What I think about AI
September 2023
The role of AI in fighting knife crime
The end of August sees the annual cornucopia of sights, sounds and smells that is the Notting Hill Carnival. An amazing event combining steel bands, Calypso music and Samba dancers. Not to mention the Jerk chicken and other dishes from all around the world.
Unfortunately, it was again marred by another alarming characteristic of London life, knife violence, 8 non-fatal stabbings on the final day. I did a quick google search on “Notting Hill Carnival”, the first four responses were articles on these stabbings followed by a call to move the event to a park. It’s such a shame that the great community work done by the carnival is continually overshadowed.
Knife violence in London is actually lower now than at its pre-pandemic peak when lockdowns saw knife crimes fall by a third. Since 2021 however, knife crimes have gradually been on the increase.
AI, or Artificial Intelligence, is on everyone’s lips at the moment, depending on who you read it’s going to transform everything for the better or destroy life as we know it. This is no different in areas of public life such as medicine and policing.
For Policing, the idea of AI scares people. Images of Minority Report or Robocop immediately spring to mind. For this reason, I will talk about AA, or Advanced Analytics, and their potential for transforming policing. The key difference between AI and AA, is that AA is human controlled, monitored and acts as a recommendation provided, not the final decision maker. If AI is ChatGPT writing your homework for you, think of AA as making some suggestions related to the references you might want to research for your assignment and some of the key arguments for and against your hypotheses that you may want to consider.
The advanced analytics of police data offers the potential to transform the way we deal with those people caught with possession of a knife. AA will help understand what interventions work to reduce reoffending and prevent further fatalities.
Currently around 60% of offenders receive an immediate custody sanction or suspended sentence. For those caught a second time the court system must impose a 6 months prison term or 4 months for young offenders. The average sentence was 7.4 months in 2021. What we know from data is that prisons aren’t impactful enough at rehabilitating people. Prison reduces post sentence employment chances and results in 60% reoffending rates.
AA has the ability to understand what works better than prison. By processing and analysing vast amounts of demographic information, offence histories, family backgrounds, school performances, and social environment data to create comprehensive profiles of knife crime offenders. Police data scientists can sift through this data to identify patterns, correlations, and risk factors that are indicative of potential success or failure in diversion programs.
Enhanced understanding of diversion strategies that work best for knife crime related offenders enables programs to channel individuals away from the traditional criminal justice system and towards rehabilitation, diversion and support. By harnessing analytical capabilities, police can gain deeper insights into the factors influencing the success of diversion programs and tailor diversion programs more effectively.
There are fantastic community organisations across the country working with offenders that produce incredible results in terms of reducing reoffending, arguably better than prison. Unfortunately very little data and research exists that truly understand what works. Three different offenders, Mary, Mark and Mikhail, different ages, educational backgrounds and other characteristics. What is the most appropriate intervention for each of them? Truth is we don’t really know. Currently we spend little effort understanding the needs of individuals and diverting offenders based on the root cause of behaviour.
We need to be bold and use alternative forms of diversion, assigning offenders peer mentors and voluntary work otherwise we will never begin to collect data around the effectiveness of these diversions.
Once we have a larger menu of choices, predictive analytics can assist police in determining the most suitable diversion strategies for each individual. By collecting this data from past cases in terms of compliance, feedback and reoffending, police can forecast the likelihood of a person reoffending based on various factors. This enables police to prioritise those at the highest risk and allocate appropriate resources.
For instance, if a model indicates that a specific type of person engages well in a peer mentoring scheme that significantly reduces recidivism, authorities can focus on directing similar individuals towards these programs. Another type of person may respond to engaging in sport and getting full time employment. Let’s connect them with a local gym and a recruiter to get them a job. Think of the menu like Netflix, with two different offenders shown different options based on the most effective programmes.
AA models improve over time. As data on intervention outcomes accumulates, models can refine predictions and recommendations. This iterative learning process enhances the accuracy of future interventions, ensuring that strategies remain aligned with evolving circumstances.
When you go to see the doctor, you don’t expect the same medicines prescribed as in Victorian times. We expect our doctors to know what’s likely to work for us. If a medicine didn’t stop the disease in 60% of cases, perhaps we would stop using it.
Budgets are tight throughout the public sector. PCCs face financial constraints when implementing diversion programs. Data can help optimise the allocation of resources by identifying the interventions that yield the highest impact. Through cost-benefit analyses, AA can assist PCC offices in deciding which programs are the most efficient in terms of reducing recidivism rates, thus maximising the use of limited resources. Moreover there are plenty of organisations that aren’t looking for funding from PCCs, they just need the referral stream to help people.
Prevention is however better than cure and AA can support this. Hotspot policing is used across many forces to make better decisions about where and when to deploy resources to have an impact in reported crime. AA can help understand this and use scarce resources more effectively. Analysing data and trends of offenders can help. AA should complement police expertise not try to replace it, help police understand neighbourhood concerns and collaborate with community stakeholders, including educators, social workers, and youth advocates, to ensure effective prevention strategies are followed. This collaborative approach can enhance the credibility of diversion strategies and foster a sense of ownership within the community.
While the potential benefits of AA in shaping diversion strategies are substantial, ethical considerations must not be overlooked. This is why I talk about AA, not a black box AA approach. It is imperative to ensure that the data used to train models is representative and unbiased. Biassed or incomplete data could perpetuate existing disparities in the criminal justice system. Moreover, transparency in AA decision-making processes is essential to gain the trust of the public and ensure accountability.
The expanded use of AA in policing has the potential to reshape the way decisions are made regarding the outcomes and diversion strategies for offenders of knife crime. Its role in shaping effective diversion strategies can contribute to a more just and rehabilitative approach to knife justice, helping police make more informed decisions that lead to better outcomes for offenders and society at large.
