It will be interesting to see if you can crack the code on what "good" content is.
I think transparent is better. I suggest that being neutral and unbiased is probably not possible. So in that regard, I would avoid presenting yourself as neutral and unbiased. The main reason I say this is, if you present yourself as neutral, you open the door to a lot of criticism and it will potentially work against your goal of curating "good" content as you will get a lot of feedback that is not collectively coherent.
“Good” is inherently subjective and trying to find the perfect formula that lasts forever may not be attainable. You should always strive to curate "good" content with the qualifier that your curation will never be perfect and you may need to adjust things overtime. (You will likely want to adjust how you vote based on the size of your stake weight for example.)
What I have found to be the best starting parameter is to curate posts with good engagement with interesting discussions. Once again, “good” is subjective. But if you want to curate “good” content then looking for content that other people think is “good” based on how they engage with that content is a good place to start.
RE: The AI Curation Experiment