United States Patent · US 9,891,792 B1 · Granted February 13, 2018
Method and system for building interactive software using biometric predictive models.
Filed October 30, 2015 at Intuit Inc., where I was Senior Lead UI/UX Design Architect. Co-invented with Yao H. Morin, Vi Joy Caro, Massimo Mascaro, Luis Felipe Cabrera, Amir Eftekhari, Nankun Huang and Damian O'Malley. Assignee: Intuit Inc.
The significance of the invention
Not just measuring the person. Changing the experience around them.
Filed in 2015, this invention described a meaningful shift in interaction design: biometric signals and user activity could inform predictive models that customize what happens next. The output could be a clearer question, a different sequence, timely assistance, a recommendation or background audio. The human state became part of the experience architecture, rather than a chart to review afterward.
- 01
Sense
Capture biometric signals during interaction.
- 02
Contextualize
Connect activity, personal baselines and predictive models.
- 03
Adapt
Customize the interface, support, sequence or audio.
- 04
Reassess
Use subsequent signals and interaction history to inform what follows.
A plain-language reading of the architecture, not a legal claim chart.
Why it still matters in the AI era
Generating an answer is not the same as understanding when a person needs a different experience. This work anticipated the broader move from static flows toward state-aware software. Today, wearables, adaptive sound and in-car wellbeing products publicly describe pieces of that same vocabulary. The enduring contribution is a UX principle: personalize the response to human context, not merely to a click.
Read the full analysis and supporting sources at PEMF MagazineThe contemporary product landscape
An early architecture. A much wider conversation today.
From adaptive audio to wearable coaching and vehicle experiences, these companies publicly describe features that overlap with parts of the biometric, predictive, adaptive design vocabulary. PEMF Magazine compares those features and their limits.
These logos identify examples discussed in that analysis, not partners, licensees or users of this patent. Public feature similarities do not establish copying, adoption or infringement. Intuit is the patent's listed assignee.
Soaak
Adaptive sound therapy
Wearable and self-report inputs, tailored sound sessions, remeasurement and ratings.
Strong conceptual alignment
Endel
Responsive soundscapes
Sound adapts to heart rate, motion and environmental context. A persistent outcome-learning loop is not clearly public.
Partial alignment
Google / Fitbit
Wearables & biometric insights
Body Response, personal baselines, mood reflection and readiness-based recommendations.
Partial alignment
Samsung
Health & wearables
Energy Score, trend comparisons, wellness tips and sleep coaching.
Partial alignment
Hyundai / Kia
State-aware vehicle experiences
Occupant sensing, state analysis and sensory cabin responses, including concepts and demonstrations.
Partial alignment
Apple
Apple Watch insights
Personal typical ranges, biometric inputs, training guidance and user-entered context.
Partial alignment
Oura
Biometric tracking
Personal baselines, readiness, stress estimates, tags and conversational guidance. A complete automatic learning loop is not publicly established.
Partial alignment
WHOOP
Performance & recovery
Continuous biometrics, personal baselines, coaching, Journal insights and breathwork results form a publicly described user-facing loop.
Strong conceptual alignment
Garmin
Wearable biometric data
Body Battery, personal baselines, stress estimates and tailored sleep guidance.
Partial alignment
Mercedes-Benz
In-car wellbeing
Publicly described wearable-informed wellbeing recommendations and an Endel research pilot.
Partial alignment
General Motors
Driver monitoring
Attention sensing, alerts and response-dependent safety escalation, rather than a documented affective learning loop.
Monitoring & safety response
Further reading · PEMF Magazine
The research behind the comparison.
PEMF Magazine examines the 2015 patent filing, the biometric predictive-model architecture and its relevance to today's adaptive wellness products. The article connects the patent record with publicly documented product features and supporting sources, explaining both the conceptual overlaps and the limits of each comparison.
- The original patent, its timeline and co-inventor attribution.
- How biometric signals, personal context and adaptive responses connect.
- Company-by-company analysis, source references and gaps in public evidence.
A sourced editorial analysis, not evidence that these companies use or license this patent, and not a finding of infringement.
In one sentence
A method for building predictive models of how a person feels, from biometric signals like heart rate, facial and voice data, and using those models to compose the software interface they see next. On my side of the work I have always called it the digitization of emotion.
Why it exists
Analytics tell you what someone clicked. They never tell you that the person was confused, frustrated or anxious while clicking it. Every product decision I have ever argued for has come down to that missing signal. The patent describes a way to capture it as structured input rather than inference after the fact.
How it works
Biometric data is captured while a user interacts with a system. That data is correlated with interaction events to train an emotional predictive model for that user and cohort. A rule engine then composes the interface, pacing, density, sequence, tone, from the model's live prediction, rather than serving one static flow to everyone.
Where it points
Adaptive learning that slows down before a student gives up. Clinical tools that surface a simpler path when the operator's stress spikes. Financial products that stop pushing at the exact moment fear takes over. All of it is the same idea: software that adjusts to state, not just to input.
In the patent's own words
Biometric data is collected to obtain more detailed, connected and reliable feedback from users of an interactive software system, on a more empirical and objective basis. That data creates emotional pattern predictive models for individual users. Those individual models are then analyzed together to generate emotional pattern profiles for whole categories of users, and used for targeted product diagnosis, targeted interventions, targeted offers, and the grouping and analysis of feedback sources.
The thinking behind the patent runs directly into my current work on frequency-based therapy at iTorus, where the body's response is the product's primary signal.
See the iTorus App case study