Rihanna Spotted Heading to César Restaurant in Paris After Céline Dion Concert
By Rodrigo Santos Andrade
Humans and machines collaborate on content, but levels of involvement can vary.
tadamichi/ShutterstockAdvertising is about persuasion. It exists to sell products and lifestyles, ideas and stories.
And while people have become pretty good at treating adverts with a degree of scepticism, artificial intelligence (AI) has added another layer of uncertainty.
When we see an advert labelled “AI-generated” on social media or anywhere else online, we know that AI was involved in making it. What we usually don’t know is how.
AI may, for example, have been one tool in a largely human-led creative process, used to edit an image or refine text. Or it might have generated much of the finished advert with limited human input. A simple “AI-generated” label can encompass both extremes and everything in between.
The distinction matters because AI can change how consumers react to advertising. There is evidence that disclosing AI involvement can reduce levels of trust.
Other research points to some overall resistance to AI from consumers, although reactions vary considerably depending on what the technology is being used to create.
Different disclosure
Disclosure itself is nothing new in advertising. When influencers use “#ad” on social media, or labels pointing to “sponsored content”, disclosures help consumers recognise the commercial side of what they are watching or reading.
Our research, however, indicates that AI disclosures serve a different purpose.
For while traditional advertising disclosures tell consumers something about why content is being shown to them, AI disclosure labels instead draw attention to how it was produced. But they do not necessarily reveal the respective roles of humans and machines.
For decades, advertisers have used digital tools to alter content. A designer might retouch a photograph for example, or remove a background while retaining control over the creative decisions.
AI blurs the distinction between editing existing content and creating new content because it can generate elements of that content by itself. A person may still develop the idea and refine the result, but the balance between human and machine contribution can vary considerably.
Disclosing that AI was involved can then invite judgments about who or what created the content, how much human input was involved, and whether the use of AI was appropriate. This may also help explain criticism of AI-led campaigns by brands such as Coca-Cola and Gucci.
Overly simple
Governments and platforms are beginning to address these questions through new approaches to AI transparency.
The EU wants to make sure that providers of certain generative AI systems are subjected to strict requirements about disclosure. Meanwhile, technology is emerging that allows digital content, including adverts, to carry accessible information about its provenance and editing history.
And the detail matters. Meta acknowledged that its “Made with AI” label appeared on content involving only minor AI modifications and did not provide sufficient context. It replaced the “Made with AI” wording with “AI info” and later made the label less prominent when AI had been used to edit, rather than generate, content.
AI can help with ads, but to what level?
Stock-Asso/Shutterstock
We are also seeing the same kind of ambiguity appearing in reverse. Labels with the words “human-made” and “no AI” are emerging to signal human authorship.
But do these mean no AI whatsoever, or a predominantly human process in which AI played a supporting role? Both kinds of label can obscure as much as they reveal.
For the time being, then, consumers should take a little time to consider how they interpret AI labels. The term “AI-generated” tells us something – but not everything – about how content was produced. It does not, by itself, mean the result is fake, poor quality or untrustworthy.
Nor will it spell out the human role – who developed the idea, selected and refined the output, checked its accuracy and decided what ultimately appeared.
Ultimately, an AI label should be a starting point for evaluating content, rather than the end. “AI-generated” tells us that AI played a role, but understanding how it was used and how much human judgement shaped the result would tell us far more.
The authors do not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.
Tags: Academic Journalism, The Conversation, The Conversation Academic News