Realism theory is the belief that many or most cognitive biases are not "errors", but instead logical and practical reasoning methods of dealing with the "real world".
Inherent in it is the assumption that subjects include far more information than cognitive experimenters want them to in their thought processes.
The practical information people use in their reasoning process includes (but is not limited to):
For example, when a research scientist offers an experimental subject $50 today or $100 a year from now, the subject is not just making a decision between $50 now and $100 in one year, but is instead also considering the possibility that the researcher will die, go broke, change his mind, or may simply be lying. They will also remember an article they read on the internet about how people use money to bait cons, without remembering the specifics of the article.
As such, the $50 today is far more valuable than the $100 tomorrow.
It can also explain things, such as the pseudocertainty effect. When making use of the pseudocertainty effect, how you frame a question determines the answer. But if you consider that human minds are actually considering how something is framed as valid evidence, then it ceases to be a bias and instead becomes a practical method of selecting the correct answer, assuming no deception. This practical reasoning works by treating the question framer's clear bias as valid input. Of course, if the question asker is intentionally attempting to trick or deceive the subject, and the subject is not aware of this, it will cause problems. But that is a result of intentional deception, not a cognitive bias.