biaseslike

Biaseslike is a nonstandard umbrella label used in some interdisciplinary discussions to refer to biases that resemble each other across different systems—most often between humans and artificial intelligence—without implying a single underlying mechanism.

It functions as a heuristic for comparing patterns of bias rather than a formal theory. Proponents use biaseslike to group biases by shared features such as reliance on limited information, susceptibility to distributional shifts, or transfer of prior expectations across tasks.

In practice, biaseslike might cover examples such as data-induced omissions due to unrepresentative samples, anchoring effects in judgment, and misgeneralization from familiar to novel contexts because of shortcut heuristics.

In AI, it can help discuss how training data, model architectures, and objective functions can produce analogous bias patterns in both humans and machines.

Critics warn that biaseslike is vague and risks obscuring important differences between cognitive biases and algorithmic biases. It is not a diagnostic category and does not identify causes; it only notes patterns of similarity across domains.

See also: cognitive bias, confirmation bias, anchoring, algorithmic bias, data bias, fairness in machine learning.