Implicit Bias ist eine automatische Assoziation oder Annahme, die Urteile ohne bewusste Absicht beeinflussen kann. In der KI-Entwicklung kann er Datenerhebung, Feature-Auswahl, Labeling, Tests und Systemdesign beeinflussen und zu verzerrten Ergebnissen oder unfairer Leistung führen.
Automatically making an association or assumption based on one's mind models and memories. Implicit bias can affect the following: - How data is collected and classified. - How machine learning systems are designed and developed. For example, when building a classification model to identify wedding photos, an engineer may use the presence of a white dress in a photo as a feature. However, white dresses have been customary only during certain eras and in certain cultures. See also confirmation bias.