Generative AI: Enhancing Creativity or Flattening Originality?

Generative AI: Enhancing Creativity or Flattening Originality?

The short and easy answer is "both," of course, or "it depends," but let's unpack a list of recent research and a dash of existential dread.

LLMs and generative AI are wildly entertaining, heavily convenient, thus increasingly inescapable. A real double-edged sword. No wonder why we’re being force-fed with them by media and giant tech advertising platforms as well (you can’t work without them in digital ads buying anymore). So, we swallow them whole. And so briefly, we have the main pillars of their success. And they keep gaining ground as we (humans) grow so accustomed to identifying AI products - I suppose many can already tell by the aesthetics of those disturbingly polished images, sounds, and maybe even texts, which are widely used for massively produced ad creatives.

It’s been more than two years since we used them broadly (it was back in late November 2022 when ChatGPT became accessible to the public), and I’m pretty sure, after reading these very important articles presented here, that the words like “homogenisation,” “formulaic,” “templated,” and “predictable” will be more and more of our concern in marketing and generally business decision making, as human creativity starts to be hijacked by LLMs and ML algorithms, which is already happening. And these AI outputs of ours have already started being associated with “cheap,” “quick,” “lazy,” “unoriginal.” A buffet of reheated leftovers made of the agony of mass production for new ideas and content.

The Homogenization of Creativity

Let’s start with the “why bother.” As Doshi & Hauser (2024) note, “Creativity is fundamental to innovation and human expression” (Doshi & Hauser, 2024). Research by both Doshi & Hauser and Anderson et al. (referencing their 2024 C&C article) arrive at very similar conclusions, and if you read between the lines, they hide very politely a very obvious and upcoming threat - generative AI tools can boost individual creativity while flattening creativity at the collective level. Writers given AI-generated story ideas produced tales rated more creative and better written – especially among less-creative writers (Doshi & Hauser, 2024). Yet, these AI-assisted stories ended up eerily similar to each other than stories written without the help of AI (Anderson et al., 2024). AI made each writer’s story better, but everyone’s stories became more formulaic and alike. The researchers caution that if the industry embraces AI-assisted writing widely, “the stories would become less unique in aggregate and more similar to each other” (Anderson et al., 2024). This is a creative social dilemma – individual gains at the risk of collective originality.

And here comes the very insightful, bold article by Robert J. Sternberg, “Do Not Worry That Generative AI May Compromise Human Creativity or Intelligence in the Future: It Already Has” where he rings the bells of our creativity muscles: “If generative AI does our creative work for us, we risk losing our creativity because we do not exercise it” (Sternberg, 2024). He compares this to the decline of foreign language skills in a world of instant translation apps: “Why learn Finnish when Google Translate does it for you?” (Sternberg, 2024). The concern is that without practice, we lose the nuance required for true understanding - or in creativity, the spark needed for connecting disparate ideas.

When Everyone Uses the Same AI

Why does AI-driven content push us into a monoculture? One issue is that large language models (LLMs) are trained on the same vast datasets, often spitting out statistically probable and familiar patterns. When multiple writers or marketers lean on the same AI questions (prompts), they get the same suggestions, so their content starts sounding homogenised. A user study by Anderson et al. (2024) found that ChatGPT users’ ideas were less distinct from one another than those of people using a different brainstorming method (Anderson et al., 2024) - the one where Brian Eno comes in.

In a collaborative setting – such as a team brainstorming with AI – people may be unconsciously fixated on the AI’s output, leading them to convergence on similar themes or wording. Anderson et al. (2024) study observed that groups who shared LLM-generated ideas experienced a “homogenisation of ideas” within the group (Anderson et al., 2024). In contrast, individuals using AI independently tend to interpret suggestions differently, injecting personal flair. Independent AI use can preserve more variation than a group rallying around the same template.

Let’s (Sternberg actually) make it a bit more scary, as he sharpens this point: “Generative AI can serve autocratic governments by reinforcing dogma... Training AI on curated datasets leads people to accept uniformity as truth” (Sternberg, 2024). The result? A creeping flattening of thought, where decision-makers - especially inexperienced youth, or those adults over-reliant on AI - risk becoming “basic” in their cognitive processes.

