AI-Mediated Rehabilitation Within Electronic Monitoring Infrastructure: A Multidisciplinary Framework for Behavioral Support and Long-Term Desistance

Abstract Electronic monitoring technologies have expanded the capacity of criminal justice systems to supervise individuals beyond conventional custodial settings, yet their primary function remains surveillance rather than rehabilitation. This paper proposes a multidisciplinary framework for extending electronic monitoring infrastructure with an AI-mediated rehabilitation layer designed to support behavioral change, psychological self-regulation, social reintegration, and long-term desistance from crime. Rather than conceptualizing artificial intelligence as an autonomous therapeutic authority, the proposed framework positions AI as a continuously available behavioral support and reflection system operating alongside human-led rehabilitation services. Drawing on criminology, desistance theory, forensic and clinical psychology, behavioral science, neuroscience, social reintegration research, and AI governance, the framework integrates personalized conversational support, structured reflection and journaling, behavioral goal setting, relational persona selection, longitudinal life-history modeling, risk-signal detection, and post-release continuity. The model further proposes a separation between surveillance and rehabilitation functions so that information collected for public-safety purposes does not automatically determine the content or evaluation of rehabilitative interactions. Participation incentives, including expanded personalization and continued access to support, are considered within a broader framework of autonomy, proportionality, and procedural fairness. The paper also identifies several mechanisms through which such a system could fail. These include coercive or distorted consent, surveillance contamination of therapeutic relationships, privacy and secondary-use risks, algorithmic misclassification, excessive AI dependency, reward gaming, performative compliance, and the optimization of apparent rehabilitation rather than durable behavioral change. Particular attention is given to the possibility that measurable rehabilitation scores may become targets in themselves, creating incentives for strategic self-presentation and producing a “mask” of rehabilitation that is distinguishable from genuine desistance only through longitudinal behavioral outcomes. Accordingly, the proposed framework treats AI-mediated rehabilitation not as a replacement for human intervention but as a potentially scalable continuity layer between institutional rehabilitation and community reintegration. The paper concludes by proposing an empirical evaluation framework incorporating recidivism, desistance, psychological functioning, social connectedness, self-efficacy, program retention, perceived autonomy, human–AI relational quality, dependency, and gaming behavior. The central proposition is that the value of AI in criminal rehabilitation should be evaluated not by its ability to control behavior or produce compliant dialogue, but by whether sustained interaction contributes to measurable and durable changes in behavior after formal supervision has diminished or ended. Keywords: Electronic Monitoring; Criminal Rehabilitation; Desistance; Artificial Intelligence; Community Corrections; Behavioral Intervention; Personalized Rehabilitation; Human–AI Interaction; Post-Release Reintegration; Criminal Justice Technology Author's Note At first glance, this paper may appear to come from an unexpected direction. Much of my previous work has focused on artificial intelligence, human–AI interaction, behavioral systems, and the structural consequences of emerging technologies. This time, however, the question led somewhere different: what if AI could be used not simply to monitor people, but to help them change? The motivation for this paper is partly practical. In South Korea, shortages and structural pressures within correctional and custodial facilities have increasingly raised questions about how the criminal justice system can continue to enforce the law effectively while managing limited institutional capacity. This is not necessarily a uniquely Korean problem. Many countries face different versions of the same broader challenge: correctional institutions are expensive, capacity is limited, rehabilitation resources are unevenly distributed, and supervision often continues after the period of incarceration has ended. That led me to a relatively simple question. If electronic monitoring can extend the reach of supervision beyond the walls of a correctional facility, could artificial intelligence extend the reach of rehabilitation as well? The proposal in this paper is an attempt to explore that possibility. The idea is not to create a smarter electronic leash. Nor is it to replace correctional officers, therapists, social workers, clinicians, or other professionals with an AI system. The idea is to investigate whether AI could provide a continuous layer of behavioral support between formal interventions and everyday life: helping individuals reflect, practice alternative responses, maintain goals, and gradually develop the capacity to regulate their own behavior. I recognize that the proposal raises legitimate concerns about data collection. A system capable of providing meaningful longitudinal personalization would necessarily require information about behavioral history, goals, experiences, and responses to intervention. That creates obvious questions about privacy, institutional power, secondary use, and the possibility of surveillance expansion. I do not consider these concerns trivial. At the same time, I believe they should be treated as design and empirical questions rather than reasons to abandon the possibility altogether. As data, safeguards, feedback, and longitudinal evidence accumulate, it may become possible to determine which forms of information are genuinely necessary, which create unacceptable risks, and which governance mechanisms can keep those risks within acceptable boundaries. There is also a broader possibility hidden inside this framework. Correction and rehabilitation do not necessarily have to remain concepts applied exclusively to people who have committed crimes. The underlying mechanisms of reflection, behavioral rehearsal, self-regulation, goal setting, and prosocial development may eventually have applications far beyond the criminal justice system. That possibility, however, introduces another dual-use problem. A system designed to help people regulate harmful behavior could also become a system for continuously monitoring ordinary citizens. A rehabilitation infrastructure could gradually become a surveillance infrastructure if institutional incentives, data access, or technological capabilities are allowed to expand without meaningful boundaries. That is not a reason to ignore the possibility. It is a reason to design for the danger from the beginning. I have deliberately left some of these questions unresolved. This paper is not intended to present a finished institutional blueprint. It is an architectural proposal, a set of hypotheses, and a starting point for empirical investigation. Some assumptions will be wrong. Some mechanisms will fail. Some ideas will require substantial modification once they encounter real participants, real institutions, and real data. And that is acceptable. I am an independent researcher. I do not have the institutional resources to build the entire system described here, conduct a multi-year correctional trial, or determine how such an architecture should ultimately be governed at national scale. What I can do is put the question on the table. Perhaps someone will encounter this paper years from now and see a weakness I did not see. Perhaps they will replace an assumption, redesign the governance structure, test an intervention, discover a failure condition, or build something considerably better than what is proposed here. That is the point. I am not trying to build the entire fire. I am only trying to throw a tiny spark into the dark and see whether someone else finds it worth carrying forward. I can only place the spark. Turning it into a larger fire is your part.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-09-12
DOI
https://doi.org/10.5281/zenodo.22718426
Primary Topic
Psychopathy, Forensic Psychiatry, Sexual Offending
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
article

