Lanna Hill: AI-generated content is perpetuating harmful stereotypes

Everywhere we look, businesses, brands, and professions are incorporating AI to improve efficiency, and it is fast becoming accepted by many marketing and branding agencies across Australia and overseas.

Yet there are many significant problems with the use of generative AI in our agencies, such as a lack of representation among the non-binary and gender-queer programmers training AI, resulting in a similar lack of representation from this community in AI-generated work.

Similarly, gender stereotypes are being emulated in AI-generated art, and we know that deep fakes have long been used predominantly against women.

Discrimination is also a real problem, and when we consider the data being crawled by generative AI, it’s not hard to determine why.

The world is grappling with significant cultural, racial and gender biases, and the widespread use of AI looks to be further perpetrating these issues with algorithmic bias, just as we are actively trying to encourage more diversity, equality and inclusivity in our communities and workplaces.

The same issues arise when companies choose to use stock images, which (unless selected intentionally and in line with DEI goals) also perpetrate both gender and racism stereotypes, confirmed by a 2020 study by Professor Fernanda Carrera. Representations of disability are also typically narrow, inaccurate and non-inclusive.

From a more simplistic, philosophical standpoint, one of the main reasons why a business will seek out a marketing or branding agency or consultant is because they need and want someone with a higher level of expertise to tell their stories, and present their brand in the best light — with an authentic tone of voice, a knowledge of their market, their strengths and weaknesses, and their objectives.

This human touch is integral to the marketing process. Generative AI, for all of its capabilities, just can’t do that. Utilising tools like ChatGPT can also have a negative impact on SEO, as search engines prioritise unique and valuable content.

In contrast, text created by ChatGPT relies on existing information that may not even be accurate and carries inherent and significant risks when it comes to copyright.

If agencies are using ChatGPT or similar (over and above the humble spelling or grammar check) to produce copy for client strategies, social media posts, or images for use on websites or other marketing collateral, they should declare it upfront.

And while it’s not always discernible, more often than not, it is quite easy to tell when someone has used ChatGPT to write an email, a media release, or a website copy.

Call me old-fashioned, but if we rely less on our abilities to hone our writing and storytelling skills, vocabulary, and ability to produce unique, creative, and engaging work, where will that leave us in a decade?

When our country’s literacy rates are dropping, and verbal, non-verbal and written communication skills are sure to become more important than they already are, why are we looking for ways to de-humanise these processes?

Why are we leaving the decisions around inclusive language, diverse imagery, or unique and interesting prose to artificial intelligence and expecting better outcomes than we can produce ourselves?

We need to draw a restrictive line in the sand as to when and how these technologies can be used — for the greater good.

Lanna Hill is a strategist, speaker and founder of Leverage Media