Case study

CogLoad

A biofeedback instrument that asks “is it my body, or my mind?”, synchronizing what a VR game demands with what your body spends, so cognitive and physical load can finally be told apart.

Type
Biofeedback research · VR exergaming
Role
Founding Researcher
Tools
EmotiBit · LSL (LabRecorder) · BSIPA plugin · Python
Timeline
Stanford Symbiotic Products · Jan 2026–present
Team
Solo (founder-as-user) · advised by Dr. Khizer Khaderi
BiosensingHuman factorsResearch instrumentQuantWearables
CogLoad
Design
Founder-as-user instrumentation pilot · n=1 · ~8 min synchronized capture
Evidence
~20 synced LSL streams (gameplay telemetry + EmotiBit HR/EDA) → time-aligned XDF · motion-artifact gating (~14%)
Analysis
HR↔difficulty correlation (r=.42–.46, p<.05) · accuracy cliff at peak demand
Problem

Is it my body, or my mind?

Halfway through a hard Beat Saber run, or a boxing round, you fall apart, and you genuinely cannot tell why. Is your body tired, or is your mind fried while your body is fine? Two tanks drain when you play: physical load (your body working) and cognitive load (your brain tracking and reacting under pressure). Both exhaust you.

2kinds of load drain you in play: cognitive and physical
1is all any consumer wearable actually measures
0tools tell you the split in the moment

Your watch shows heart rate; the game shows your score. Nobody tells you, in real time, how much of your collapse was body versus mind. I call that gap the embodied split, and the fix depends on the answer: if it is your body, you rest; if it is your mind, that is a different kind of break entirely.

Approach

Measure both channels, or measure nothing

The split is only knowable if you capture both sides on the same clock: what the game demanded and what the body spent. A controlled VR environment is the unlock: Beat Saber labels every second with exactly how hard it was pushing you. That is ground truth no wearable has, because no wearable knows what you were being asked to do.

So rather than guess at the split, I built the instrument that records both channels, synchronized, and keeps the physiology honest.

Process & artifacts

Building the ground-truth engine

Beat Saber telemetry pluginA BSIPA plugin streams every note hit, every miss, and saber motion live over Lab Streaming Layer (LSL).
EmotiBit physiologyA wrist-worn EmotiBit streams heart rate and skin response (EDA) onto the same clock as the gameplay.
One time-aligned fileLabRecorder lands ~20 synchronized streams in a single XDF: game demand and body response, second-for-second.
Motion-artifact gatingHard swings corrupt the optical heart-rate signal, so the system flags and removes the high-motion samples (~14% of a run) instead of trusting bad data.

A post-hoc Python analysis turns that raw capture into the Golden Graph: game difficulty, heart rate, and accuracy on one timeline.

Golden Graph: game difficulty, heart rate, and accuracy aligned on one timeline
The Golden Graph: heart rate climbs alongside difficulty, and accuracy falls off a cliff at the single hardest moment, the body and the game agreeing.
LabRecorder with ~20 synchronized LSL streams landing in one XDF file
Roughly twenty synchronized streams, gameplay telemetry plus EmotiBit physiology, captured on a single clock for second-level alignment.
Impact

The body and the game agree

r = .42–.46Heart rate tracked game difficulty: HitRate↔HR r=.420 (p=.024), NoteDensity↔HR r=.456 (p=.013)
0.59At the hardest moment (the note-density spike), accuracy fell off a cliff, performance collapsing as the body peaked
~14%Motion-artifact gating kept the physiology trustworthy by vetoing corrupted samples instead of trusting them
ValidatedThe instrument captures the ground-truth-vs-physiology channel the split requires. The readout is the scoped next build

Stated honestly: this is an n=1, ~8-minute pilot (founder-as-user). The instrument and these correlations are real; the literal cognitive-vs-physical split readout, consumer dashboard, and Fitbit / Google Health bridge are the near-term build, not finished output. I presented exactly that line (validated capability, scoped roadmap) to an industry panel including the President of Riot Games.

My role

CogLoad is a solo project: I conceived, built, and validated it end to end: the Beat Saber telemetry plugin, the EmotiBit + LSL synchronization, the motion-artifact gating, and the Python analysis that produced the Golden Graph. I ran the pilot as the first user and presented the work to a commercialization panel. Advised by Dr. Khizer Khaderi (Stanford Symbiotic Products).

Reflections & takeaways

The hardest call wasn’t technical. It was refusing to overclaim. In front of a sharp commercialization panel, the tempting story was “I separated mind from body.” The honest one was “I built the instrument that can, validated the channel on a pilot, and here is exactly what’s proven versus designed.” Protecting the gap between a validated capability and a finished product is what makes a result trustworthy. I also held two ethics lines that matter for biometric work: the data stays the player’s, and this is a sensory budget (energy spent on purpose), not a deficit to diagnose. The deeper insight is methodological: a game is a rare place where you get ground truth for free, because the game already knows how hard it is pushing you.

CogLoad: solo founding research, Stanford Symbiotic Products (Dr. Khizer Khaderi), Jan 2026–present. Instrument: BSIPA plugin + EmotiBit + LSL/LabRecorder; pilot n=1, ~8 min. Final presented to a panel including Hoby Darling, President of Riot Games.