What did you select to leave out/leave in?

This is difficult to answer really, as I work with several different types of language models, chatbots, etc. and with different uses.
There can be chatbots that will be used on gaming chats, for instance, and my job would be to filter out speech that can be hurtful or insulting etc. Or innappropriate sexual content. This way, we teach the bot to recognize this behavious and flag it.
Some might call it censorship, but in my opinion, some language just doesn't belong in those chats. Especially since often there are minors present.
In others, I'll just send prompts and check if it's true or not, and unbiased. Sometimes it's very simple, and other times very complex. It all depends on the day and the project. Either way, I think it's fun and I'm learning a lot while I'm working. So a win win situation for me. I feel like a computer sometimes LOL. So to answer your last question: yes, I do get involved with checking output.