Understand bias in AI

Because an AI model is trained on a data set ​to recognize patterns and perform tasks, ​the model is only as good as the data it receives.

The output from the AI tool may be affected ​by both systemic bias and data bias.

Systemic bias is a tendency upheld by institutions ​that favors or disadvantages certain outcomes or groups. ​

  • Systemic bias exists within societal systems ​like healthcare, law, education, politics, and more. 
  • ​Even if the people who design ​and train an AI model think they’re using high-quality data, ​the data may already be biased ​because humans are influenced by systemic biases

Data bias is a circumstance in which systemic errors ​or prejudices lead to unfair or inaccurate information, ​resulting in biased outputs. 

  • ​Maybe you’re developing a work presentation, ​you ask an AI image generator to create a photo of a CEO. All of the images generated appear to be white males. ​Based on this result, ​you might assume that all CEOs are white men. ​Obviously, this data is biased.
  • Therefore, the more the data represents ​a wider variety of people, ​the more inclusive the outcome ​of the image generation will be

Just as AI models reflect the biases ​of the data used to train them, ​they also reflect the values of the people who design them. ​In other words, AI models are value-laden.