In my Future of Marketing class at the Schulich School of Business – York University, students create AI-generated images for public service campaigns.
One student, Erika Wassilko, created a visual supporting bee conservation for Pollinator Partnership Canada. They are a wonderful organization dedicated to protecting pollinators and their ecosystems.
I loved her image. The heart-shaped honeycomb immediately captured my attention, and the message was memorable. I gave Erika an A+.
I encouraged her to submit it for possible use on the organization’s social media channels. We sent it off, and within minutes, we received a response that caught us completely off guard!
Dr. Victoria Wojcik, Director of Pollinator Partnership Canada, called it “the biggest transgression or unforgivable sin in the bee conservation world in North America.”
What the heck?
On a subsequent Zoom call, Dr. Wojcik kindly explained that the image reflected a serious underlying bias. Can you identify what it was?
The Problem: An Abundance of Honey Bee Data
The answer is that the picture exclusively portrayed western honey bees (Apis mellifera), honeycomb, and the cultural symbols most people associate with beekeeping.
In North America, however, honey bees are not a native wild species, and they’re not endangered. They were introduced from Europe and are managed largely as agricultural livestock.
Honey bees play a key role in agriculture and food production. However, a conservation campaign that focuses on honey bees can unintentionally draw attention away from other native bee species that may face more serious ecological threats.
Erika’s image was biased because it reinforced the common belief that protecting pollinators primarily means protecting managed honey bees. There are over 20,000 bee species, and almost 30% are at risk of extinction.
The western honey bee is not on the list.
The AI model put honey bees in the image because that species appears most frequently in the training data. The internet is absolutely filled with stories, pictures, songs, children’s books, folklore, movies/videos, product packaging, etc., all about the honey bee. So, it is no wonder they appear as the default bee in AI images.
Test it yourself. Ask your favourite AI image generator to create a visual of bees and see what image comes up.
If the image features a small bee with a golden brown body and dark brown or black bands across its abdomen and translucent wings … congratulations! You’ve generated an image of a honey bee.
The Lesson for Marketers
AI-generated images are now widely used in advertising and promotional materials. This real-world case demonstrates why marketers must evaluate visuals for subtle, unintended biases.
Beginning this fall, students in my Future of Marketing class will be required to submit their AI-generated images for a systematic bias review before using them in their campaigns.
Here’s a prompt you can use to identify hidden bias in your own images:
“Analyze this image for hidden bias, stereotypes, exclusions, or misleading assumptions. Examine what is emphasized, what is absent, whose perspective is treated as normal, and whether the image oversimplifies a complex issue. Consider how subject-matter experts, affected communities, or people from different cultural backgrounds might interpret it differently.”
When I consult with companies I have also begun advising them to examine every image closely for hidden bias. By stepping back and looking at the image from a critical lens, they may be able to identify and adjust areas of concern.
Lastly, my thanks to Dr. Victoria Wojcik for educating Erika and me on bee bias. I encourage you to visit Pollinator Partnership Canada, to learn more about this wonderful organization and the work they do.
This post was written by M. David Rice, Executive Director, Future of Marketing Institute and an Associate Professor at the York University Schulich School of Business.
This post was originally published in the FMI Newsletter on LinkedIn.