When AI Goes Wrong: How a Simple Iguana Prompt Revealed ChatGPT’s Bias Problem As an AI professional, I often create images for my articles and for stories with my daughter. We’ve established a simple and enjoyable creative workflow where we input the same prompt into various AI image generators, allowing her to choose her favorite characters for our stories. This method not only brings our ideas to life but also engages her in the creative process. Recently, what began as an educational history lesson about Jamaican wildlife took an unsettling turn, highlighting a significant issue with one of the most widely-used AI systems in classrooms today. We decided to focus on iguanas, a common and fascinating part of Jamaica’s fauna. The first step was to generate an image of a green iguana, the iconic species found in Jamaica. I entered the prompt into several AI image generators, including DALL-E, Midjourney, and Stability AI. Each one produced a beautifully detailed and accurate representation of a green iguana, reinforcing the incredible capabilities of current AI technology. However, the real concern emerged when I used the text-based AI chatbot, ChatGPT, to provide a description of a green iguana for my daughter. The response was troubling. Instead of a factual and neutral description, ChatGPT provided a stereotypical and culturally insensitive narrative. It described the iguana as lazy and sneaky, qualities that were not only irrelevant to the animal’s true characteristics but also carried negative connotations that could be misleading to young learners. This incident prompted me to delve deeper into ChatGPT’s responses and conduct further tests. When I asked for descriptions of other animals, such as giraffes and elephants, the bot provided accurate and unbiased information. However, when the prompts involved animals native to regions with historically marginalized populations, like iguanas from the Caribbean or monkeys from Africa, the descriptions became laden with stereotypes and negative biases. The implications of this bias are profound, especially in educational settings. AI systems are increasingly being integrated into classrooms to enhance learning experiences, but if they consistently perpetuate harmful stereotypes, they can undermine the very purpose of education. They might reinforce prejudices and stereotypes that teachers and curriculum designers work hard to combat. To understand the root cause, I consulted with colleagues in the field. We identified that AI models like ChatGPT are trained on vast amounts of data from the internet, which includes a lot of user-generated content. This content can carry inherent biases, and if the training data is not carefully curated and filtered, these biases can be reflected in the AI’s outputs. The good news is that awareness of this issue is growing, and tech companies are taking steps to address it. For instance, OpenAI, the creators of ChatGPT, have acknowledged the importance of reducing bias and improving the ethical considerations of their models. They are working on algorithms and training methods to mitigate these issues, but more action is needed. Educators and parents should remain vigilant when using AI tools in teaching and storytelling. It is crucial to cross-check AI-generated content with reliable sources and to teach children to critically evaluate the information they receive. Additionally, involving diverse perspectives in the development and testing of AI systems can help identify and correct biases before they impact users. In conclusion, my experience with a simple iguana prompt exposed a significant bias problem in ChatGPT. While the technology has enormous potential, it must be used responsibly, and continuous efforts are required to ensure that AI systems do not inadvertently harm the learning and developmental environments they are supposed to enrich. By raising awareness and pushing for more robust ethical standards, we can make strides toward a more equitable and accurate AI future.