CHONGYANG.IO

My Works. Skill Set Interest About Me
My Photo
CV / RESUME
My Story
My name is Chongyang Wang, usually folks call me Gary! I'm a PhD student in Computer Science at Northeastern University and a research assistant in the Relational Agents Group with Prof. Timothy W. Bickmore. I build and study AI agents that teach: embodied conversational agents, virtual reality, and LLM-driven tutoring, mostly for patient and health education and increasingly for education in general.
AI has driven the cost of telling to zero. It has not lowered the cost of learning.
That one judgment runs through all my work. Today's models can tell you almost anything, often better than most people, yet none of them can take a person from not knowing to actually knowing, start to finish, on its own: the whole job of a competent tutor. I split that gap into four layers, one for each part of a tutor's job: the content, the learner, the tutor-learner relationship, and institutions. You can watch the whole argument in the video below, then walk through it node by node in the Problem Map.
The full story, 28 minWhy AI Still Can't Teach on Its Own
Every node in the Problem Map below also has a ▶ Watch this part button that jumps straight to its moment in the video.
My PhD work tackles the first three layers with patients: Dr. Clara, an LLM-driven agent that walks patients through their education materials in VR; a pedagogical layer that takes teaching decisions out of the language model; teach-back, where the agent checks whether anything actually stuck; and life-sized VR health coaches, whose rapport and physical distance shape how people respond. Outside campus I build learning tools with K-12 teachers (AI PBL Agent, Keyi) and teach middle schoolers to make games with AI (First Principle Agent). That is where I learned the fourth layer.
Before Northeastern I was at Columbia University. In Prof. Joey Lee's TC Game Research Lab I learned that a well-designed game is a well-designed learning experience, and in Prof. Steven Feiner's CGUI Lab I worked on XR and 3D interfaces for NSF-funded human-robot collaboration in future manufacturing (ACM SUI 2023).
Teaching and making have always gone together for me. As an instructional designer I authored official Unity Learn courses and ran game-development workshops, and I still design and build games, these days a card tactics roguelike called For Boaria!. That mix is why I care less about what an AI can say and more about what a learner walks away with.
Education
Ph.D. in Computer Science (in progress), Northeastern University
M.A. in Instructional Technology, with program track - Design and Development of Digital Games, Columbia University
Cert. in Game Design, University of Wisconsin - Madison
B.S. in Computer Science, University of Wisconsin - Madison
Active Roles
Research Assistant at Relational Agents Group, Northeastern University, Supervisor: Timothy W. Bickmore
The Problem Map
AI tells you things very well, often better than most people. What it cannot yet do is teach: do the whole job of a competent tutor and carry one person from not knowing to knowing, start to finish. This map is how I take that gap apart. It is also the architecture of my research: understand the material, understand the learner, deliver in real time, plus a fourth layer I only saw by building things outside the lab. Click any node.
The root questionWhy can't AI teach a person on its own, from start to finish?Teaching on its own = the whole job of a competent tutor
Why this is worth the long haul

Suppose AI becomes a hundred times stronger. What must a person still do for themselves? Make the decisions about their own life: whether to take this drug, have this operation, believe this piece of news. To decide, you have to understand at least a little, so human cognition cannot be fully outsourced, and people must keep the ability to learn.

The most effective teaching we know, one-on-one and personalized, is also the scarcest, so it goes to those who can pay or know the right people. An AI that can truly teach could fill that gap, but only if it gets past telling. For most people the bottleneck was never a shortage of material; it is motivation, trust, and whether they can keep going. Flipped over, the map becomes four design goals:

EfficientMore of what is taught actually stays, per minute of learning.
Low painThe effort of learning can be reduced, never removed.
Stable over timeNot the novelty of one session, but months and years of company.
Through a narrow channelStill leaving a trace through text, voice and a screen.
skill set
An Interdisciplinary Skill set and tool mastery:
Discover My Work from 5 Aspects
I've arranged things this way to show the evolution my work and interests/focus.
The sequence (left > right; top > bottom, not strictly though) represents the proximity of each item to my current focus and interests. Dive into my:
EXPERIENCE: roles
PUBLICATION: recognized or published papers and products
PROJECT: formal, systematic series of tasks with professional guidance
THOUGHT: thoughts/prototypes with the potential (don't have time or not polished yet) to evolve into projects, or simply fun ideas
AWARD: some recognitions.


