A response bias, a systematic deviation from a neutral response (the animal reports its "true" hearing threshold), is usually handled in most studies by designing the reinforcement scheme in such a fashion, that animals are only reinforced with the same food item and quantity for both true positives (the animal reports it heard the sound while a sound is played) and true negatives (the animal reports it did not hear a sound, while no sound is played). If, however, correct rejections (true negatives) are enforced with double amount of food, the experimenter can train the animal to only report well hearable sounds. If correct reactions to played sounds are enforced favourably for the animal, it will listen closely to the sound and rather accept a false detection (false positive), than reporting a false non-detection (false negative). These systematic deviations can be problematic when comparing audiogram data. At the moment the audiogram database does not report on such differences, but we have created a GitHub repository and issue where these questions can be discussed and will lead to a change in the Audiogram Metadata Scheme (AMS).
A response bias is not the same as a cognitive bias.