Recent advances in machine learning and artificial intelligence (AI) have the potential to transform the way we regulate ads. Computers are now increasingly able to interpret images and text, and complete complex tasks. We believe harnessing these technologies will allow us to deliver effective regulation at greater scale and speed. As Head of Data Science, my focus is building the capability the ASA needs to achieve this important step-change. Our new and growing Data Science team has exciting plans ahead, and is already making an impact today.
Digital advertising brings new challenges for regulation, one of which is scale. People in the UK now consume content online across an ever-growing list of social media platforms, apps and websites. They also increasingly see targeted content, based on things like their location and interests. These trends mean that there are more ads than ever for the ASA to regulate.
Our regulation is increasingly proactive, but in the digital world staying on top of an issue might mean monitoring tens of thousands of ads, by hundreds of companies across dozens of platforms. And then doing it again the following week to see if anything has changed. The scale of digital advertising means it simply isn’t possible to do this comprehensively without taking advantage of machine learning and automation.
Today when we want to monitor a set of ads to make sure they follow the rules, this starts a collaboration between our data science and compliance experts. The Data Science team is developing tools that can capture ads from a range of sources and use machine learning tools to make sense of them. These tools can do things like search for potentially misleading text or identify ads that are visually similar to problematic ads we’ve seen before. This means that instead of searching manually, we can let the computer sift through big piles of ads and only put the most relevant ones (those most likely to be in breach of the rules) in front of the experts for their analysis.
We’re already applying these tools to high-priority topics such as influencer marketing, environmental claims and cryptocurrency ads. By helping to spot the most relevant content we are turning the task of monitoring online ads into a simpler one, allowing our compliance colleagues to review a handful, rather than thousands of ads. For example, we are now monitoring the ads produced by dozens of cryptocurrency companies daily, regularly sharing any potentially non-compliant ads with compliance experts. In each of these areas and beyond this is allowing us to find more problematic ads and act on them more quickly.
Of course, the scale of digital advertising also means the ASA handles large numbers of complaints – more than 40k in 2021. We are also exploring how machine learning can help us streamline this process without compromising our high standards of service. Decisions relating to complaints are often subtle, requiring a detailed understanding of interpretation and context, not traditionally strengths of AI systems. However, we strongly believe that there are ways for AI to assist our experts in reaching the right decisions more quickly.
When thinking about standards it’s also important we are mindful of the potential risks that come with automation. Humans often have an innate judgment of what’s right or sensitive in a way that computers do not. Fairness, transparency and consistency are all core principles of our regulation, and we will ensure that our introduction of AI-based systems upholds the same values.
We still have a long way to go in this journey, but already we’re seeing how integrating machine learning and AI into our processes can deliver value. We’re continuing to invest in this area, growing our team, and we will be sharing more examples of the way our work is changing ad regulation in the UK.
More on
-
Adam Davison
Head of Data Science
-
Keep up to date
Sign up to our rulings, newsletters and emargoed access for press. Subscribe now.
Comments are closed.