Loading Thumbnails...(feel free to view and click tho)

PROJECT-under review
Ongoing

Dr. Clara: VR Patient Educator

an LLM-driven embodied agent that turns patient education pamphlets into guided, multimodal lessons in VR; NIH-funded · map 1.2, 1.3, 2.1, 3.2
PROJECT
Ongoing

Presence Is Not Pedagogy

a pedagogical layer between the LLM and the embodied agent: concept maps, 24 teaching patterns, 363 typed teaching actions · map 1.1, 1.3, 2.1, 3.2
PROJECT-under review

Did You Get That? Teach-Back

an embodied agent asks people to explain in their own words, an LLM judges, the agent reteaches; randomized study, N=50 · map 2.3, 3.2
PROJECT-published

Life-Sized VR Health Coaches

rapport-building behaviors and agent approach in VR health coaching; people stood closer afterwards; IVA 2026 · map 3.1
EXPERIENCE
Ongoing

Research Assistant

at Relational Agents Group, Northeastern University, 08/2024 - Now
PROJECT-personal
Ongoing

AI PBL Agent

turns a teacher's lesson plan into an interactive project-based course, with a Socratic AI companion for every student · map 1.1, 3.3, 3.4, 4.2
PROJECT-teaching
Ongoing

First Principle Agent

a weekly AI workshop I design and teach for middle schoolers: they build games with AI, test them on real players, and iterate · map 3.3
PROJECT-personal

Keyi: Aha-First Courses

drop in a document, get readings, 2D interactives, quizzes and 3D scenes built around the few insights a learner must leave with · map 1.1, 1.2
PROJECT-personal
Ongoing

For Boaria!

an indie medieval-fantasy card tactics roguelike; 24 playable prototypes in seven weeks and an AI ladder to test faction balance
PUBLICATION

FMRG-MultiRobot

accepted publication by ACM SUI 2023
PROJECT-for publication

FMRG-MultiRobot

NSF-funded project; future manufacturing human-robot remote collaboration; multi-task management; 3D GUI; VR/AR; pre-cueing
PROJECT-for publication

FMRG-GazeCollab

NSF-funded project; future manufacturing human-human remote collaboration; eyetracking; cognition visualization; hand-gaze coordination; 3D GUI; VR/AR;
EXPERIENCE

Research Assistant

at CGUI Lab, 05/2022 - 08/2024
PROJECT-for publication

Unity Dev "Cheat Mode"

an Unity officially verified case-based tutorial for GPT facilitated Unity game full-stack production for the Unity Learn platform CN
EXPERIENCE

Instructional Designer

at Unity Technologies., 05/2023 - 08/2024
PROJECT-awarded

Physics Wonderland

a VR physics learning sandbox edu game, inviting learners to build, test, and play with their hand-crafted machinary with immersion
THOUGHT

"Walk" into Metaverse

prototyping alternative ideas for novel VR locomotion systems, 11/2022
PROJECT

Mesh Math

visualize complex 2-variable functions, procedural function 3D graph generation in AR
PROJECT

FPS Aim strategy analysis

comprehend player behavior and FPS game crosshair pre-placement strategies using gaze analysis
PROJECT

"Time Traveler" Museum

an immersive VR museum experience enabling users to travel back to the time when the displayed artifact was born and developed
PROJECT

Fallacy Finder

an AI-powered web plugin-liked tool promoting the public's awareness of media's frequently-used, contentious propagation strategies
THOUGHT

Conquer Physical Impairment

designing prototypes for AR/Neural Interfaces for harsh environment (like deep water or space) task assistance and collaboration
EXPERIENCE

Grad Student Researcher

at Columbia, 04/2022 - 05/2022
PROJECT

Transformative Game for Misconception

a VR role-play game design project that aims to combat public misconceptions on climate change
PROJECT

Workshop: GPT Multiplayer Dev

holding a workshop teaching online game basics and discussing the potential of GPT and prompts in multiplayer game development
PROJECT-awarded

"Flooded"

a serious board game that encourages international collaboration on solving global issues like climate change
PROJECT

Gamified Web Profile

my legacy homepage, a gamified personal profile exploration experience
PROJECT

Workshop: Learn Unity in Unity

holding a workshop teaching game development and Unity Engine basics, research novel learning material substitutes for powerpoint
EXPERIENCE

Lab Assistant

at TC Game Research Lab, 09/2021 - 05/2023
PROJECT

Topdown Pixel BR Game

a topdown pixel multiplayer battle-royale game, leading a community to iteratively build a commercial game
PROJECT

HDRP Tank Simulator

a demo for practicing Unity High-Definition Render Pipeline, post-processing, etc.
PROJECT

"One Man Army"

a topdown multiplayer combat game crafted with Unity URP, Blender, Garageband, self full-stack development
EXPERIENCE

Founder, Director

at Spectrum Forge GameDev Community, 10/2019 - 08/2024
PROJECT

"Mission Complete"

a first-person sandbox story-driven serious game
PROJECT-awarded

Game Haus

an alternative product for discord, a client-centered communication tool for esport fans
AWARD

Tespa ESport League

Top 50 Overwatch competitor Team, 2019
EXPERIENCE

Overwatch Team 1, Main DPS

at Madison ESport Club, 09/2019 - 01/2020
AWARD

Overwatch Top 200

NA server, 2019
AWARD

StarCraft2 Grandmaster Top 30

CN server, 2017
AWARD

National Physics Olympiad

3rd Prize, 2015
AWARD

Go Chess 3 Dan

2014
Publications

Contact Me:

I quitted Facebook and Twitter in 2019, realizing they were more distracting than beneficial. I have multiple GitHub accounts, this one is for contact.