Corrections_Today_Fall_2024_Vol.86_No.3

TECHNOLOGY

carefully considered, thoughtfully developed and inten tionally implemented. The promise of AI AI is a powerful and adaptable technology that is well-suited for solving an array of problems. AI systems are already being applied across different human ser vices and justice-related contexts. Doctors, for instance, can use AI applications to detect and diagnose cancer and create customized treatment plans (Sufyan, 2023). Facial recognition systems are used in law enforcement to verify a person’s identity at border crossings (U.S. Customs and Border Protection, n.d.). AI systems can also be used to detect fraud or intrusions into computer systems (Jada & Mayayise, 2023). “AI algorithms are trained on historical data that may reflect existing societal biases that can be perpetuated or even amplified in correctional contexts.” In fact, we’re just starting to scratch the surface of what’s possible with AI, including in correctional set tings. For example, AI can be used to monitor inmate communications inside and outside of a prison. Large language models can review phone call transcripts, inmate letters and other interactions to identify patterns that impact the facility’s safety and security. Humans can then review any flagged content to decide on an appropriate response. AI technology can also be used to detect contraband more easily by analyzing body scanner imagery. AI can also provide administrative support in the form of chatbots that can answer routine questions from callers, automate visitation schedules and even provide comprehensive reports of a facility’s operational data to identify problems and suggest solu tions (Redden et al., 2020).

In community corrections, AI offers a promising range of applications, including risk assessment and as an adjunct to supervision. AI systems can analyze large data sets to identify individuals at higher risk of recidi vism or violation. AI-powered tools can also automate tasks, such as monitoring a client’s compliance with conditions, and flag any potential problems for further investigation. For example, some systems can be used to monitor and review a probationer’s internet history to flag inappropriate content that a human can then re view for any violations of supervision conditions (IPPC Technologies, n.d.). A supervising agent could use such information to assess a client’s risks and needs in near real time. AI could also create a customized case plan closely tailored to a client’s individual needs (van Dijck, 2022). In fact, anywhere a data-driven decision could be made, AI could augment human analysis. Ethical concerns The potential benefits of using AI in corrections must be balanced against critical ethical questions. One crucial issue in correctional contexts is the possibility of bias and discrimination that may affect decision-making. AI algorithms are trained on historical data that may reflect existing societal biases that can be perpetuated or even amplified in correctional contexts. For example, studies have shown that AI-based risk assessment tools used in policing can disproportionately target minority communities (Heaven, 2021). Biases based on gender, socioeconomic status and related personal characteris tics could also find their way into an AI system, which could lead to discriminatory decisions. The technological reasons underlying the bias problem are complex, but the presence of bias in any form is a serious ethical con cern that must be addressed for AI systems to be used effectively, especially in correctional settings. One of the challenges to addressing bias is the “black box” nature of many AI systems, which makes it difficult to understand how the AI’s underlying programming is arriving at a decision. This raises concerns about both accountability and due process. AI systems are trained using data consisting of millions or even billions of data points collected from around the internet. Biased infor mation in that training data can and has led to biased outcomes (IBM, 2023). The algorithm that the system

Corrections Today | Fall 2024

26