Of all the tools available to civil servants, Artificial Intelligence will have “the biggest impact” on the quality of citizens’ lives, says Joel Tito, Programme Manager for AI and the Future of Government at the Centre for Public Impact (CPI).
There are three key areas where they believe it will make a big difference: healthcare, policymaking and public safety. GovInsider caught up with Tito to showcase how AI will improve these areas.
1. Healthcare and disease management
One area CPI is exploring is healthcare. Machine learning algorithms “have the potential to detect emerging disease outbreaks quickly and accurately”, allowing healthcare officials to respond faster, Tito writes in a new report on the impact of AI. This is crucial, because cutting the detection and response times by even two days can reduce deaths by six times, according to the US’ defence research agency, DARPA.
In the UK, the National Health Service has tested disease surveillance systems in a number of hospitals. The system alerts infection control officials to take preventive measures against possible outbreaks. If this system were to be implemented across the entire NHS network, it could save as much as £38.5 million (US$51.2 million) annually.
Two areas ripe for improvement are patient management and radiology. A physician’s day-to-day job includes predicting how long a patient will need to stay in hospital based on their profile. “That’s a task very suitable to predictive analytics, that AI can do. In fact, AI can get a lot more information and notice more minute differences in a case,” Tito says.
Radiology, on the other hand, can benefit from “computer vision”, he says, where a computer can ‘see’ a cross-section of an organ and tell whether it is cancerous or not. While radiologists go through “decades of training” to tell whether a shadow is a malignant tumour or not, an algorithm can be trained to do that job faster. Radiologists can then “see what the system flags and perform a visual check afterwards,” Tito adds.
AI can make policy decisions more personalised. At the moment, the government “needs to be close to a constituency in order to determine what the most appropriate intervention is,” he says. Typically, this means having local officials in the field, gathering and housing information on communities’ needs and preferences.
AI is able to automate that information-gathering on a large scale, saving time and resources. It can “source information from social media platforms to identify problems and gauge public sentiment,” Tito writes in the report.
Governments in Asia are already using this approach: Jakarta uses social media data to track traffic jams, while Singapore has built the Risk and Horizon Scanning tool to scour big data for potential threats.
AI can then also analyse the information and make decisions based on the preferences of each neighbourhood, community and person, if necessary. “The real power of AI is its unconstrained decision-making,” Tito says. “With AI, governments can more easily identify individuals, entities,
regions, or other factors in the greatest need of assistance or at the highest risk of a particular issue.”
3. Public safety and courts
In the US, a number of large cities, like Los Angeles, Chicago and New York, are developing machine learning-driven approaches to policing, in what is now known as “predictive policing”. “Police departments have the ability to use the predictability of criminal activities to their advantage,” Tito says. It has primarily been used to “predict crime hotspots, but they’ve also used it to predict gun violence and adverse police interactions”, he adds.
Los Angeles trialled predictive policing in selected neighbourhoods in 2011. Burglaries fell by 27% that year and property theft by 19% the following year, relative to neighbourhoods where predictive policing was not use, Tito explains.
Meanwhile, AI can improve the way courts assess the likelihood of someone re-offending, he says. These decisions are traditionally made in an “information-constrained, largely subjective” environment. “Machine learning algorithms can improve not only the efficiency of the court – for example, increase the speed at which bail decisions are made – but also, the quality of those decisions,” he explains.
“Machine learning algorithms can improve not only the efficiency of the court but also, the quality of those decisions.”
In the US state of Pennsylvania, the government is drafting a plan to allow judges to use such algorithmic risk assessment tools to assist with sentencing decisions. A study by researchers at the University of Pennsylvania showed that the algorithm was more accurate at forecasting than conventional modelling approaches, Tito says.
However, the “big caveat” officials should be aware of here is the risk of building in prejudices into AI. If an algorithm is trained with race- or gender-biased crime data, it will learn a correlation between the crime and gender or race. “When it comes to public safety, the most crucial thing is the transparency around training datasets, and artificial intelligence literacy and numeracy,” Tito notes.
While plenty has been said about the risks of AI, its potential for good is undeniable. The Centre for Public Impact is here to help.
To learn more about Artificial Intelligence and its public impact, fill in the form receive weekly updates from CPI.