I'm currently working on a system that manages big data and has a classification AI. However, the AI is not yet optimally trained, and I would like to gather some True Positive (TP) and False Positive (FP) examples to further train the AI and increase it's precision.
Since the data belongs to the users, it means I need to validate with them if the AI's predictions are correct, and let them be the judge of that. Taking that into consideration, I was trying to think of a way to ask the users for feedback regarding the data, something like "Is this a [true statement]?". However, I don't want them to think that the AI has poor performance and that the processed data cannot be trusted.
How could I ask the user for this feedback? Should I be honest and upfront? If so, how minimal can this process be? I was thinking of just an approach along the lines of the question and two buttons, meaning yes or no.