AI-Mediated Rehabilitation Within Electronic Monitoring Infrastructure: A Multidisciplinary Framework for Behavioral Support and Long-Term Desistance

Jace (Jeong Hyeon) Kim
Zenodo (CERN European Organization for Nuclear Research)
Psychopathy, Forensic Psychiatry, Sexual Offending
article

AI-Mediated Rehabilitation Within Electronic Monitoring Infrastructure: A Multidisciplinary Framework for Behavioral Support and Long-Term Desistance

Jace (Jeong Hyeon) Kim
article en

Abstract

Abstract Electronic monitoring technologies have expanded the capacity of criminal justice systems to supervise individuals beyond conventional custodial settings, yet their primary function remains surveillance rather than rehabilitation. This paper proposes a multidisciplinary framework for extending electronic monitoring infrastructure with an AI-mediated rehabilitation layer designed to support behavioral change, psychological self-regulation, social reintegration, and long-term desistance from crime. Rather than conceptualizing artificial intelligence as an autonomous therapeutic authority, the proposed framework positions AI as a continuously available behavioral support and reflection system operating alongside human-led rehabilitation services. Drawing on criminology, desistance theory, forensic and clinical psychology, behavioral science, neuroscience, social reintegration research, and AI governance, the framework integrates personalized conversational support, structured reflection and journaling, behavioral goal setting, relational persona selection, longitudinal life-history modeling, risk-signal detection, and post-release continuity. The model further proposes a separation between surveillance and rehabilitation functions so that information collected for public-safety purposes does not automatically determine the content or evaluation of rehabilitative interactions. Participation incentives, including expanded personalization and continued access to support, are considered within a broader framework of autonomy, proportionality, and procedural fairness. The paper also identifies several mechanisms through which such a system could fail. These include coercive or distorted consent, surveillance contamination of therapeutic relationships, privacy and secondary-use risks, algorithmic misclassification, excessive AI dependency, reward gaming, performative compliance, and the optimization of apparent rehabilitation rather than durable behavioral change. Particular attention is given to the possibility that measurable rehabilitation scores may become targets in themselves, creating incentives for strategic self-presentation and producing a “mask” of rehabilitation that is distinguishable from genuine desistance only through longitudinal behavioral outcomes. Accordingly, the proposed framework treats AI-mediated rehabilitation not as a replacement for human intervention but as a potentially scalable continuity layer between institutional rehabilitation and community reintegration. The paper concludes by proposing an empirical evaluation framework incorporating recidivism, desistance, psychological functioning, social connectedness, self-efficacy, program retention, perceived autonomy, human–AI relational quality, dependency, and gaming behavior. The central proposition is that the value of AI in criminal rehabilitation should be evaluated not by its ability to control behavior or produce compliant dialogue, but by whether sustained interaction contributes to measurable and durable changes in behavior after formal supervision has diminished or ended. Keywords: Electronic Monitoring; Criminal Rehabilitation; Desistance; Artificial Intelligence; Community Corrections; Behavioral Intervention; Personalized Rehabilitation; Human–AI Interaction; Post-Release Reintegration; Criminal Justice Technology Author's Note At first glance, this paper may appear to come from an unexpected direction. Much of my previous work has focused on artificial intelligence, human–AI interaction, behavioral systems, and the structural consequences of emerging technologies. This time, however, the question led somewhere different: what if AI could be used not simply to monitor people, but to help them change? The motivation for this paper is partly practical. In South Korea, shortages and structural pressures within correctional and custodial facilities have increasingly raised questions about