Every probation and parole officer has had the experience: a client sitting across from them who seems different — more focused, more hopeful, more ready to leave the old life behind. But how do you capture that? How do you prove it? And in a caseload of dozens or even hundreds, how do you spot it at all?
These are the questions The Change Companies set out to explore at the 2026 APPA Winter Conference in Atlanta, where Chief Program Officer Valerie Bagley presented a session titled Narratives of Change: Leveraging AI to Identify Transformation in Community Corrections Clients.
The session introduced a theoretical framework for using AI-powered narrative analysis to detect markers of desistance in clients' digital journaling responses. The goal: to help community corrections staff do what they already try to do, only faster, more consistently, and with greater precision.
The backstory behind this work is a question posed by the behavioral health director of a large corrections agency — one that many in the field have probably asked themselves: "How can we see if participants are actually gaining insight from the digital tools we're providing?"
The agency was already using Atlas, The Change Companies' digital platform for delivering risk-need-responsive interventions through the evidence-based practice of Interactive Journaling®. They had engagement data. They had program completion numbers. But they wanted something deeper: evidence of real change happening inside a person.
Participant feedback was positive, and peer mentors within the agency were stepping up to encourage new residents to engage. Something was happening. The question was how to recognize it — and use it.
Desistance is not simply "going crime-free." Research defines it as the process by which individuals move from criminal behavior toward a more prosocial lifestyle.
Studies show that individuals on a desistance path begin to show changes in identity, future orientation, social connections, and prosocial behaviors long before recidivism data reflects those changes (McNeill, 2016; Weaver, 2019). The problem for practitioners is that these shifts are subtle, qualitative, and easy to miss — especially under the weight of a full caseload.
"When staff know where an individual is in the process of desistance, they can tailor conversations to the person's readiness to change, offer encouragement at the right moments, and avoid one-size-fits-all approaches that might not align with current needs."
Desistance-aligned supervision isn't just good practice. Research shows it increases the likelihood of long-term behavior change (McNeill & Weaver, 2010). The challenge has always been identifying who is where — and doing so efficiently at scale.
Precision medicine transformed cancer care by using genetic data to personalize treatment rather than applying generic protocols to every patient. The same logic can be applied to community supervision.
With enough of the right data — in this case, narrative data from clients' own written reflections — it becomes possible to identify where an individual is on their change journey and tailor interventions accordingly. Not to punish, not to predict, but to support.
This is what The Change Companies is calling a "precision desistance" model: using AI-powered analysis of Interactive Journaling® responses in Atlas to detect meaningful signals of change, and surfacing those insights for staff to act on.
Drawing on decades of desistance research, the team identified four domains that serve as meaningful indicators of where someone is in their change journey. Each one can be detected — to varying degrees — in the open-ended journaling responses clients write in Atlas.
1. Prosocial Actions
Primary desistance — the behavioral stage — often shows up first. Clients start doing things differently: taking a class, getting a job, supporting a family member, walking away from a conflict instead of escalating it. These behaviors represent early, concrete signals that something is shifting.
In test data using anonymized Atlas responses, this domain captured responses like: "I'm studying right now to be a building inspector... speaking to people in my field and learning new things" and "I was arguing with my oldest and he said something that really upset me so I walked away until I was a lot more calm to handle the situation."
2. Changing Identity
Secondary desistance goes deeper: it is a shift in how someone sees themselves. This domain captures the formation of what researchers call a "redemptive script" — a narrative in which a person reframes their past as something that prepared them for a meaningful future, rather than defining them forever.
It also includes a sense of meaning and purpose: joining a cause larger than oneself, contributing to others, connecting to something that transcends material gain. Real responses flagged in this domain included: "I want to be the person who went through hell but never gave in to the failure... the one who forgave and learned how to love myself and others after such a long battle with anger."
3. Future Orientation
Hope matters. Research consistently finds that individuals who are desisting from crime develop a growing belief that a better future is possible — and that they have some agency in creating it. This "internal locus of control" is a significant predictor of sustained behavior change.
This domain captures two subdomains: self-efficacy ("I can," "I've done this before," "I will") and goal-setting (specific, time-bound, actionable plans). Atlas responses in this domain included: "I have a hearing coming up for a major infraction. I want to be able to sort through my emotions, speak clearly with my side of things and accept my punishment without crying or yelling."
4. Social Capital
No one desists alone. The development of meaningful, prosocial relationships — with family, peers, helping professionals, faith communities, or support organizations — is one of the strongest predictors of sustained desistance (Farrall, 2021).
