The real-time role of AI assistance in command and control can deliver improved decision making and outcomes

The use of artificial intelligence (AI) and machine learning to crunch large (usually historical) data sets to identify and extract valuable and in some cases vital information has long been recognised across the private sector, and is becoming increasingly prevalent in public sector arenas.

However, while such technology has made some inroads into policing, its adoption within law enforcement has not been without concerns, mainly relating to issues around the ‘big brother’ role of policing, and suggestions that some AI machine learning algorithms can perpetuate historical biases.

For Hexagon Director Nick Chorley and the Public Safety & Security team within the company’s Safety, Infrastructure & Geospatial division, those concerns are recognised and understood.

But their focus is on a different role for AI and machine learning – as a real-time assistive tool in dispatch and command and control – where such concerns can be minimised, while delivering significant operational benefits.

The company’s Smart Advisor solution is available as part of its CAD system (HxGN OnCall Dispatch) and HxGN Connect, and uses AI, advanced statistics, and machine learning to continually mine and analyse operational data so that it can alert users to trends, anomalies and similarities.

And as Nick explained, having that additional overview and analysis in what can be time-critical and emerging scenarios can be invaluable.

“This is very much assistive technology – it’s not making decisions for you, it’s acting as an extra pair of eyes when you need them,” said Nick.

“That could be an extra pair of eyes looking at information across boundaries. For example, the repeat incidence of a crime or certain behaviour in areas that you are responsible for may not be very high, but just over the border or the boundary, there may be another set of cases that you’re not aware of, and which could have major relevance.

“The AI can make you aware of that activity, identify commonalities or key pieces of information that are linked to those various incidents, and highlight that the activity you’re looking at may be part of a bigger cluster, and therefore may be more significant than you realise.

“Those boundaries don’t necessarily have to be geographical, they could also be temporal. For example, you may have only just come on shift, and you’re seeing one or two reports or incidents taking place that, on their own, may seem isolated or not of concern.

“But you may not be aware of the 10 similar things that happened in the hour before you started your shift; so the AI can assist you by bringing things to your attention that you may not have otherwise seen.

Crucial contribution

While those well-known data crunching functions of AI and machine learning can save huge amounts of police time and resource – for example, when trawling through gathered data within investigations – it is the immediacy that the technology can deliver which is also crucial in the Smart Advisor package.

Nick is keen to emphasise the benefits that the technology can bring to real-time decision making: “The whole point of command and control is presenting information to an operator while they’ve still got time to do something about it, and make useful decisions about it.”

“So the immediacy of the assistance element of AI in these scenarios is critical, because it is being triggered by real-time events. The single new event or new call that has come in might be the thing that triggers the AI to say ‘this is now significant – you need to be aware of it’.

“Getting that notification about a significant event as soon as possible, rather than at the end of a shift when it’s too late, is crucial.

“An hour after something has happened, and with the benefit of hindsight, then of course you can easily work out what to do. But if you never had that information at the time, you may well have made the wrong decision.

“It’s all part of bringing that information together earlier in the timeline, so you have a better chance of making the right decision, rather than that information coming to you five to 10 minutes later, maybe even an hour later, and you’ve not had the opportunity to make the right call.”

Counteracting concerns

Hexagon are currently working with a number of public sector agencies to develop their AI in the public safety arena, although Smart Advisor is only available for use by existing customers.

The first to ‘go live’ with the AI solution is the city of Manaus in Brazil, which has integrated the technology into its City Cooperation Center (CCC) as part of its ‘smart city transformation’, but Nick and the team have been working hard to set the foundations for other agencies to benefit from the new technology.

“We’ve very much focused so far on trying to allay people’s fears about AI and the ‘big brother’ element. Once people have accepted the benefits of AI, it’s also important to remember that it is only as good as the information it learns from,” said Nick.

“A common concern is that if you’re giving AI biased information to learn from, you’re going to get biased results. But all we’re giving this technology to learn from is our customers’ historical data, which is just what they are collating in their everyday business.

“Users do need to look at their historic data and ‘fine tune’ the various agents, so that they will receive useful notifications at the right threshold – one that’s not going to completely overwhelm them.

“It’s a bit like the little boy who cried wolf; if the system is going to be shouting at you all of the time, that’s really not helpful, so you do need to spend some time testing the agents against the historical data.”

So are people still scared away from the use of AI in policing because of those concerns over ‘big brother’ and possible bias, or have forces and agencies now come to terms with some of those challenges?

“I’m not getting enough feedback either way to express a firm opinion, but we’re very conscious of the way some of those fears have been voiced and concerned in the media, and we’ve been working hard to counteract some of those fears.

“We’re also clear that we don’t personalise the way we use AI. There’s a lot of anonymity in what we’re doing; when users notice a pattern or correlation of events, it doesn’t relate to individual identities or characteristics.

“We’re not doing anything with facial recognition, we’re not analysing or trying to match names; we’re analysing calls and events, and the use of resources in responding to those events. That’s the scope of the data we’re working with, all in a locational or temporal context.”