automatically pre-score a content based on various points which I won't reveal as to not allow users to find a way around it when it will be implemented
If you could do this transparently and still have it work, it would actually be useful. Bad actors will eventually figure out fragile but hidden algorithms. It's generally not seen as a good form of security to hide implementation details. Good IP, but not necessarily the most trustworthy.
The pre-scoring is just a first layer to help the human being that will actually perform the upvote actions. We don't want a fully automated upvote process as as you said, it's not necessarily the most trustworthy.
Thank you for your valuable input.