This domain also captures something subtler but powerful: a growing concern for others' well-being. The shift from self-focus to other-focus is often a sign that a person is building the kind of identity that sustains long-term change.
Atlas already uses AI to generate individual progress summaries — highlighting strengths, flagging areas of concern, and offering recommendations for staff to explore in supervision sessions. To build on the precision desistance framework, The Change Companies envisions opportunities to break down individual desistance indicators and map client journaling responses to desistance domains, giving staff a snapshot of where momentum is building and where intervention may be needed.
Importantly, this is a tool for support — not surveillance. The vision is to help officers ask better questions, affirm the progress clients may not even recognize in themselves, and design case plans that meet people where they actually are.
The precision desistance model is presented with clear-eyed transparency about its current limitations and the ethical guardrails that must accompany it.
"Oncologists would never use insights on the origin of a patient's cancer to shame or degrade them about lifestyle changes that might have prevented the disease. Rather, these insights are used to treat the diagnosis in the most effective and precise way possible for that individual."
The same principle applies here. AI-generated narrative insights are not a basis for sentencing decisions, release determinations, or case plan completions. They are a lens for understanding — one that should be used to improve services, not to judge or restrict.
Future work on these features includes correlating narrative indicators with real-world behavioral data, testing for accuracy across diverse populations, and building the evidence base needed to responsibly deploy this capability at scale.
Community corrections is at an inflection point. Caseloads are high. Staff time is limited. And the field is increasingly asked to demonstrate outcomes — not just compliance. Meanwhile, AI tools are becoming more capable and more accessible than ever.
The precision desistance model represents a bet that these two realities can meet in a way that is both practical and humane: that technology can help officers see the change happening in front of them, and respond to it in ways that actually help.
As Valerie put it in closing: "Community corrections is ready for smarter tools — ones that respect the complexity of human change. By combining desistance theory with narrative analysis and ethical AI, we can move from a system that monitors behavior to one that supports transformation."
The backstory behind this work is a question posed by the behavioral health director of a large corrections agency — one that many in the field have probably asked themselves: "How can we see if participants are actually gaining insight from the digital tools we're providing?"
The agency was already using Atlas, The Change Companies' digital platform for delivering risk-need-responsive interventions through the evidence-based practice of Interactive Journaling®. They had engagement data. They had program completion numbers. But they wanted something deeper: evidence of real change happening inside a person.
Participant feedback was positive, and peer mentors within the agency were stepping up to encourage new residents to engage. Something was happening. The question was how to recognize it — and use it.
References:
Farrall, S. (2021). International perspectives and lessons learned on desistance. In: National Institute of Justice, Desistance from Crime: Implications for Research, Policy, and Practice. Department of Justice, Institute of Justice.
Johnston, T. M., Brezina, T., & Crank, B. R. (2019). Agency, self-efficacy, and desistance from crime: An application of social cognitive theory. Journal of Developmental and Life-Course Criminology, 5(1), 60–85.
Kosorok, M. R., & Laber, E. B. (2019). Precision medicine. Annual Review of Statistics and Its Application, 6, 263–286.
Maruna, S. (2000). Making good: How ex-convicts reform and rebuild their lives. American Psychological Association.
Maruna, S. (2001). Making Good: How Ex-Convicts Reform and Rebuild Their Lives. American Psychological Association.
McNeill, F. (2016). Desistance and criminal justice in Scotland. In: Croall, H., Mooney, G. and Munro, M. (eds.), Crime, Justice and Society in Scotland. Routledge.
McNeill, F., & Weaver, B. (2010). Changing lives: Desistance research and offender management. The Scottish Centre for Crime & Justice Research, 3.
Rocque, M. (2021). But what does it mean? Defining, measuring, and analyzing desistance from crime in criminal justice. In: National Institute of Justice, Desistance from Crime: Implications for Research, Policy, and Practice. Department of Justice, Institute of Justice.
Weaver, B. (2019). Understanding desistance: A critical review of theories of desistance. Psychology, Crime & Law, 25(6), 641–658.
Evidence-based, behavioral health Interactive Journaling® curricula are available digitally on Atlas. Atlas can save staff time while supporting fidelity to evidence-based practices.
Ready to see what Atlas can do for your program? Visit our website to schedule a personalized demo today. Learn more about Atlas →
Provide your information below for a complete overview of Atlas for your setting.