where it could fail. By identifying patterns in trust-eroding responses and recommending appropriate guardrails, I helped the organization make informed decisions about the responsible use of AI.
Reflection
This project reinforced that successful AI experiences are not just about making models smarter. They are about designing experiences with appropriate guardrails that inspire user trust.
It challenged me to think beyond traditional usability and consider trust, transparency, and responsible AI behavior as equally important parts of the user experience.