unconscious bias

Terms from Artificial Intelligence: humans at the heart of algorithms

The glossary is being gradually proof checked, but currently has many typos and misspellings.

Unconscious bias rerefrs to the way that as humans we make all sorts of assumptions about people based on class, ethnicity, gender or other factors such as how they speak or where they studied. This is opposed to more conscious bias, such as explciitly not liking people from a certain country or town based on historic enmity.

Unconscious bias may be based on cultural attitudes that we have implicitly learnt (from media, family or friends) or personal experiences (for example, having had a bad experience with someone of a particular religion). The boundary between learning and bias is thus often unclear, but we make choices as individuals and societies as to which are acceptable (e.g. refusing to serve someone acting agressively) or not (e.g. doing the same based on the colour of their skin). When a machine learning system is traned using previous human decisions or behaviours, it may learn from human bias; however this is not the only source of bias in computer systems.

Used in Chap. 18: page 281