We’re using the GenAIE platform to help service personnel transition to civilian life. We’re working with local councils and social workers to support the long-term sick, the long-term unemployed, home carers, disabled people, and people with mental health issues.
There are things you just couldn’t do five years ago, but now with generative AI, all of a sudden lots of things have become possible.”
The challenge
Teachers and key workers supporting learning in prisons and probation face a variety of difficulties.
Working in learning environments where resources and preparation time are often limited, prison educators have to meet the diverse learning needs of a varied cohort. The individuals they teach may be pursuing vocational training, literacy and numeracy goals, or their own personal interests.
To further complicate the task, prisoners and probationers may have very different learning levels, existing skills and academic backgrounds. Some might need to be taught in a completely different language.
As daunting as this sounds, the potential rewards – for both the individual offenders and British society – are considerable. Education and training help to reduce the risk of prisoners reoffending – a problem which is currently estimated to cost UK taxpayers £18bn every year.
The breakthrough
Working with Dan Brown, UCL’s Entrepreneur in Residence, UCL Professor of Computing Philip Treleaven is leading the Generative AI Education (GenAIE) programme.
GenAIE offers tools and training for key worker professionals to build courses, lessons and resources using a custom-built set of transformer-based (GPT) natural language models.
The GENAIE education platform harnesses agentic AI workflows. This means that individual natural language models within the platform act as independent agents, with different motivations and tasks to help achieve the desired educational outcomes.
Each AI agent takes responsibility for one of four main areas – curriculum content and compliance, student needs, pedagogical design, and moderation. This means course materials can be more easily fine-tuned to meet the evolving needs of the educator, the student and the penal institution, while still delivering education to accredited standards.
Overall, the aim is to provide moderated content (in a ‘walled garden’ network environment appropriate to the setting), that keeps human educators in control of the course generation process, while also automating the development of course content tailored to the individual learner.
GenAIE is fully compatible with existing e-learning platforms and builds on the Community Campus software already used in the UK Probation Service and Prisons (created by Meganeux, a successful UCL Computer Science spinout company).
Real-world impact
In the two years since the launch of the GenAIE programme with UK HMPPS in December 2022, the programme has seen exponential growth in usage. In a UK prison population of around 80,000 people, GenAIE has gained over 53,400 users and delivered over 596,600 hours of learning.
Given the sensitivity and restrictions on information and communication with learners in the prison and probation services, the GenAIE team can only share limited evidence of the programme’s real-world impact. However, operational managers within the UK HMPPS have reflected that the tools provided by GenAIE and Meganexus are a welcome respite from ‘fluffy’ courses, providing a greater level of control for their educators.
GenAIE are now exploring the potential of their tools and platforms in other sectors, through programmes such as the ‘Open Health’ collaborative research initiative.
They’re also hoping to expand on their work with prisoners and probationers, through partnerships with local authority Social Services and Jobcentre staff.
Additional information
Philip Treleaven, Professor of Computing, UCL Computer Science
Dan Brown, UCL Entrepreneur in Residence
UK HM Prison and Probation Service
Meganexus
Read the Royal Society article about GenAIE’s work with HMPPS
