AI 會否因訓練資料或重複提示詞而產生確認偏誤？

Suppose that you are told something [a piece of information or a statement] again and again. Now you start not just believing it, but thinking it's the ultimate truth, and that nothing can contradict or oppose it.

Now, every time something occurs that contradicts the statement, you start finding points based on nothing just to justify that the statement is true and that contradiction is not a part to be considered.

This is Confirmation Bias.

You know something. It can be true in some cases but can be false in some cases, but for you it's the ultimate truth. Your brain (human brain) doesn't have the capacity to take in new information that actually contradicts your old information. Now you start finding interpretations and explanations that actually support your statement or decision.

It is the tendency to search for, interpret, favor, and recall information in a way that confirms or supports one's prior beliefs, values, or decisions.

Now, the main point of considering this.

Our neural networks are built specifically so that

• We try to prove ourselves and our thoughts correct in every situation.
• Things (informations/decisions/any thought) that get repeated stay in our brain like forever. The more they are repeated, the more we believe in them.

So, I had two things in mind while studying this.

• Can AI have confirmation bias based on the dataset/database it is fed?
• Can we build confirmation bias in an AI by telling it something again and again? (ofc for personal use case cuz AI's actual dataset can't be changed according to our interaction)