Not Everything Is What It Seems: Perception, Bias, and Judgment

Family and friends' WhatsApp groups are endless sources of random content — sometimes funny, sometimes just curious, but almost always helping to make the day a bit lighter.

This week, I received the image below with the following caption:

"After two hours sitting looking at the lake, the alcohol effect was gone, and I realized it was the neighbor's wall."

Cultural note: In Brazil, it's common for brick walls to separate neighboring properties.

At first glance, it does look like there's a lake in the image — even something resembling a small boat floating in the middle. But with a closer look, we realize it's not a lake at all. It's just a wall.

Besides making me laugh, the image made me reflect on mental models and the stereotypes we carry — often unconsciously. Just as AI tools are trained on data, our experiences, environments, and the culture in which we live shape our judgments.

A relevant example: when ChatGPT became popular, I participated in activities to test the tool's biases, such as generating an image of a "CEO." Most of the time, the result followed a familiar stereotype — a white man, wearing a suit and tie, in a traditional corporate setting. Months later, when I repeated the same prompt, the system provided me with an image of a young woman in a futuristic and far less formal setting. What changed? The tool was updated with new data, and the model was adjusted — tasks we should consider with our lenses.

These lenses — whether cultural, social, or emotional — that we use to interpret the world can limit our judgment, influence decisions (sometimes negatively), and perpetuate bias. Stereotypes around gender, race, age, and culture tend to oversimplify complex realities.

In a leadership course focused on women, I participated in a powerful exercise: images of people from diverse ethnic backgrounds were shown, all dressed in executive attire. Participants were asked to guess their professions. The interpretations varied widely — revealing subtle (and sometimes uncomfortable) biases.

I came across a similar reflection in a great Microsoft Research blog article that explores how text-to-image AI models often reproduce — or even amplify — those social biases:

 🔗 Understanding Social Biases Through the Text-to-Image Generation Lens

Especially with the increasing use of generative AI and algorithmic support in decision-making, being aware of our own biases is essential. After all, not every "lake" is what it seems — and what we see with confidence may, in reality, be something entirely different.

#PerceptionMatters #UnconsciousBias #MentalModels #CriticalThinking #BiasInAI #GenerativeAI #Bias #Stereotypes