For the 28th Annual McWilliams Probation Lecture on 14th July 2026, Professor Melissa Hamilton will speak on 'Artificial Intelligence in Probation: Opportunities, Risks, and Responsible Use'.
Melissa Hamilton is a Professor of Law & Criminal Justice at the University of Surrey and a Surrey AI Fellow with the Surrey Institute for People-Centred Artificial Intelligence. She holds a Juris Doctorate (law) and a PhD in Criminology and is a member of the Royal Statistical Society, International Corrections and Prisons Association, American Psychological Association, and the Association of Threat Assessment Professionals.
Her research is interdisciplinary and focuses on the use of AI and related technologies in criminal justice, sentencing practices, interpersonal violence, and trauma-informed approaches to legal and correctional decision-making. Before entering academia, Melissa worked as both a police officer and a prisons officer, experience that continues to inform her research and teaching.
Melissa’s work has been published across law, social science, and criminal justice journals. She also regularly contributes to public and professional discussions through print media, radio and television broadcasts, and online platforms including blogs and podcasts.
The respondent is David A Raho. David is a PhD researcher in Law and Criminology at the Institute of Law and Social Sciences at Sheffield Hallam University, investigating AI Maturity Models, AI Cultural Readiness, and the comparative adoption of artificial intelligence in probation and rehabilitation services across England & Wales, Brazil, and Japan. He has extensive frontline experience as a Probation practitioner spanning nearly four decades and now works as a member of the AI Team at HMPPS HQ. He has contributed to publications on the use of technology in Probation for both CEP and the UN, and is a member of the UNESCO expert network on AI and a Tutor at the University of Oxford on AI Governance. He is both a Fellow and a Trustee of the Probation Institute and a proud Napo member, having previously served as a Branch Chair in London and also as a National Vice Chair.
This event will be held at the Institute of Criminology in the lower ground floor seminar rooms.
Lunch will be served at 1pm. The lecture will begin at 2pm. Tea will be served at 3.45pm.
Please register for in-person attendance here.
Please register for online attendance here.
AI isn’t a future prospect in probation, it’s already here. It’s not hard to see where this leads: keyboards and IT hardware replaced by transcription recordings and voice control, intake points using biometric scans, individuals tracked from entry to exit through linked systems. Electronic tagging evolves into continuous, remote supervision, potentially layered with facial recognition. Risk assessments, licence conditions, even sentencing recommendations, already, much of this can be generated at the press of a button.
ReplyDeleteAnd the justification is familiar. We’re told the same story each time: workloads are too high, resources too thin, and this technology will make things more efficient, freeing up time for meaningful work. But in practice, that time rarely materialises as more time in the room with people under supervision. Instead, it’s absorbed elsewhere, redirected into managing the system itself and the PO becomes ignored and obsolete.
The trajectory is clear, but the real question isn’t what AI can do, it’s where human judgment remains non-negotiable. At what point does a practitioner step in, rather than defer? When does a relationship, built through presence, trust, and discretion, stop being central to supervision? Because if decisions are increasingly shaped upstream by automated systems, there’s a risk that practitioners become implementers of outputs rather than authors of them, asking AI what to do before deciding what should be done.
And underpinning all of this is a quieter, murkier issue: who is actually designing these systems, and whose assumptions and bias are being coded into them? If the logic behind risk, compliance, and intervention is embedded in software, then those choices don’t disappear, they’re just harder to see, and harder to challenge.
See you at the lecture.
What stands out in the comment above is the shift from “AI as a future tool” to “AI as a present influence on practice.”
DeleteIn probation, the key issue isn’t whether technology can support us, it clearly can in areas like transcription, information access, and consistency. The concern is more subtle: how much of the decision-making process is being shaped before a practitioner even begins their assessment.
If risk classifications, suggested licence conditions, or intervention pathways are increasingly pre-structured, then professional judgment risks becoming a form of confirmation rather than authorship. That has implications not just for practice, but for accountability, particularly if the underlying logic of those systems isn’t transparent or open to challenge.
The discussion I’d be interested in hearing at the lecture is where the boundary sits. What decisions remain firmly human, and how is that protected in policy and system design, not just in principle, but in day-to-day use?
Because the value of probation has always been relational and discretionary. If those elements become secondary to system outputs, then the role itself quietly changes, even if the job title doesn’t.
