t-test data Box plot for data Procedure design diagram
Project silhouette

AI Telehealth Persona Research

Fall 2025 · UX Research for PM

Type

UX Research · Product

Role

Group Leader

Team

Team of 4

Method

Controlled Experiment

Overview

Conducted a controlled HCI experiment measuring how an AI chatbot persona's facial expression and tone affect user perception in a telehealth setting. We compared a smile + friendly-tone persona against a neutral-expression + neutral-tone persona, collecting quantitative satisfaction and trust ratings from participants. Statistical analysis with t-tests revealed results that directly challenged our initial hypothesis — and conventional UX wisdom.

01 · Research Question

Does a warmer, friendlier AI chatbot persona — one that smiles and uses a friendly tone — produce higher user satisfaction and trust in a healthcare context? Our hypothesis going in was yes: more human-like warmth would feel more approachable and build rapport.

02 · Key Findings

Neutral AI outperformed the friendly AI

Contrary to our hypothesis, participants rated the neutral-expression, neutral-tone persona significantly higher on satisfaction metrics. The difference was statistically significant (p < 0.05), making this a reliable finding rather than noise.

In healthcare, neutrality reads as competence

Qualitative notes and rating patterns suggest users associated a calm, neutral AI with professionalism and trustworthiness in a medical context. The smiling persona felt out of place — warmth backfired when the stakes felt clinical.

Next Project

SettledIn