As far as I know fairness in retrieval evaluation and fairness in AI models are related concepts, but they have some important differences.
For example: Consider a healthcare search engine that is used to find relevant medical information. If the search engine returns different results for different racial or gender groups, that could be seen as unfair in retrieval evaluation.
However, if the search engine is built on an AI model which is biased against certain groups, it might also return inaccurate or harmful medical advice, which could be seen as unfair in the context of AI model fairness.
To my knowledge it refers to the principle of evaluating search engines or information retrieval systems in a way that is unbiased and equitable to all users, regardless of their demographic characteristics such as race, gender, or age.