Should we be worried about Artificial Intelligence in regulation?
11 October 2021
● News and media
Big data and technology are reshaping the world of regulation with Artificial intelligence, machine learning, improved data collection and analytics, remote sensing and satellite imagery tools which can change the relationship between regulators and their communities.
The technologies available to regulators have transformed the technical side of regulation and opened new possibilities that are simultaneously exhilarating and terrifying.
An ANZSOG National Regulators Community of Practice webinar, RegTech: the good, the bad and the seriously scary examined how technology could and should be used, and the responsibility of human regulators to balance the opportunities of technology with an understanding of their mission.
The panel was chaired by UNSW Business School Professor Rob Nicholls and consisted of Anne Lenz, head of the Compliance and Regulatory Services branch at Brisbane City Council, Katie Miller, General Counsel and National Manager, Legal and Enforcement at AUSTRAC, and Monika Sarder, former Senior Strategic Data Analyst at InSight.
The panel covered the practical benefits of technology in assisting regulators, as well as issues around what decisions should and should not be made by technology and how to ensure transparency.
Ms Miller said that RegTech created a range of legal and ethical issues of which regulators might not be aware.
“The first issue is to know your project and work out the ‘why?’ of what you are doing and who is affected. Look at where you are applying the technology, because doing it at the decision-making stage can be tricky for a lot of governments. You need to be careful of issues of discrimination and bias, and when you are ready to let go of human made decisions and move to tech-made ones.
“I would advise anyone in regulation thinking about a new technological tool to get legal advice, not just about the data collection issues, but to recognise that our laws were written with human public servants in mind.
“The third thing is transparency – this stuff is challenging for the community and it sounds pretty scary if they don’t understand the full picture. When they don’t, they fill it in with their own information and that is usually not good.”
Ms Lenz said that machine learning and AI tools were vital aids to the Council, which has a large, regulated community and needed RegTech to look at how it could prioritise resources.
“At the most basic level our IT systems were old and clunky and led to officers waiting in cars for things to upload. Beyond that we have algorithms that predict compliance, and which allow us to prioritise activities. RegTech enables us to separate wheat from chaff, for example we can look at which properties in Brisbane have pools.
“One of our regulatory responsibilities is barking dogs. In the past someone needed to listen to an audio recording to ascertain how and when a dog was barking. Now, an algorithm can separate that noise from background noise and send it to the officer to check. It’s a huge time saving and means that we’ve halved barking dog complaints by reacting early.”
“This sort of technology is a tool of the trade. The practice is the same as far as a member of the community experiences it, it’s only our internal processes that are changing.
“We will never use technology to make decisions, it’s used to help officers to do their job in a more efficient way.”
Ms Sarder said that issues around discrimination were still being worked out, and that over-reliance on quantity of data over quality could lead to problems for regulators.
“We have a form of data fundamentalism where big data is confused with good or representative data. We need to ensure that the data we use works as well in one cohort as another.
“Regulators should have a good statistician on staff and ensure that there is someone who can contribute on representative data.”
Building new regulatory technology and managing risk
Regulators looking to use RegTech in new situations should seek to pilot first, by finding some ‘use cases’ to test systems and allow staff to learn the skills needed to manage new technology.
Ms Lenz said regulators should look at existing sources of data – such as imagery recorded by buses or garbage truck moving around and what could be done with it.
She said that Brisbane Council had recruited specialist staff to support its use of technology, including a data scientist.
“The important thing is to recognise that it is part of the toolkit – what do the operational areas need to know, it’s no different to any other risk you need to know what might go wrong and what the checks and balances are.”
Ms Miller said that regulators needed to think about the “techno-administrative challenge”, in which situations could technology help them make quicker decisions, what challenges did it raise.
“I think Ombudsmen will be increasingly scrutinising the use of technology. Regulators who are statutory need to have their Boards manage some of these risks,” she said.
With many sources for regulatory tools and algorithms, regulators are working through issues of whether to buy off-the-shelf products or invest in building their own technology.
Ms Miller said that off the shelf could be a useful way of seeing what is possible, but sometimes a small “bake-your-own” can be a useful proof of concept and convince people of the benefits of technology.
Ms Sarder said that there were issues with proprietary technology that did not give regulators or citizens the information to understand how decisions were being made.
“I’m a big fan of open systems, and there is a really rich community of people developing open-source tools,” she said.
“Off-the-shelf can be OK but is that vendor going to be transparent with you so that you can be transparent with the community? What happens when citizens want to challenge decisions and ask why they were made?”
For more information on regulation by Brisbane City Council, read Anne Lenz’s editorial for the NRCoP which examines the importance of regulatory tools and ‘tech-smiths’ to modern regulators.