Bias

Bias means systematic distortions in data, methods or results that favour particular perspectives or groups and can impair quality and fair treatment.

Past decisions shape future priorities

A customer-prioritisation model might learn from previous deals. If some customer groups received little sales attention, there may have been few opportunities to observe their actual demand. The model may interpret this missing evidence as low potential. If fewer resources are allocated again, the next dataset reinforces the original assumption. This is a plausible feedback mechanism: recorded revenue reflects past coverage and selection as well as customer demand. It does not automatically describe the full market opportunity.

Examine bias through its decisions

A useful assessment separates data coverage from the consequences of incorrect prioritisation. Which industries or company sizes are repeatedly deprioritised? Do the selected criteria have a defensible relationship to the task? An evaluation should examine relevant subgroups and consider alternative explanations. Small samples require cautious conclusions. Deliberately selected additional cases can help address missing evidence. This makes it easier to determine whether the system recognises a valuable opportunity or mainly reproduces historical resource allocation. One aggregate accuracy or fairness score rarely settles that question on its own.

Sources

• What is AI bias?