we can't 'know' for certain how it would go unless it is first tested

We agree on this

I think we both have a pretty good guess as to what is most likely to happen with free downvotes... and that is them being used primarily for personal reasons as opposed to responsible and altruistic curation.

I think they are likely to be used for both. In fact they already are, just on a very small scale, and that small scale introduces a severe imbalance in the system which opens it up to a vast degree of milking and other value-destroying behaviors. That is a far bigger problem than a relatively small amount of trolling. That is my view.

But, again, we don't know until we try. The rate of iteration on Steem/Steemit is far too low in my view. If we only get to try something every few years there is really no chance of reaching a better outcome before Steem sinks permanently into obscurity.