Machine Learning Ethics & Safety Intermediate 2 min read

What is bias?

A systematic error or distortion in a machine learning model's results

bias explained in plain English

Bias in AI refers to the unfair or prejudiced results produced by a machine learning model, often due to the data it was trained on or the way it was designed

Analogy

Think of bias like a pair of tinted glasses - if you're wearing glasses with a red tint, everything you see will have a red hue, even if it's not actually red. Similarly, a biased model will produce results that are 'tinted' by its own prejudices

Example

For example, a facial recognition system that is biased towards white faces may not work as well for people with darker skin tones, leading to inaccurate results

How is bias used?

Bias can be used intentionally or unintentionally in AI models, and it's often used to describe the unfair outcomes that can result from these biases

Common misconceptions about bias

One common misconception is that bias is always intentional - in reality, bias can often be unintentional and a result of the data or design of the model

History

The concept of bias in AI has been around since the early days of machine learning, but it has become more prominent in recent years as AI models have become more widespread and powerful

Also known as

prejudice distortion error skew