Types of bias

In machine learning, there are generally five categories of bias that warrant attention. Although the list provided isn't exhaustive, these categories represent the most prevalent types of bias, each of which can be further subdivided.

Easy to identify bias

Some types of bias can be easy to identify using active monitoring and by conducting analysis. These include the following.

Reporting bias

This type of bias occurs when the data producers, data annotators, or data capturers miss out on important elements, which results in data not being representative of the real world. For instance, a healthcare business might be interested in patients’ sentiments toward a health program; however, the data annotators may decide to focus on negative and positive sentiments, and sentiments that were neutral may be underrepresented. A model trained on such data will be good at identifying positive and negative sentiments but may fail to accurately predict neutral...