My bad. I think the situation should continue to be improved, as it surely has in the last couple of generations. You're welcome to think otherwise, of course.
I've looked around a little bit and I simply cannot find any evidence that police violence is on the decline. A cursory look at a graph of police killings over the last few years shows, if anything, a slight increase (https://www.vox.com/2020/5/31/21276004/4-charts-anger-police... - first graph).
I can only speculate about what your exact argument is: I'm guessing that what you're saying is that, surely, today's police are behaving better than the police in, say, the civil rights movement era (I couldn't find any data on that but I'm willing to accept that much for the sake of argument). And I think you infer from that that police violence is downward trending and that therefore, we should just let matters run their course: any upset to the delicate improvement might plunge us in the other direction.
But your inference is incorrect. You're looking at a function over time, see that this function was (probably) higher at some point in the past than it is now, and conclude from that that the function is still trending downwards. Essentially your argument seems to be based on the entirely unjustified assumption that the function has to be monotonically decreasing.
But if you don't assume monotonicity (and you really shouldn't), then function values in the far past don't give you any information about the derivative of the function at t=now. And the derivative at t=now seems to be (ever so slightly) positive, not negative as you imply. In which case, evidently, things are not improving on their own.
