top of page
Search

ML in Policing Meet-up Round-up

Our one-year STAR-funded project came to an end with an ML in Policing launch!

It was a productive day with great speakers covering various parts of the model development journey which instigated discussions with a wonderful and engaged audience.




Day Overview


We had 5 speakers covering a variety of subjects that can be encountered when developing a machine learning algorithm from conceptualisation to deployment. This inspired a lot of discussion on reality-based issues that affect police forces attempting to leverage their internal data and knowledge to improve their systems through machine learning. The day was hosted at Middlesex University by Dr Ruth Spence and the Centre for Trauma and Abuse Studies (CATS).


Expand the titles below for videos and slides from each of the presentations

What is Machine Learning? by Tamara Polajnar

Dr Tamara Polajnar, an expert in ML and Natural Language Processing with experience in academia and industry, delivers a (very) brief introduction to how machine learning works and how it can be used to look at data and leverage the patterns hidden within.

  • Using a toy dataset to help visualise how different algorithms work

  • A quick run through different types of learning, their strengths, and how they can be useful

  • How neural networks and deep learning differ from some of the more standard models




Algorithms in Policing: Problems, solutions, and more problems! by Hazel Sayer

Hazel Sayer, a PhD candidate at Bournemouth University as well as a Research Fellow who has worked on issues surrounding violence against women and girls on projects like Operation Soteria Bluestone, gives us an overview of current uses of ML-based algorithms in policing.

  • The history of our project and the reason this website exists

  • What are some of the past failures, and how can we help to avoid them in the future?

  • How can algorithms be beneficial and what can we do to ensure that they are developed and implemented in a responsible way?

  • A brief introduction to RUDI and why frameworks that offer guidance for ethical ML development are necessary but often impractical.




How to Conceptualise and Quantify Problems in Policing by Simon Clifford

Simon Clifford, CEO of Cliff42 and a perennial digital transformation advisor, covers practical ways for quantifying problems in policing.

  • What does modern policing space look like?

  • How do you identify the points in the different processes where automation would produce the maximum impact with least amount of expense?

  • Acceptance of the inevitable failure in the innovation space and how to ensure process continuation within the ever-changing command structure

  • Designing your processes for information sharing without data oversharing




Demystifying AI by Stephen Anning

Dr Stephen Anning leverages his experience working on CESIUM, an Ethical AI platform for tackling child exploitation, and his PhD work to explain AI and how we can make it explainable.

  • AI vs Machine Learning and where are we now in the current wave of technologies?

  • Data sharing and how can it be used to enable better and more useful ML development

  • Model development, testing, and explainability

  • Explainability in NLP using toxicity in text as an example and why out of the box methods may not always translate to every domain




AI in Policing by Tony Joslin

Innovation Team Inspector (soon to be retired) from Devon and Cornwall Police, Tony Joslin delivers insights into what is needed to make AI in policing a success.

  • Necessity of understanding the current technological landscape

  • Embracing "futures thinking" as a strategy to plan to stay with if not ahead of the times

  • The current state of AI in policing

  • Learning to learn from mistakes and to accept the growing pains of adopting AI as a springboard for better practices with transparency and cooperation with the community

  • A brief interruption where by technical error (what's a live event without one, eh?🍁) we lost the latter part of his slides but Tony continued his most excellent presentation like a total pro.

  • Some examples of AI going pear shaped





Acknowledgements

This website, RUDI, and the launch meet-up exist due to funding Office of the Police Chief Scientific Advisor (OPSCA) via the Science, Technology, Analysis and Research (STAR) and the association with our project partners at the West Midlands Police.

The project contributors include Dr. Ruth Spence and Dr. Tamara Polajnar via Middlesex University; Hazel Sayer, Dr. Kari Davis, Dr. Terri Cole from Bournemouth University; and Prof. Miranda Horvath from University of Suffolk.

The event was hosted at Middlesex University on March 27th 2024.

Comments

bottom of page