What if computing could help people better understand whether an old legal conviction can finally be left in the past? Computer science PhD student Josh Alele-Beals is making online tools to help people navigate the complexities of statutory law. He’s currently focused on Michigan’s Clean Slate Law, which allows people to have some old convictions expunged from their public records.
Alele-Beals learned about the law from his advisor, Professor Charles Wallace, and from Social Sciences Associate Teaching Professor Susanna Peters, an attorney who works pro bono on Clean Slate cases.
“This law has helped thousands of Michiganders leave old convictions in the past, but the confusing nature of the legal language has limited its impact,” said Wallace.
Researchers said Clean Slate proved to be a strong fit for the aim of the project: to harness the power of computational modeling, visualization, and artificial intelligence (AI) to give ordinary citizens answers to their legal questions
“I became interested in this project because of its direct impact on people’s lives,” said Alele-Beals. “Clean Slate laws can make a real difference in whether someone is able to move forward after a past conviction. Even minor offenses from years or decades ago can continue to limit employment opportunities, prevent people from securing housing, restrict access to credit, and block educational pathways.”
Alele-Beals said the barriers presented by past legal convictions affect more than individuals. “They ripple out to families and communities by limiting economic mobility and long-term stability,” he said.
After consulting with Wallace and Peters, Alele-Beals delved further into the practice of expungement through conversations with legal professionals across the state, including the non-profit organization Safe & Just Michigan.
To make the tool accessible, Alele-Beals developed a user-friendly interface designed to assist individuals and their legal advisors. “The interface identifies which convictions on a person’s record can be expunged under various legal interpretations and provides clear explanations for why certain convictions may not qualify,” he explained, adding that along with the immediate benefit of helping people navigate the complexities of expungement law, the tool is also proof of concept, demonstrating the broader potential of applying these kinds of tools in other scenarios.
In this College of Computing Q&A, Alele-Beals, who was recently awarded a Michigan Tech Doctoral Finishing Fellowship, provides a closer look at the motivation, support system, and impact of his research.
Q: Why did you decide to focus on Michigan’s Clean Slate Law?
JAB: Michigan’s Clean Slate Law stood out to me because it represents a meaningful effort to remove barriers and give people a fair second chance. At the same time, the law itself is complex and difficult for non-experts to navigate, which means many people who could benefit may not realize they’re eligible or understand how to take advantage of it.
That’s where the technical side comes in. Professor Wallace and I saw an opportunity to use technology to make complex statutory law more accessible and understandable.
“The combination of a challenging technical problem and the potential to help people rebuild their lives was something I couldn’t pass up.”
– Josh Alele-Beals, PhD student, Computer Science
Q: How is mathematical modeling helping you get around the problem of potential AI hallucinations and bias?
JAB: Mathematical modeling is central to how I address hallucinations and bias in AI. I’m using formal methods, specifically Alloy, which is a lightweight modeling language backed by a SAT solver, to encode the legal logic into a precise, deterministic framework. In this approach, the statutory rules are translated into a formal representation that serves as the authoritative source of truth for reasoning.
This means that instead of relying on AI to “interpret” the law on its own — which can introduce non-determinism, hallucinations, or bias—the core decision-making is handled by the formal model, where outcomes are constrained, verifiable, and reproducible.
At the same time, I leverage AI for what it does best: interacting with users, collecting inputs through a human-in-the-loop process, translating those inputs into the formal representations the model can consume, and explaining the model’s outputs in clear, natural language.
This hybrid approach allows me to combine the strengths of both systems — the determinism and rigor of formal methods with the flexibility and usability of AI, resulting in tools that are both reliable and accessible.
Q: How have you made the tools accessible and user-friendly?
JAB: Accessibility has been a core design priority from the beginning. The tools are built to allow non-experts who have no prior legal knowledge to apply the statute to their unique cases and understand not just the outcomes, but the reasons behind those outcomes.
We’ve taken a hybrid approach to meet users where they are. Some people prefer a chat-based interface where they can ask questions in their own words. Others want a structured form that guides them step-by-step through the relevant factors. And for those who already have their criminal history record through the Internet Criminal History Access Tool (ICHAT), we built a feature that lets them simply upload it and get a direct assessment. AI has been essential in making this possible by unlocking natural language interactions, translating the formal model output into plain explanations, and adapting to each person’s situation without requiring them to understand legal jargon or navigate complex eligibility rules on their own.
The goal is to make the law accessible in the fullest sense: approachable, understandable, and actionable for anyone who needs it.
Q: Please tell us about your sandbox vision for this project and its potential applications.
JAB: The sandbox vision is about creating an exploratory environment where users can interact with legal scenarios without consequence. Instead of just asking “Am I eligible?” users can test what-if situations. For example, how waiting another year might change their eligibility, or what would happen if a pending charge were dismissed. This helps people understand not just their current status, but the pathway forward and what factors are actually weighing on their case.
On the research side, the sandbox enables experimentation with new AI techniques, rule interpretation approaches, and modeling methods in a controlled, transparent way. We can test how well the system handles edge cases, measure accuracy against real statutory outcomes, and refine the reasoning before deploying changes to users. It’s both a user-facing educational tool and a development lab where we can iterate on fairness, explainability, and precision.
Long-term, I see the sandbox evolving into a broader platform for computational law—where researchers, policymakers, and legal aid organizations can model the downstream effects of proposed legislation, test eligibility rules at scale, and make statutory systems more navigable for the people they’re meant to serve. Clean Slate is the first use case, but the approach could extend to housing law, benefits eligibility, immigration pathways, and other areas where complex rules create barriers to access.
Q: You’re a software engineer who earned your master’s degree in computer science in 2024. Why did you choose Michigan Tech and why did you decide to stay here to complete your PhD?
JAB: I chose Michigan Tech because of its strong emphasis on applied research and the opportunity to work closely with faculty on meaningful projects. The environment encourages building systems that solve real-world problems, which align well with my interests.
I decided to stay for my PhD because I saw an opportunity to continue developing this work at a deeper level. The combination of supportive mentorship, interdisciplinary collaboration, and the ability to pursue research with tangible impact made it an easy decision. It’s a place where I can both advance technically and work on problems that matter.
Q: What’s next for you and your research?
JAB: Going forward, I’m focused on expanding the robustness and scope of these tools. That includes improving the integration between AI systems and formal legal models, scaling to additional areas of law, and continuing to refine explainability and user trust.
I’m also interested in exploring how these approaches can be used more broadly in domains where rules, uncertainty, and human impact intersect—not just law, but areas like policy analysis and decision support systems. Ultimately, the goal is to build systems that people can rely on in high-stakes contexts.
About the College of Computing
The Michigan Tech College of Computing, established in 2019, is the first academic unit in Michigan dedicated solely to computing, and one of only a handful such academic units in the United States. The college is composed of two academic departments. The Computer Science department offers four bachelor of science programs in computer science, cybersecurity, data science, and software engineering; four master of science programs in applied computer science, computer science, cybersecurity, and data science; and a doctoral program in computer science. The Applied Computing department offers four bachelor of science programs in cybersecurity, electrical engineering technology, information technology, and mechatronics; two master of science programs in health informatics and mechatronics; and a doctoral program is in computational science and engineering.
Questions? Contact us at computing@mtu.edu. Follow the College of Computing on Facebook, Instagram, LinkedIn and Twitter.