UX Research Projects
Agentic AI Role-Play for Insurance Sales Training
Applying learning science, behavioral evaluation, and human-AI interaction principles to professional skill development
@Samsung Life Insurance, AI/PI Center
2025–Present
Role: AI Product & Learning Science
Focus: Learning Science · GenAI Evaluation · Human-AI Interaction · UX Research · AI Product Strategy
This is an ongoing internal product. Product details, interfaces, proprietary methodologies, and research findings are intentionally omitted.
Overview
I work on an enterprise GenAI learning experience designed to help professionals develop interpersonal and communication skills through AI-powered role-play and coaching.
My work focuses on a central learning-science question: How do we distinguish a good AI conversation from a good learning experience?
The goal is not just to create engaging conversations, but to ensure that learners demonstrate the target skills and continue to grow.
My Role
I bridge learning science, research, AI, and product development to translate learning goals into experiences that can be practiced, evaluated, and improved.
Designing for Learning, Not Just Conversation
Conversational AI can create engaging and realistic interactions, but a good conversation is not necessarily evidence of learning.
My work focuses on designing experiences where learners practice meaningful skills, demonstrate observable behaviors, receive evidence-based feedback, and apply that feedback in continued practice.
My Conceptual Approach
Learning Objective → Observable Behavior → Evaluation → Feedback → Continued Practice
My conceptual approach connects learning goals to measurable learner growth by translating objectives into observable behaviors, evaluating performance, delivering actionable feedback, and using that feedback to guide continued practice.
Translating Learning Science Into Product Decisions
I use learning science as an input to product design rather than as an evaluation layer added after development. Several principles consistently shape how I think about practice, interaction, and feedback.
From Research to Product Decisions
I lead human-centered research throughout product development to understand learner needs, evaluate emerging experiences, and translate evidence into product and AI decisions.
Methods: Interviews · Surveys · Usability Testing · Think-Aloud Studies · Prototype Evaluation · Pilot Research · Behavioral Analysis
Process: Research → Analyze → Insight → Decide → Validate
Research findings inform decisions across learning design, AI behavior, evaluation, feedback, UX, and product requirements.
I work closely with domain experts, UX, product, engineering, AI teams, and business leadership to translate research and learning-science concepts into implementable requirements.
What This Work Reinforces
01. Engagement is not evidence of learning. A fluent or enjoyable AI interaction does not independently demonstrate skill development.
02. Learning constructs must become observable. Concepts such as empathy, communication, or problem solving become useful for evaluation only when translated into observable evidence.
03. Learning science has the greatest impact upstream. Learning science can shape product hypotheses, interaction design, evaluation, and research rather than simply evaluating a finished experience.
Related Research
My approach to this work builds on earlier research in learning science, educational technology, and human-AI interaction.
AI-Generated Pedagogical Agents
Experimental research examining AI-generated instructors, learning outcomes, cognitive load, and learner attention.
Evidence-Based AR Design Principles
Translating learning and cognitive principles into operational design criteria for educational technology.
*Confidentiality Note
This case study describes my individual areas of responsibility and general approach.
Samsung Life Insurance's unreleased product interfaces, proprietary evaluation frameworks, prompts, system architecture, internal research findings, and business information are intentionally excluded.