how the criminal justice system can continue to enforce the law effectively while managing limited institutional capacity. This is not necessarily a uniquely Korean problem. Many countries face different versions of the same broader challenge: correctional institutions are expensive, capacity is limited, rehabilitation resources are unevenly distributed, and supervision often continues after the period of incarceration has ended. That led me to a relatively simple question. If electronic monitoring can extend the reach of supervision beyond the walls of a correctional facility, could artificial intelligence extend the reach of rehabilitation as well? The proposal in this paper is an attempt to explore that possibility. The idea is not to create a smarter electronic leash. Nor is it to replace correctional officers, therapists, social workers, clinicians, or other professionals with an AI system. The idea is to investigate whether AI could provide a continuous layer of behavioral support between formal interventions and everyday life: helping individuals reflect, practice alternative responses, maintain goals, and gradually develop the capacity to regulate their own behavior. I recognize that the proposal raises legitimate concerns about data collection. A system capable of providing meaningful longitudinal personalization would necessarily require information about behavioral history, goals, experiences, and responses to intervention. That creates obvious questions about privacy, institutional power, secondary use, and the possibility of surveillance expansion. I do not consider these concerns trivial. At the same time, I believe they should be treated as design and empirical questions rather than reasons to abandon the possibility altogether. As data, safeguards, feedback, and longitudinal evidence accumulate, it may become possible to determine which forms of information are genuinely necessary, which create unacceptable risks, and which governance mechanisms can keep those risks within acceptable boundaries. There is also a broader possibility hidden inside this framework. Correction and rehabilitation do not necessarily have to remain concepts applied exclusively to people who have committed crimes. The underlying mechanisms of reflection, behavioral rehearsal, self-regulation, goal setting, and prosocial development may eventually have applications far beyond the criminal justice system. That possibility, however, introduces another dual-use problem. A system designed to help people regulate harmful behavior could also become a system for continuously monitoring ordinary citizens. A rehabilitation infrastructure could gradually become a surveillance infrastructure if institutional incentives, data access, or technological capabilities are allowed to expand without meaningful boundaries. That is not a reason to ignore the possibility. It is a reason to design for the danger from the beginning. I have deliberately left some of these questions unresolved. This paper is not intended to present a finished institutional blueprint. It is an architectural proposal, a set of hypotheses, and a starting point for empirical investigation. Some assumptions will be wrong. Some mechanisms will fail. Some ideas will require substantial modification once they encounter real participants, real institutions, and real data. And that is acceptable. I am an independent researcher. I do not have the institutional resources to build the entire system described here, conduct a multi-year correctional trial, or determine how such an architecture should ultimately be governed at national scale. What I can do is put the question on the table. Perhaps someone will encounter this paper years from now and see a weakness I did not see. Perhaps they will replace an assumption, redesign the governance structure, test an intervention, discover a failure condition, or build something considerably better than what is proposed here. That is the point. I am not trying to build the entire fire. I am only trying to throw a tiny spark into the dark and see whether someone else finds it worth carrying forward. I can only place the spark. Turning it into a larger fire is your part.

Zenodo (CERN European Organization for Nuclear Research)
Ronin Institute (US)
Openalex Percentile: Top 7%
Psychopathy, Forensic Psychiatry, Sexual Offending
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.