The practitioner has become the administration assistant for the tools they are forced to work with.
DeleteBe it AI or algorithms, the 'tools' have constricted the practitioners ability to make decisions to such narrow parameters it has become just about redundant.
https://www.globalgovernmentforum.com/uk-civil-service-trade-union-sets-out-demands-for-use-of-ai-in-government/
'Getafix
"As part of the "Transforming Rehabilitation" reforms introduced by then-Justice Secretary Chris Grayling in 2013-2014, the probation service in England and Wales began supervising an additional 40,000 to 50,000 offenders annually."
ReplyDeleteThe total annual probation caseload in England and Wales increased by 39% from ~175,000 in 2000, reaching 243,434 in 2008.
caseloads at 30 Jun 2010 = 239,041.
Total annual caseload reached approximately 241,000 by the end of 2015, following a low of 217,359 at the end of 2014.
Total Caseload @ end of September 2025: 246,502 offenders.
https://data.justice.gov.uk/probation/offender-management/caseload-total
So where did the additional 50,000 cases go, as promised by grayling?
Staffing crisis? Caseload crisis? What crises?
Based on data from the UK Parliament, the total number of probation staff in England and Wales in 2000 was approximately 15,240
175,000/15,240 = 11.48
At the end of 2010, according to Q4 2010/11 data, there were approximately 19,369 FTE total staff
239,041/19,369 = 12.34
By December 2025, there were 21,407 full-time equivalent (FTE) staff
246,052/21,407 = 11.49
______________________________________________________
Smoke, mirrors, pisstake, scam.
The figures are useful, but I think they risk oversimplifying what practitioners are actually experiencing on the ground.
DeleteA relatively stable caseload-to-staff ratio doesn’t necessarily mean the work is comparable over time. The nature of probation practice has shifted significantly since the early 2000s - more complex cases, higher levels of risk management, increased recall activity, and substantially greater administrative and compliance demands.
“Caseload” as a number doesn’t fully capture:
- the intensity of supervision required per case
- the volume of recording and system interaction
- or the fragmentation of time across multiple platforms and processes
So while the headline ratio may look similar, the weight of each case is arguably much higher.
The question isn’t just “how many cases per officer,” but “what does each case now involve?”—and whether staffing levels have kept pace with that complexity.
That’s where the sense of a workload crisis comes from, even if the raw figures appear broadly consistent.
I don't think there's any increase in complexity of cases or "intensity of supervision required per case", but would certainly agree with "volume of recording and system interaction" which comes with the necessary arse-covering & the burden of increased recalls.
DeleteBefore trusts there were comprehensive & intensive one-to-one supervision sessions & weekly team meetings which, in my experience, weren't the waste of time the 'new' managerialism labelled them as; they were key to the successful management of complex cases & intensity of case management.
But no-one has yet managed to explain the magic trick aka where grayling's additional 50,000 tr cases went.
Seen on the internet:-
ReplyDeleteGood Morning all
Payslips for April indicate employer contributions for pensions have tanked to 18.5% from 26%. Wider civil service contributions remain around 28.5% across other sectors.
The whole sweetener to get us to go into the civil service on terrible pay following TR was the pension contributions.
Was this part of the vote which they neglected to share with us?
So we got 6% pay rise and HMPSS saved paying 7.5% into the LGPS?
DeleteI had a look into this. The LGPS is a defined benefit scheme, so it won't affect the value of the pension at retirement, however, it means that the 6% pay rise is being funded by our own pension surplus! HMPPS and the complicit unions are laughing at us.
DeleteA public official called the Certification Officer oversees the registration, annual returns, mergers and finances of trade unions and determines any complaints about elections, as well as some other ballots and union rules.
Deletehttps://www.gov.uk/guidance/complain-to-the-certification-officer-about-a-trade-union-or-an-employers-association
I’ve looked at my March and April payslips alongside the GMPF valuation, and it looks more like a change in what’s being shown rather than an actual reduction.
DeleteMarch showed the full employer contribution (~26.5%), which includes both the primary rate (~18%) and secondary rate (~8%). April appears to show only the primary rate (~18.5%).
So the “missing” 8% is likely still being paid, just not visible on the payslip anymore.
That said, without any explanation, it understandably looks like a cut. It would be helpful for payroll to clarify exactly what’s changed and why.