AI Tool Speeds Government Feedback – Experts Say Exercise Caution
Now a new A.I. tool is hoping to plow through such reams of feedback, and truly comprehend what the public means to say.
“No one in the public sector should have to waste time doing something an A.I. program can do in a fraction of the time and at a fraction of the cost — and definitely not at the expense of supply teachers and teaching assistants. And in the future, no one will,” said Peter Kyle, the UK technology secretary.
“Why spend millions of taxpayers’ pounds outsourcing simple work to contractors, when technology can do it increasingly better every year?”
The digital bot – which goes by the name of Consult – has already completed an early test run with the Scottish Government.
The Scottish Government used Consult in this trial to consult the public on plans to regulate non-surgical cosmetic procedures such as lip fillers and laser hair removal. The AI’s conclusions closely mirrored those of human analysts.
The idea now is to roll out this technology across different government arms. As it stands, reviewing the public’s input is laborious and delayed – with costs for contractor review frequently in the millions.
Consult is also, however, at the heart of a wider drive to reshape the UK state so that it becomes smaller and more responsive — one that can deliver on the heavily trailed “reform agenda” without splurging too much cash and without taking too much time.
So, how did Consult do in Scotland?
It analyzed thousands of responses and deeply considered feedback from more than 2,000 respondents. Powered by generative AI, it detected the pivotal themes and concerns expressed in responses to six primary questions.
Then again, Consult wasn’t alone as an actor. These early insights were then scrutinized and developed by experts within the Scottish Government. AI then returned to the refined theme level to code individual responses. This allowed officials to think more deeply about what people were saying and how it could inform policy.
Given this was Consult’s debut live run, the Scottish Government also manually checked every single response to ensure accuracy. Besides, figuring out what someone meant and what the category of their comment is can be subjective — even when it comes to humans, who don’t always agree.
When Consult was tested alongside human analysis, it was largely accurate. To the extent that there were qualms, they were subtle ones that didn’t change the broad message of the public.
Part of a Bigger Picture
Consult is one part of a broader AI toolkit, Humphrey, a group of digital assistants aimed at liberating civil servants from mundane administrative tasks as well as cutting down on contractor costs. It’s all part of a vision to use high tech to make public services better and save £45 billion in productivity.
“It will help us to cut the costs of governing and make it easier to collect and widely review expert and public input into key issues,” Peter Kyle added.
“The Scottish Government has made a brave first move. Before long, colleagues across my department and others in Whitehall will also be using Humphrey to support us to turbocharge the reform agenda.”
In Scotland, Public Health Minister Jenni Minto said:
“This tool was hugely beneficial in helping the Scottish Government understand what people wanted us to hear and the full spectrum of views our respondents held.”
“It enabled us to home in more quickly on the key policy questions and look into the evidence further – all while making sure we were hearing the strong views people had.”
Still in the Early Stages
Like many government AI rollouts right now, it is still early days. Technically, Consult is still in testing. Additional testing and analysis will be needed before any widespread roll-out to ensure it works. But the potential is massive.
The British government holds around 500 consultations annually. If it delivers as promised, Consult could save officials the equivalent of 75,000 days of analysis each year.
Feedback from Civil Servants
So what was the verdict from the civil servants who had made use of Consult?
They were, apparently, “pleasantly surprised” – with the AI’s initial analysis a “useful starting point.” Others said that it “saved them a lot of time” and allowed them to “rapidly find, and extract what was required.”
Curiously, several claimed that Consult had given a new level of fairness.
“It takes bias out of it, makes things more consistent,” said one official, explaining how it helps avoid personal views, even unconsciously, influencing results.
And some consultations have tens of thousands, in some cases hundreds of thousands, of responses. As well as Consult has performed in early experiments, it’s probably only a matter of time before it is used for those large consultations.
Human Oversight Remains Key
And rest assured that people will not be completely eliminated from the process. Consult is meant to keep experts in the picture at every step. Officials will still review the AI’s proposed themes and how it organizes the responses. They’ll also gain entry to an interactive dashboard where they can filter and search for particular insights.
This is where AI is stepping in to take the strain – so that people can get on with the smart thinking.
A.I. Experts Sound the Alarm Over a ‘Killer Robot’ Arms Race
The push for AI in government is not happening in a vacuum, and some experts are keeping track.
Stuart Harvey, CEO of Datatix, said:
“AI-powered faster public consultation is a brilliant example of how tech can improve productivity and save cash. But AI is only as impressive as the data that underlies it.”
“For systems like these to function fairly and effectively, government departments need to make certain their data is accurate, current and well-managed.”
“We need people to trust decisions made with AI. That’s going to depend on those being rigorous, well-run and ethical processes. Bad, neglected data can result in biased or unsound conclusions.”
“Any time government is introducing AI in the use of public services, it is important to invest heavily in strong data practices — by code that is consistently built; by systems that are built to review data; and by keeping people inside of the mix, especially when it’s a business of people listening to others.”
Academics Agree
Academics share similar views.
For example, Professor Michael Rovatsos, of the University of Edinburgh, welcomed the potential of the tools, but also warned about the dangers of allowing bias to creep in — and even the potential for outright manipulation. He advocated for strong protections and continuous investment to maintain that government AI is trustworthy and fair.
“Building a Smarter State”
Stuart Munton, head of group operations, AND Digital, said:
“The government’s use of AI to accelerate the time to run public consultations is a step in the right direction for smarter, more efficient public services. But as we embrace AI, we need to make sure that humans – not just ‘tech’ – are at the center of that transformation.”
“This is the only way that such tools can reach their full potential — by investing in training public sector teams with the right skills. Giving a diverse array of talent the opportunity to use AI will not only raise results but make innovation more inclusive and rooted in real world needs.”
Looking Ahead
The ambition is to roll out Consult fully across departments by the end of 2025.
