AI marketing campaigns risk missing cultural nuances and authenticity

You can also listen to this podcast on iono.fm here.

CIARAN RYAN: Artificial intelligence is changing the way companies market themselves. It can generate advertising copy, it can translate campaigns into multiple languages, create images and personalise messages for millions of customers almost instantly.

But what happens when AI gets the culture wrong? A message can be grammatically perfect and still completely miss the mark. Humour, idioms, names, imagery, even seemingly innocuous cultural references can carry meanings that an AI system may simply not understand.

And in Africa there’s an additional concern: AI models trained largely on global data sets may reproduce stereotypes or impose Western assumptions on local audiences.

It is quite a fraught and problematic area. The central argument is that while AI can make marketing more efficient, it cannot replace cultural intelligence – and brands need human judgement if they want their communication to be authentic and trusted.

To discuss this we’re joined by Silas Matlala from the marketing faculty at Gibs, the Gordon Institute of Business Science. Hi Silas, thanks very much for your time.

Give us an example of what it looks like when AI gets culture wrong. What are the kinds of mistakes that brands are already making when they rely too heavily on AI-generated marketing?

SILAS MATLALA: Thanks Ciaran, and thank you for having me. This is a subject that is very close to my heart, because we are at the intersection of marketing.

Great marketing is usually translated in storytelling that is rooted in culture, because it is through authentic storytelling that we can communicate and, most importantly, evoke emotions.

Unfortunately, emotions can either be positive or could also be negative – and that’s where the issue of artificial intelligence comes in. If we understand how artificial intelligence works, without belabouring the point it’s about using multiple sets of data to generate whatever has been prompted to give meaningful answers.

As you’ve rightly said, marketers, communicators and advertisers are using AI, artificial intelligence, to generate creative campaigns, to generate communication, to talk about brands, to talk about launches.

But the problem, to which you’ve already alluded, is to a large extent that artificial intelligence is over-representing Western societies to such an extent that African nuances, African culture, African texture run the risk of being misinterpreted or just plainly misconstrued.

For example, something as simple as the culture of umgidi.

Umgidi, for those that know, is the South African practice of becoming of age – particularly in the Xhosa culture.

The problem that happens is, once we get words like ‘South African black culture’ and ‘becoming of age’, no matter how much data points, no matter how much of the internet the AI can scrape, the nuances that AI can scrape, the nuances that go with umgidi are lost because umgidi is particularly relevant in the Xhosa culture.

And then umabo, umwabo – which is again a becoming of age – is particularly relevant in the Zulu culture. And as you would know, becoming of age in the Jewish community could be about the bar mitzvah and so forth.

So we run the risk of misinterpreting those things if we do not apply our authentic self to what artificial intelligence can generate.

CIARAN RYAN: All right. Now, is there a particular problem for African businesses here? AI systems learn from existing data. Much of the material about Africa is available online, but it can be stereotypical.

So they going to be trained on datasets about poverty and wildlife in rural Africa – sometimes almost a little cartoonish.

Does that mean AI simply reproduces those stereotypes at scale, and how does this then project onto businesses and how they project in their marketing?

SILAS MATLALA: The quick answer to that is yes, and I sincerely hope that as AI tools and AI capabilities grow and develop, we will be correctly represented. The comments that you are making – to say because of the over-representation of words like poverty, wildlife, and everything – that unfortunately is still the case. You still get all of those, and the risk therein is, first, stereotyping, which is negative by itself.

It could be about prejudgment, which is negative in itself.

And the third way is it could be a misrepresentation of the message that you want to bring.

When we do any of those things, stereotyping, prejudging and misrepresentation, the creative just loses impact; it just becomes wallpaper.

For example, if you were to go and say ‘Give me a key visual of a South African black township funeral’, the image that’s going to come out there might miss the mark, because the essence of not just a funeral but a burial occasion is particularly layered and nuanced within the African and particularly South African culture.

What makes it even more complicated is that black South Africans handle death, funerals, and burials differently from their black Nigerian counterparts. Totally different – that’s a topic by itself.

I’m afraid to say that, currently as we stand with such texture and such nuances and differences, AI is not yet able to grapple with those. And then you run the risk of misrepresentation which sometimes can even come across as disrespectful.

