The UK’s Advertising Standards Authority is increasingly turning to artificial intelligence to help its experts monitor the torrent of advertising that consumers are now faced with online.
Since 2021, the ASA has been scaling up its Active Ad Monitoring system, which captures ads by relevant advertisers from a range of social media platforms, and applies machine learning algorithms to identify and flag likely non-compliant ads, which are then sent to experts to review and act on.
“There are millions of ads live at any given time in the UK, and to be an effective regulator you need to understand what consumers are seeing,” explains the ASA’s head of data science, Adam Davison.
“There’s a limit to how much a person can deal with the scale of advertising we’re now faced with as regulators,” continues Davison, in an exclusive interview with HBW Insight. “On the other hand, computers aren’t good at dealing with nuance.”
“When you put the two together, you create a system that's more effective than either, which can deal with the scale but retains that ability to think through the detail and understand the nuance that you get from expertise.”
Consumer Health Focus
The system now processes more than two million ads a month, Davison reveals. “So it's starting to become really significant in terms of data volume. It’s still very much a work in progress, but it has scaled up a lot in the last couple of years.”
The Industry View
The UK consumer healthcare industry association, PAGB, which provides advice to companies on advertising compliance, told HBW Insight that it welcomes the ASA’s new Active Ad system.
“It’s great to see the ASA using AI technology to crack down on misleading advertising for food supplements online,” comments PAGB’s director of Advertising Policy and Operations Laura Kelly. “At PAGB we support our members to ensure their advertising is responsible and follows all the relevant regulations, but social media can be seen as a route for unscrupulous companies to make problematic claims. The use of the Active Ad Monitoring system to proactively search for online ads that might break the rules is a step in the right direction towards creating a level playing field for advertising and ensuring consumer trust in consumer healthcare products.”
Commenting on the recent application of the Active Ad system to menopause products, Kelly adds that “advertisers need to ensure they do not make medicinal claims for unlicensed products, whether explicitly as in this recent case, or implied.”
“Although menopause itself is not considered a medical condition, any claims about treating or managing its symptoms are considered medicinal/medical, which can make it difficult to advertise food supplements for women going through menopause compliantly. Whilst this case demonstrates a clear disregard for the rules, even conscientious advertisers should take care when navigating complex topics like this.”
This process of “scaling up” has involved focusing on key areas, and gradually expanding outwards. Consumer health has been one of these key problem areas, particularly food supplements that make unauthorized medical or health claims.
“Earlier this year, for the first time, we started to do very broad searches of ads that make medical or health related claims,” Davison explains. “And this raised a selection of issues that weren't necessarily prevalent in the complaints we received from the public, but when we looked at them, we thought they were significant areas that needed more attention.”
This has resulted in a number of rulings in the last two months against ads for supplements claiming to treat anxiety and stress, menopause, autism/ADHD and weight loss. (Also see "Infographic: Mental Health, Menopause And Weight Loss Ads Targeted By ASA's AI Monitoring System" - HBW Insight, 6 August, 2024.)
Pre-Emptive Regulation
When deciding where to focus its attention within consumer health, the ASA chose to prioritize the protection of vulnerable groups, Davison continues. “We think these are ads that are not only problematic, but also in some cases represent significant potential for harm.”
The fact that the consumers targeted by these kinds of ads are vulnerable is part of the justification for using the Active Ad Monitoring system, Davison suggests, which represents a further step on the ASA’s journey towards pre-emptive regulation.
“The ASA for a long time now has been in the process of being more proactive, doing more project-based work, tackling some of those issues that we think are really high priority,” Davison notes.
“Complaints remain really important to us as an organization,” he insists. “You get a really good sense from complaints about what's really impacting people's lives.”
“But complaints aren't universal. There are things that complaints miss as well, especially in a world of very personalized, very targeted online ads.”
“Maybe some of the most vulnerable people wouldn't be the kind people who would know to come and complain to the ASA,” he adds. “So, we think that it's important that we supplement that with other sources of intelligence.”
Compliance Back-Up
Complaints are also only the most visible part of the ASA’s output, Davison points out, and require a lot more work than other aspects of its day-to-day regulation.
“Formal rulings are a very robust process that includes lots of opportunity for the advertiser to provide their own evidence as to why they think their claims, for example, aren't misleading,” he says. “They also get judged by our counsel, which is a group of representative individuals.”
“But obviously, this is a very heavyweight process, and if we put every ad that we saw that was problematic through this process it would be very challenging to regulate effectively.”
The formal rulings process, then, is reserved for “particularly challenging” or “particularly interesting” cases, Davison notes. The rest go through the “less visible but no less important” work of the ASA’s compliance team.
Similar to legal precedent, once rulings are established, if other advertisers run similar ones that break the same rules the compliance team can work through “various channels” to make sure those ads get taken down as well.
“It's about proportionality as well,” Davison adds. Before the ASA’s Active Ad system, there was no real way of knowing whether a complaint represented a trend, or just a one-off issue involving one or two companies.
The new system, by contrast, allows the regulator to tell “all sides of the story,” he says. It enables the ASA to “focus very rapidly” on worrying trends but also equally to “tell the positive story that “the vast majority in most sectors is behaving well.”
“And, you know, that consumers should trust their advertising. So it cuts both ways. It has a variety of uses, I think, for us in enhancing the system.”
Future Plans
As for the future, Davison says that the AI system will continue to become core to the way that the ASA regulates.
This means not just adding more data – Davison predicts that somewhere in the region of 20-30 million ads will go through the system this year – and increasing the number of issues it is used for, but also “integrating it more deeply into how the organization works.”
“I think that's essentially the direction for us, to make it core to the way we regulate,” Davison reflects. “We want to be world-leading with this tech, because we think it will lead to world-leading regulation, better protection for consumers, a level playing field for advertisers and better trust in advertising overall.”
Consumer health will continue to be a “high priority area,” Davison predicts, because of the potential for consumer harm.
