Event Round up: 'De-mystifying the Foundations of AI in Justice and Public Safety'

Event Round up: 'De-mystifying the Foundations of AI in Justice and Public Safety'

On 14 October techUK's Justice & Emergency Services Management Committee's (JESMC), AI sub-group hosted and an afternoon of panel discussions and networking on the theme of 'de-mystifying the foundations of AI in Justice and Public Safety. Attendees heard from leading industry stakeholders and senior members of the Civil Service.

For our first panel discussion, we were delighted to be joined by a panel of SMEs and Police Digital Service, reflecting on their lessons learned 'from proof of concept to deployment' when adopting AI. We were joined by Milan Bollegala, Business Analyst - Police Digital Service , Julia Muraszkiewicz, Head of Programme Against Exploitation and Violence - Trilateral Research, Abbey Cooper, Faculty AI, Kuba Misiorny, CTO & Partner, Untrite Ltd, Caroline Barnett, Justice and Emergency Services Committee (AI sub-group member) as chair.

Kicking off the conversation, attendees heard directly from SME's their first-hand experience delivering on the Police Digital Service (PDS). Whilst being a pathfinder is never easy, the panel discussed the benefits of having a shared production platform and the opportunities this would create for future projects. We heard, for some, that whilst the technology was "easy" from an engineering perspective, there was a lack of trust around it which had to be dealt with to obtain 'buy in' from stakeholders on what was being proposed, the risks that it would entail, and the benefits achieved. Explaining what the technology does and getting stakeholders comfortable was key.

Defining the problem

When strategising a plan for AI deployment, the panel advised to start small, explain the solution, discuss the results, bring along the front-line policing staff with you, discuss the risks, and let them see the benefits for themselves. This was key to getting a wider buy in and being able to progress beyond proof-of-concept stage. From the outset, and before anything else, the panel highlighted the importance of 'understanding the problem you are trying to solve' right from the word go, whilst incorporating different perspectives from the wider team when identifying and tackling the problem

Collaboration is key

Attendees heard how driving a collaborative environment, establishing and maintaining effective relationships with the parties involved is crucial. This will save time later on, as it ensures continued support and will allow the project to progress smoothly, particularly when it moves to deployment stage.

The panel then went on to emphasise the importance of co-design sessions in the early stages of a project. Conducting these will help you attain a strong baseline which everyone is involved with, and everyone buys into. This will not only provide a more well-rounded design but will also help build the needed trust for buy in.

Scalability

Drawing on the work and role of PDS, the panel discussed scalability and 'the pursuit of innovation'. Where there are opportunities for forces locally and nationally, through a mechanism, to share best practice, case studies and proof of concepts to avoid 'reinventing the wheel'.

Addressing the natural difficulties of this, it was emphasised that nnovation could happen quicker if all vendors had open API as part of their systems, using industry best practice, making it easier to integrate with their systems.

The panel also gave us insight into the barriers encountered by police when deploying AI. As whilst they seek agility when it comes to working with vendors where flexibility is key, SMEs have an important role to play in providing innovative solutions to the Police.

Overall, any AI solution has to be explainable to stakeholders and can't just be a black box. Explaining the solution is key to earning the trust of users so they understand what the technology is doing and the risks that might be involved. "Involve your stakeholders early as part of your co-design so their knowledge of the solution is agreed".

The second panel considered the role of AI in creating a better justice system, through ensuring access to skills, maintaining ethical and transparent practices, avoiding vendor lock-in and interoperability. We had the pleasure of hearing from Sapna Sanghvi, Deputy Director, Data Science and AI - Ministry of Justice, Matthew Cain, CDIO - Crown Prosecution Service, Manish Garg, Director VE3, Tess Buckley, Digital Ethics and Artificial Intelligence Safety Programme Manager - techUK, and Cinzia Miatto, Justice and Emergency Services Programme Manager, techUK (Chair).

Human in the loop

Firstly, for the effective use of AI it is important the human element of the interaction is properly accounted for, through ensuring that staff are AI literate. This is integral for any successful data strategy when getting the most out of AI, as it will allow data stewardship across the organisation and ensures that there is accountability within the organisation when things go wrong.

Progressing with the responsible AI stance, the panel touched on ethical and transparent practices within the criminal justice system which can be established through an ethical framework. It was suggested that these are built upon as good sources of ethical principles including UK Government principles and guardrails that have been published, OECD framework, NIST framework, and BSI framework.

The panel then considered the methods to avoid vendor lock in, from moving away from legacy systems, microservices to promote innovation to diversifying the eco-system through partnerships. Diversity in perspectives for innovation can be promoted through the involvement of SMEs. With this in mind, the Ministry of Justice is initiating a piece of work to look at how a case goes through the justice system end-to-end as a way of providing a joined up integrated service.

The final focus, exploring how AI could help in creating a better criminal justice system was interoperability. As initial applications of AI were "on top of" existing systems but now using the technology to envision systems differently from the ground up. In both instances, data challenges can persist and "we shouldn't forget the design principles and good practice that have been applied to system development when developing AI solutions". For this to be successful, it is crucial that the quality of the existing data is strong, preventing any data bias from persisting within the new systems.

Concluding the event, the AI subgroup had several key takeaways from the panel. To ensure that with any potential deployment, the challenge is understood to ensure that any AI adoption is outcome driven. Whilst AI is a transformative technology, it is also important to keep sustainability in mind and to make progress towards achieving Greener IT. Alongside this, trust and collaboration are key to using and developing AI systems. Once these conditions are attained and what 'good looks like' is consolidated, future use of AI can be explored.

Public link

Please use the above public link if you want to share this noodl on another website.