Bias in AI
Bias is when a model’s predictions are skewed and wrongly favor certain things or people.
This is a big problem since it brings unfairness into AI producing prejudiced results.
Bias can be embedded in the society around us and in the values and beliefs, we have when making a decision. It is critical to be aware of the common bias types and learn how to prevent them before putting AI models into production.
Read more about common bias types
Bias can occur in different forms and can be created during all steps of the machine learning pipeline;.
Example: When does bias occur?
Let’s make a simple test here:
What’s in 1st image?
What do you see in the image below? Please, write them down on a paper with a maximum of a few words.
What’s in 2nd image?
Now let’s have a look at this image and write down what you see in the image below?
What’s in 3d image?
Let’s repeat the same task for the next image.
What’s in 4th image?
Let’s do it one last time, and write down what you see in the image.
We asked these questions to a few of our friends and they told us what they see.
Our friends replied:
First image: “a truck”, “a fast truck”
Second image: “three trucks”, “trucks in a parking lot”
Third image: “lots of trucks”, “trucks in a parking lot”
Fourth image: “a food truck”, “a pink street food truck”
As a result, while nobody mentioned the color of a white truck, many people mentioned that it is a pink food truck. Bias has been introduced.