Bias Machine Learning / AI Ethics & Fairness
noun
Definition 1: A bias parameter or offset term in a linear or neural model that shifts the output independently of the input features [Machine Learning Glossary].
Example in context: “In neural networks, the output of a neuron is typically determined by the weighted sum of its inputs, which includes a bias term.” [Akhtar et al. 2024]
Definition 2: Statistical, sampling, or social distortion in data or models that causes systematic deviation from neutrality, representativeness, or fairness [Bakanach 2023/2024].
Example in context: “Despite this success, these models can learn, perpetuate, and amplify harmful social biases.” [Gallegos et al. 2024]
Synonym: offset term (sense 1)
Related terms: bias term, intercept, sampling bias, data bias, algorithmic bias, social bias, fairness