In short, the “hallucinations” and biases in generative AI outputs result from the nature of their training data, the tools’ design focus on pattern-based content generation, and the inherent limitations of AI technology. Acknowledging and addressing these challenges will be essential as generative AI systems become more integrated into decision-making processes across various sectors.

Problems with bias in AI systems predate generative AI tools. For example, in the Gender Shades project, Buolamwini (2017) tested AI-based commercial gender classification systems and found significant disparities in accuracy across different genders and skin types. These systems performed better on male and lighter-skinned faces than others. The largest disparity was found in darker-skinned females, where error rates were notably high.

When conceived of in this manner, implicit bias is a normal behavioral phenomenon: It happens to everyone all of the time. From a moral point of view, however, implicit social bias is a highly controversial phenomenon. Many of us do not want to be implicitly biased, that is, we often find it undesirable to be influenced by social cues, such as when we try to hire the best person for the job.

You probably have the impression that line B is longer than line A, but in reality, both lines are equally long. What happens is that you are influenced by the arrows at the end of the lines even though you do not pay attention to the arrows or might even have the conscious goal not to be influenced by the arrows.