About AltReform
Builders With
Front-Row Seats
AltReform was founded in 2026 by Joseph McLean-Arthur, a Black man who carries the statistical reality of this system in his body every day, and Erin McGuire, a legal professional whose family has felt its weight firsthand. We are not observers. We are builders who happen to have front-row seats to both the problem and the technology that can help solve it.
Operating at the intersection of AI and criminal justice reform, we are based in Northern Virginia, in the shadow of the world’s largest data center corridor. Proof that the most powerful technology on earth can be redirected toward the communities that need it most.
Our Mission
To build the analytical tools that the criminal justice reform movement has never had access to, and to put them directly in the hands of the people who need them most.
The Team
A Black man who carries the statistical reality of this system in his body every day. Built AltReform because he was tired of watching brilliant people fight uphill battles with inadequate tools. Based in Northern Virginia.
A legal professional whose family has felt the weight of the system firsthand. Brings deep expertise in criminal justice policy, state legislative strategy, and the legal landscape of reform.
Our Values
Our model is trained on outcomes, not ideology. If a conservative reform produces better recidivism results than a progressive one, the score reflects that. We publish our methodology so you can audit every result.
Reforms that improve aggregate outcomes while worsening racial disparities are not successes. We score equity separately and weight it equally. A 90-point reform with a D on racial equity isn't a 90.
We build tools for advocates, public defenders, and community organizations. Not for prosecutors, prison contractors, or the political class that benefits from the status quo.
Every data source is cited. Every model is documented. Every score is explainable. If you can't understand how we reached a conclusion, we haven't done our job.
Our Methodology
We ingest from 40+ public datasets updated quarterly: Bureau of Justice Statistics, Vera Institute Incarceration Trends, National Conference of State Legislatures, state DOC annual reports, and RAND Corporation criminal justice research.
Each initiative is classified by type, jurisdiction, status, policy domain, and target population. Our NLP pipeline extracts key provisions and maps them to our 8-dimension outcome taxonomy.
Our gradient-boosted ensemble is trained on historical outcomes from 1990 to 2023. We predict recidivism impact, racial equity effect, cost-effectiveness, and political feasibility as separate dimensions, then composite-score them.
We apply the AltReform Equity Adjustment Factor to every score. Reforms that worsen racial disparities in sentencing, incarceration, or policing receive a mandatory score reduction, regardless of aggregate outcomes.
High-stakes scores (above 85 or affecting populations over 100,000) are reviewed by a criminal justice policy expert before publication. Machine learning informs. It doesn't decide.