CIARAN RYAN: Language is an obvious danger area. AI is pretty good at literal translation, but can it understand humour, idiom, emotion, cultural context? If the AI model is too literal, it can miss some of those very, very slight nuances.

SILAS MATLALA: My view is simple: we need to use AI for a purpose. Even before marketers could adopt AI, I’ve had a particular campaign in my head – please allow me to elaborate.

There’s a campaign by Coca-Cola, and the campaign was about how we properly pronounce people’s names.

The story is that inside that, taking time to understand and to pronounce someone else’s name correctly is a sign of respect. That’s a human truth everywhere in the world.

But the way that you can creatively communicate that as a brand will be different for every country, and Coke demonstrated that by running a campaign asking people how they pronounce their names.

The way it showed up in Britain used humour. The way they did it in South American countries was different. The South African way, my word, was also very South African.

When you just look at that campaign you are saying you can never just take this campaign to another country and play it as it is, because it is so nuanced.

Therein lies the magic: the insight is universal, the human truth is universal.

But the way it gets articulated and expressed in terms of the way we talk – not only the way we talk, but the dialect and everything – AI, I’m afraid, can never replicate it to that level.

So we always need to be careful to go beyond literal translation, especially when coming to marketing creative and marketing communication.

When we want to write emails or business languages, by all means let’s do that. But when we want to represent communities, townships, suburbs, or the ‘gees’ of the country – there’s a word that we South Africans use: the South African ‘gees’ – what does that mean?

Only you and I who have a South African lived experience understand the layer, tonality, and meaning behind the word gees, the South African gees. It’s a combination of English and Afrikaans (the word for ‘soul’), but only we will understand that.

When AI comes in and says, ‘the South African soul’, it just loses the meaning. We use phrases like, ‘It’s not DNA, it’s South African’. Only we know what that means. It’s an inside joke, it’s what we know. If you took that same line and said ‘This is not England, it’s DNA’, they might look at you very oddly.

In my view there are a couple of campaigns that I’ve seen. You’ve come across this one – by Vaseline, Vaseline Petroleum Jelly, a brand by Unilever, which was named Umtanaka Gogo. You cannot translate it to ‘grandma’s kid’, you just can’t. It will lose the gravitas that it carries.

But if you put Umtanaka Gogo into a language learning model, it will come out and say ‘grandma’s kid’ – but it just loses it.

CIARAN RYAN: All right, so finally here, if you could just very quickly summarise, how serious can these mistakes be for a brand if they get it wrong, if the AI brand doesn’t get the nuance of the culture, as you’ve just described? What can the consequences be?

SILAS MATLALA: The consequences are dire. At the very least, if things are misrepresented, misconstrued, mispronounced, or even incorrectly interpreted, an average person will just switch off. It doesn’t grab attention, it doesn’t break through, and it doesn’t make people react – that’s the safer part.

The danger is when AI produces things that can come across as insulting or disrespectful.

Then you find people reacting by either going on social media, or slating the brand and you are trending for the wrong reasons.

The worst at the basic form is that people just don’t respond, it’s all paper, and like, what happened there?

And then at the other extreme people are just offended, just slating you and so on. We have seen a couple of those incidents, even though they were not AI generated, where brands or the marketers have got it very wrong, and they just get slated.

It goes beyond just being slated. There is money behind investing in communication. There is a lot of money, there are a lot of resources, there is a lot of time – which then means it’s just spend that has been misspent and wasted.

There is also reputational risk.

How do you come back from reputational risk and so forth, because right now information travels at the speed of a nanosecond? So it is very risky.

CIARAN RYAN: It’s a fascinating discussion, AI and culture, and no doubt this will have many more chapters written before it’s done.

We’re going to leave it there.

That was Silas Matlala from the marketing faculty at the Gordon Institute of Business Science. Silas, thank you very much for your time.

SILAS MATLALA: Thank you very much, and thank you for having me here. I appreciate it.

Brought to you by the Gordon Institute of Business Science (Gibs).

Moneyweb does not endorse any product or service being advertised in sponsored articles on our platform.

#marketing #campaigns #risk #missing #cultural #nuances #authenticity