Even more troubling, Sternberg points out that people may falsely believe AI-generated ideas are their own original thoughts (Sternberg, 2024). Think about it: how many students (and professionals - rapidly look at your colleagues, now!) have already handed in AI-written essays or reports and taken full credit? That’s a dangerous blurring of authorship. It’s like engaging autopilot and then congratulating yourself for the smooth landing – eventually, you forget how to fly and you start believing the autopilot’s landing was your own. Over-relying on AI can lead to a kind of creative delusion and complacency, where decision-makers lose the very ingenuity and critical thinking that made them successful in the first place.

Staying Original in an LLM Era

If you are in a profession where decision making is vital (that means all of your life’s dependent on decisions, after all), you might normalize or make “cheugy” the way you think, particularly if your decision making is not strong enough yet (you’re young) or it relies too much on AI (it can be atrophied). For digital marketers and creative strategists or any other creative professionals, these findings are a reminder to stay vigilant against creative flattening. If everyone’s content starts following the same AI-driven formula, it becomes harder and harder (and harder as it keeps on being trained to the same homogenised inputs) to stand out. To maintain originality, consider a few exhausting, as of post LLMs epoch, strategies:

• Invest in deep domain knowledge. And yes, I understand, being educated by institutions becomes more and more a privilege worldwide today, and LLMs can cover this gap, but do they? Try reading books, and follow in them the human thought process - that’s gonna help your synapses more than you can imagine. Deep, firsthand knowledge gives you raw material for insights that no generic AI output can provide. The more expertise you build, the less you’ll feel the need to lean on canned answers. As Sternberg warns, “Paradigm-defying creativity is unlikely to come from AI... Big-C ideas that change the world will emanate from humans for a long time” (Sternberg, 2024).
• Cultivate curiosity and diverse thinking. As much as it’s self-assuring, convenient, narcissistic to blame groups of people (genders, ethnic groups, generations, sports teams) for their inferiority, that very particular difference is exactly what can bring evolution and innovation to the table. When diverse minds collaborate and challenge each other, they’ll hatch ideas far outside the AI-generated box. Human curiosity and a mix of perspectives spark creative leaps that an AI – trained on the average of human ideas – simply can’t replicate.
• Diversify Prompts and Approaches: Just like the parable of the tree of knowledge, it’s very hard not to taste the fruit. Taste it, but taste some other fruits as well. Don’t always ask the AI the same questions. As Anderson et al. (2024) found, be creative with prompting, just like Oblique Strategies cards, create chaos to the algorithmic patterns of LLMs, as they foster greater diversity because they “introduce randomness that LLMs inherently filter out” (Anderson et al., 2024).
• Begin Cross training, Cognitive Cross training: “Periodic ‘No-AI’ brainstorms” are essential, but frame them as creativity Cross train. Sternberg’s “use it or lose it” principle applies here: “Exercise becomes especially important with age... The same applies to mental function” (Sternberg, 2024). Try writing a campaign brief without AI, then evaluate it with colleagues. The goal isn’t to reject AI but to prevent what Sternberg calls a “reverse Flynn effect,” where creativity declines not due to circumstance but “through circumstances we could control but choose not to” (Sternberg, 2024).

Generative AI is a powerful tool – and like any tool, its value comes out from how we use it. We can opt-in to let it dull our creative instincts, or we can wield it to enhance our originality. There’s always a trap here, because it speaks like a human, but actually, we discuss in a human way with algorithmic functions and models, in the language and culture of the means, standard deviations, and p-values. Creativity and innovation require an honest connection to ourselves, and this connection builds our own creative muscles, so we can flex to our surroundings.

After all, if everyone has access to the same AI, true originality might just become the ultimate competitive advantage.

Sources:

• Doshi, A., & Hauser, J. (2024). Science. Link
• Anderson, J., et al. (2024). arXiv. Link
• Sternberg, R. J. (2024). MDPI. Link