One of the most pressing concerns is whether AI can genuinely reflect the diversity of the audiences it aims to engage. AI models learn from data, and the quality of the data they’re trained on directly impacts the content they produce. If the training data lacks diversity whether in terms of race, gender, culture, or socioeconomic background—then the content generated by these models is likely to reflect those gaps, potentially perpetuating stereotypes or excluding significant portions of the audience.
I am currently testing this via Midjourney when producing IF IT APPLIES design concepts by creating hyper-realistic diverse models. This campaign focuses on a concept skincare brand called “Lather” aimed at Gen Z & Young Millennials; they are active on social media and prefer brands that align with their values such as sustainability, inclusivity, and transparency.
My aim with these models was to show deeper skin tones with different faces, with a select group of prompts to see if there were any hidden biases with the AI data.
What do you think these prompts were? Would love to hear your thoughts to see if it successfully reflects in this campaign.
-Creator Stephanie Taitte
