Bias
Also: algorithmic bias, biased
A pattern of skewed or unfair results. It can come from the data, the model, or the way a tool is used.
The fuller definition
In AI, bias is systematic error or unfairness in outputs that reflects patterns in training data, labels, or design choices. The word also has a narrower meaning as a learned parameter inside a network.
- ExamplesInputs paired with the expected output.
- PredictionThe model produces an output.
- LossHow far off it was, as a number.
- UpdateParameters nudged to reduce the loss; repeat.
Training ends at the knowledge cutoff. Whatever the examples contained, including their gaps and biases, is what the model learned.
A concrete example
A screening tool trained on past decisions reproduces historical patterns against a group.
The distinction that matters
Bias in this sense is measured against a defined standard. Detecting it needs evaluation on representative cases, not assurance from a vendor.
Related terms
Dataset
A collection of examples used to train or test a model.
Reliability & securityEvaluation (eval)
Testing an AI tool on a set of tasks to see how well it actually does.
Training & generationAlignment
Work to make a model act as people intend. For example, it should follow rules and avoid harm. This is a goal, not a promise.
Where it is taught
How models learn from examplesSource and further reading
Anthropic · Evaluating AI agentsThe definition and example are written for this library. The link gives background, official documentation or legal authority as labelled. Legal scope depends on the jurisdiction and current text; product features depend on release and settings.