It
will modify the behaviour of
some of the people who vote.
Which is the point.
The interesting part is whether they can group out users who are more predisposed to upvote violence, develop a measure of tendancy, and observe a change for some subgroup.
If the admin want to downplay violent content they have a range of options, eg: if they have suitable measures then those who do not modify behaviour can have their upvotes diluted or negated on violent content or any other actions up to shadowbanning or beyond(?).
Reddit literally has a decade+ database on how people vote.