Not quite; you can put an upper bound on the "badness" of unknown harms, because if it was bad enough, we'd have detected it already.

For a reductio ad absurdum proof, imagine that microplastics kill you on contact. This would show up in the death statistics very quickly. So in practice we can say with confidence that the possible unknown harms don't include mortality effects above a certain level.

There's definitely room for effects that we'd consider to be serious like "reduces male sperm counts by 50% over decades of exposure" or "increases cancer rates by 100%", but it's certainly possible (indeed, the correct and rational way of approaching this problem) to attempt to bound the upper limit of harm based on what we should be able to detect.

Note, I'm not commenting on the object-level question of _whether_ the benefits outweigh the costs or whether this is known; I would like to see a citation for that as I've not seen such an analysis. I'm just addressing the meta-level question of if it's possible to know.