Game On: Stop talking about data and start putting it into play

A recurring theme at every policing event we attend is the need to improve training to enable officers to make best use of the ever-increasing range of digital data sources that are at the heart of every investigation.

The training curriculum, however, is already bursting at the seams. In person courses can be expensive, hard to schedule and draw too many people away from operational commitments and E-Learning generates an allergic reaction from many.

So, how can we address the skills gap? What if we were to turn training into gaming and let people learn by actively exploring data as part of an online game?

Gaming tutorials are a critical part of any game. They help to immerse players into their new reality with less at stake than when you are playing for real. They teach players how to navigate the world they find themselves in, the rules and limitations of their actions while working towards the ultimate goal of the game. What if we applied this same thinking to digital training for policing?

Making games feel real

We've explored this idea over the last months to see what’s possible when we bring together police digital investigation expertise, gaming platforms, data science techniques and the very latest synthetic data capabilities.

We didn’t want players to be shown an investigation; we wanted them to take part in the investigation.

Our prototype was built upon the classic concept of a Whodunnit – because let’s face it, everyone loves a crime drama. The game, Last Orders, starts with a murder scene in the upstairs office of a pub. CCTV shows the landlord entering his office and drinking poison when he pours himself a drink, however no prior footage is available to see who the suspect might be.

The player has to piece together the evidence collected from the victim’s phone, plus digital information collected from three possible suspects. The data available to the player includes communications data, ANPR, CCTV, financial records, intelligence reports and even social media posts. Working against the clock, the players build up a narrative, collate evidence and decide who they think is the murderer - just as they would in a real murder investigation.

If the player selects the correct suspect, they are taken into the next level of the game - a manhunt. This leads them to search through additional call data, cell site and ANPR to locate and arrest the suspect.

Think of it like a digital version of Cluedo – with all of the data sources that form part of today’s digital investigations and a far more complex set of decisions to make beyond whether the murderer was Professor Plum with the lead piping in the library. By applying Generative AI techniques and drawing on our Synthetic Data capability, we can generate comprehensive, realistic and coherent digital data at scale – including plenty of noise for the investigator to search through to solve the case.

Learning while playing

While the prototype was originally developed to showcase our Synthetic Data capabilities and test out our ideas around gamification, we have found it has struck a chord across the operational community. While the importance of high volume, realistic synthetic data is now recognised for developing new capability, its full potential as a training tool has yet to be explored.

Improving the understanding of the different types of data available to policing, such as what it tells you, where it is stored, how to acquire it, and how to piece it together is a key component in addressing the skill gaps around digital data and could directly influence the pace, direction and effectiveness of investigations. Building on the initial prototype, we have now developed a new set of games, focused on broader investigator training needs.

One of the key challenges an investigator faces is to understand what different datasets might tell them and their relevance; which data will help them find the missing person, the outstanding suspect or the missing evidence. They also need to understand what data is available to them and how to request it – and how this data could be the missing piece of their investigation jigsaw. This tradecraft is needed long before they even begin to analyse the data itself.

Then they must identify the gaps, inconsistencies and assumptions they have to make when interpreting the data. They must compare records, recognise links and piece together events that may have occurred across different locations, systems and periods of time.

This is a skill that is learnt through experience and an understanding of how digital data can be exploited to find the missing link. With E-Learning packages, who hasn’t been guilty of trying to get through it as quickly as possible, scoring just enough in the quiz not to have to do it again and getting off the naughty list? There is no margin for error in a real investigation, however, and as real data cannot be used as a training aid, it can be a bit of a shock to see raw data – haystack as well as the neat, tidy needle you are looking for – in your first investigation.

Part of the tradecraft of investigations is identifying when the data you are analysing is wrong. Perhaps the address is old. Maybe the phone is being used by somebody else. What if the vehicle has been sold? To keep on top of this, the investigator must constantly assess reliability, relevance and risk.

Delay the enquiry and a suspect disappears. Focus too heavily on one theory and another line of enquiry goes cold. Apply for data without sufficient justification and the request is refused. Fail to bring the information together and the crucial patterns remain hidden.

By using a broad set of Synthetic Data, built up against the patterns of life of the characters in our game, we create a far more powerful learning environment than a conventional training course. It is an environment that develops curiosity, judgement, collaboration, critical thinking and investigative instinct. It teaches people not just how to search for information, but how to question it, connect it, sequence it and justify what they do next. It can be adapted for different roles, crime types, technologies and level of expertise and can help individuals test investigative processes, systems and policies before they encounter the same challenges in the real world.

This gamified training environment contains agents who operate within a synthetic world, moving throughout the simulator to show everyday commutes, school drop-offs, food shops as well as criminal activity woven into the fabric of the noise across the UK landscape. The data is then presented to the player based on an initial scenario to encourage them to start their investigation and request new lines of enquiry based on the data they currently have access to and what they learn from it.

Alongside victims, witnesses and suspects, other NPC agents can be included to represent gatekeepers, such as SPoCs or other authorisers. For example, requesting communications data during the investigation would trigger the data acquisition process and interactions with an Agent to guide the player through and ‘grant’ them data based on their given justification.

This provides players with informative hints and tips throughout the game and could also help should they get stuck. Just as in a real investigation, winning isn’t about requesting as much data as possible, it’s about knowing what to look for, justifying why it’s needed, and spotting what everyone else missed. So, do you think you’ve got what it takes to find out Whodunnit?

To find out more about data gamification for training and wider synthetic data innovation, contact policingapps@principleone.co.uk.