IT Systems and Data Protection

Artificial intelligence: reality or hype?

As society comes to terms with the fear and loathing of artificial intelligence (AI), is it time for councils to start considering what benefits the technology can bring to their organisation? LGiU chief executive Dr Jonathan Carr-West and associate Neil Reeder explore the possibilities of a digital future.

AI has massive potential to transform lives. There are some wild claims: a quick Google search suggests AI could learn to recognise galaxies, save us from the next cyber-attack, and even help stop religious violence. That’s in addition to the more well-known possibilities like driverless cars and machines diagnosing cancer. With all of these possibilities seemingly just around the corner, how is AI already contributing to modern life?

Data analytics

Every day, our behaviour and habits are tracked and stored as data. AI can analyse data incredibly fast, spotting patterns in large datasets and predicting consequences. This information helps decision-makers determine what actions are given priority.

AI has been applied to infrastructure by assessing stresses in the system and the probabilities that major problems will occur, such as by reviewing financial flows. HMRC has been a pioneer in AI usage, with its tax inspectors using Connect, an analytical system designed by BAE Systems, to quickly carry out preliminary investigative work. Connect collates data from sources ranging from banks to social media, then references this against tax records to flag up individuals or businesses to investigate.

We are also seeing signs of AI being used to assess human situations and circumstances, including the sensitive agenda of determining when children are at risk of maltreatment. A recent article by The Guardian identified that “at least five local authorities in England have developed or implemented a predictive analytics system for child safeguarding,” and that more than 350,000 people’s data “has been incorporated into the different predictive systems.”

Automation and augmentation

Another important use of AI by local authorities is replacing staff with call-centre chatbots. Such bots are said to be increasingly capable of analysing natural language, understanding context, sensing emotions, and providing answers. Enfield Council in North London, for example, has introduced Amelia, a robot technology dedicated to frontline council services – such as taking resident queries or authenticating licenses.

Experience from the private sector does, however, suggest that such systems can be a source of frustration to members of the public if they have more complex issues, particularly if the option of speaking to a human is made difficult.

These kinds of changes can affect morale at work as people worry about jobs. The lesson for local authorities is that there is a frequent trade-off between customer experience and cost (a trade-off that will change over time as the technologies develop).

AI also has the potential to bring together people and technology. Several councils such as Wigan and Hampshire are already trialling the use of voice-activated assisted technology. These enable people with a physical disability to achieve simple but desirable actions that others take for granted, such as opening and closing curtains, turning lights on, and opening the front door.

Shaping AI and adapting to it

AI is likely to result in fundamental changes to the way that local authorities operate, to the way that citizens receive services, and to citizens’ wider lives. However, as with any new technology, choices made as to how it is introduced and controlled make an important difference to the results. Consequently, before widening their usage of AI, local authorities need answers to the questions:

• How can we get AI analysis that is fair and unbiased?
• How can we address the legal consequences of adopting AI?
• How can we prepare internally for greater use of AI?
• How can we prepare communities for transformations in the use of AI?

Moving towards fair and unbiased analysis

AI tools are normally taught to spot patterns amongst large quantities of data, which form their ʻtraining datasets.ʼ In theory, this has the ability to be objective, and represents an “opportunity to better address longstanding prejudices and inequalities in our societies.”

Yet a major risk is that when the training datasets are unrepresentative and/or reflect historic patterns of prejudice, problems of injustice can arise. For example, an Amazon AI tool designed to assist in recruitment has recently been stopped precisely on these grounds, since it “taught itself that male candidates were preferable,” in extrapolating from a database that had many more male candidates than female.

Preparing citizens for AI

When a technology becomes available, it does not necessarily follow that it will become widely adopted. Public opinion has a role to play, and a key issue is to ensure that citizens are informed of, and consent to, points at which AI is used to make decisions about them.

Demos, in a survey of 2,000 adults in Britain in 2017, found that 35% thought there was “a risk [to their current jobs] from future developments in AI and automation,” compared with 53% who believed there was no risk. More than half (54%) felt that technological benefits will not be shared evenly across society, with this figure rising for those aged 55 or more.

Those who are most likely to gain – or suffer least loss – from AI are those who either have the skills to work directly with artificial intelligence systems, or those who are able to use AI to augment their own skilled work. This in turn points to a role for the UK education system; a House of Lords report indicates the need for a system that “prepares children for life with AI and for a labour market whose needs may well be unpredictable.”

AI can perform at a level that vastly supersedes even the greatest human talents. The opportunities of AI appear compelling. However, given the lack of expertise in AI that many local authorities have, the difficulty in differentiating between AI reality and hype, the need to keep data security carefully under control, the unclear legal framework, and the concerns over AI that some staff and members of the public may well have, this indicates that a careful experimental approach to implementing AI may well be the most suitable route for many local authorities.

This implies a process of prototyping, building up skills, learning from others, and potentially working with other local authorities. It will also require an approach that openly engages staff and customers, strikes the right balance between meeting cost pressures, and improves the outcomes for local communities.

Enjoying PSE? Subscribe here to receive our weekly news updates or click here to receive a copy of the magazine!