I would recommend more critical thinking as we go. Every step we could be asking ourselves about the data on which a decision is being made. If it's "AI told us to do this," we might want to talk about how we know or how much risk we're willing to accept.

Early, proper usability testing would help here too. Don't just look for "preference" or "did they make it through the prototype." Give tasks that will show a good Researcher if we really solved the right problem, and solved it well. Did we create new problems?

I just realized that I never published my article on signs during a project and after its launch that you are working from bad or not enough research. I think all of those would apply here. Some of those signs included:

* We don't really know much so we fight over opinions. I like the button here. I think it should say this. I think users might want this.

* We reverse engineer use cases. We have an idea or solution. We don't know if anybody will need or want it. But we make up possible reasons they might. Good early research would help us know earlier what problems we can solve... versus solutions in search of problems.