Donald Harris Ph.D.

Projects

Findings are only useful if someone can use them.

Tools, models and prototypes built on top of the research, some shipped, some in development, all aimed at putting an evidence base in the hands of people making decisions this quarter.

In active development

The MHMA
Friction Compass.

A user-testing tool for mental health app development teams. It reads a new app or feature, scores it against a taxonomy of friction points and motivators drawn from real user data, and predicts a range of user outcomes, before a single live participant is recruited.

  • Prototype
  • Structured app assessment
  • Calibrated prediction
  • Behavioral health

The gap it fills

Automated user-testing tools already exist. They're good at catching surface-level friction, a confusing button, a broken flow. Reviewers consistently note the same limitation: they miss the nuanced reactions that drive the most valuable design decisions, and they're best treated as an early rough signal that still needs human calibration.

They're also horizontal. The same generic engine points at e-commerce, fintech and a suicide-prevention app. In behavioral health, that's not a small compromise.

The wager

  • Two independent dimensions, not one score. Inhibitors and enablers move independently, removing friction and adding value are different interventions with different returns.
  • Satisfaction is not a retention signal. Friction keeps a direct effect on abandonment after controlling for satisfaction and usefulness. A generic tool conflates "satisfied" with "will stay." This one doesn't.
  • Healthy churn is separated from real churn. A user who leaves because the app worked and a user who leaves because it didn't look identical in every engagement dashboard. One is a success. One is a failure.
  • Domain specificity as a moat. The taxonomy is derived from behavioral health users, not borrowed from general UX heuristics.

How it works Two stages

Stage 01, Reading the app

Structured assessment

The app is assessed the way a designer would read it: onboarding flow, check-in cadence, paywall placement, feature set, navigation depth. Each factor is scored on the same scale as the original survey items, so the results feed straight into the prediction model rather than living in a parallel universe of invented metrics.

Where a factor genuinely can't be judged from a static design, crash frequency or inconsistent behavior, the tool withholds a score and says so, instead of producing a confident-looking guess.

Stage 02, Predicting the response

Calibrated modeling

That profile feeds a model built on item-level survey responses from real users, preserving how the friction points relate to one another. The output is a range of likely user responses across satisfaction, perceived usefulness, and, the part that matters, which of the four trajectory types they land in.

Keeping the two stages separate is deliberate. Assessment and prediction are different jobs, and collapsing them into one unanchored judgment call is precisely the critique of current tools.

Where it stands Openly

Survey respondents did not identify which specific app they used, so individual responses can't be linked to the review-mined dataset. Validation therefore happens at the inhibitor-category level rather than app-by-app. A planned pilot compares simulated predictions against real review data for a set of known apps, drawing on the Study 1 corpus of 170,127 reviews.
The published model operates at the composite level; disaggregating to per-inhibitor weights is buildable but not yet done, and item-level modeling estimates more parameters from the same sample, so per-inhibitor standard errors are expected to be wider. The underlying sample is moderate, cross-sectional and specific to young adults. Claims are scoped accordingly, and stated that way in the product rather than in a footnote.
Two inhibitors, inconsistent behavior and glitches or crashes, require a second data source entirely: crash analytics, beta-test logs, or review mining. A deeper-analysis path that ingests that data directly is scoped as a premium tier rather than smuggled into the base read.

Also built

Tools in the field.

Shipped ยท v7

Systems Resource Mapping Tool

The working instrument behind Systems Resource Mapping, the method for walking a community through its own crisis response system, step by step, to find where people fall out of it. Used with state, county and federal partners.

  • Facilitation
  • Systems mapping

Application

Community Systems Mapping

A Python/Streamlit application for building and interrogating stakeholder network maps of community behavioral health systems, turning a whiteboard exercise into something that persists, updates, and can be handed to the next coalition.

  • Python
  • Streamlit
  • Network analysis

Prototype

Inhibitor-Abandonment Risk Index

The scoring layer beneath the Friction Compass, an index translating the Inhibitor-Abandonment Matrix into a comparable risk figure for a given product or feature set.

  • Risk modeling
  • IAM

Frameworks

Models I work from.

Some are mine, some are borrowed and adapted. All of them exist to make a decision easier for someone who has to make it with incomplete information.

Data Empowerment Model

A three-tier framework for building data fluency inside organizations that have plenty of data and no capacity to act on it, moving a team from reading a dashboard to designing the question behind it.

Inhibitor-Abandonment Matrix (IAM)

The four-quadrant taxonomy operationalized for product teams: classifying users by friction load and exit behavior so that healthy churn and product failure stop looking like the same number.

Systems Resource Mapping

A structured facilitation method that maps a community's crisis response as a sequence of steps and resources, exposing the handoffs where people are most often lost between services.

Theoretical foundations

Cenfetelli's dual-factor theory of inhibitors and enablers; the NASSS framework for non-adoption and abandonment of health technology; goal-attainment plateau; and coping self-efficacy.