When a social media site throws out an advert for a product you were just discussing over the phone, it’s easy to jump to the conclusion that ‘they’ must be listening, says Aura CEO and co-founder Warren Myers.
The truth, however, is that the site employed artificial intelligence (AI) to predict your behaviour.
“You searched for a yeast starter last week and commented on a friend’s photo of sourdough bread yesterday. The ad for a bread-making course that seemingly pops up out of the blue was shown to you because the data predicted you might be interested in it, based on your own and previous users’ behaviour,” notes Myers.
“Those same principles can be applied to fight crime – and soon will.”
Aura is a security and medical response marketplace, gathering and centralising data from more than 170 private security companies before applying AI to dispatch the closest vetted response vehicle to its more than 500 000 active users.
There is a wealth of information to predict the likelihood of crime, says Myers.
“The data exists to predict, and thus prevent, crime, but it’s just not being analysed as yet.”
However, when it comes to predicting crime, it’s not quite as simple as scraping social media to find people who did successive searches for ‘crowbar’ and ‘balaclava’.
Data of use here spans from the obvious indicators, such as a person’s presence in a bad part of town, to the more surprising, which includes weather patterns and the days of the week.
A Finnish study, for example, showed that a 1 °C increase in temperature results in a 1.7% increase in criminal activity, based on two decades’ worth of data.
Another study in the US proved that vehicle theft spikes on weekends and in the evenings, notes Myers.
Also, science has even shown that when a local football team loses unexpectedly, domestic violence incidents increase by 10%.
By collating all this data, as well as past crime statistics, information from tipping lines, social media scraping, CCTV, and more, crime can be predicted, and emergency services proactively dispatched before it is too late, says Myers.
The same principle can also be used to predict spikes in traffic accidents and to dispatch emergency services to nearby locations for even faster response times.
The AI employed here is similar to Uber’s algorithms which predict when and where there will be a high probability of ride requests, so the service provider can dispatch drivers proactively.
Again, the same principle can also be applied to riots, adds Myers.
“Riots, like the ones parts of South Africa was subjected to recently, are one of the most difficult emergency incidents to manage because they have such a staggering snowball effect.
“History has shown that once a riot escalates past a certain point, almost nothing can be done that won’t be to the detriment of everyone involved. By predicting it, it can be prevented or contained in the early stages.”
Using AI Proactively
Emergency services, from ambulances to private security and police, currently largely operate on a reactive basis, explains Myers.
“A call comes in, and a vehicle is dispatched to assist. As the country’s crime and emergency statistics keep increasing, it’s clear that a proactive approach is the answer.”
This will happen soon – predictability fuelled by AI and big data can reduce violent crimes by 25% as early as 2023, believes Myers.
“It is unclear how accurate our predictability will be within this timeframe, but we have pegged a really high expectation.
“We are aiming for crime and risk prediction to be the primary creator of real dispatches on the Aura platform – in other words, we want to be responding more to predictions than to actual panics. Once a person has panicked, the crime/damage has usually been done.”
Data engineers at Aura are already working on expanding its existing security and medical response algorithms to become the centre repository for risk data, says Myers.
There will be a number of data pools which will feed into...