I think this is
heavily subjective, because you have criteria like the attention you give him, directly or indirectly, and that triggering people donating to him as a consequence, which cannot be quantified. We're also determining how "bad" or "retarded" something is, which again is subjective.
Example: some whales/trolls have tipped large amounts due to Kiwifarms & this board's conversation about them, but that's an indirect consequence from the discussion that happens here, i.e: the "attention" that this board is giving him.
But that's just on one end, you have more stuff to throw into the equation. Like for example, how is money actually earned through Kick? Seems like you don't agree on this topic with some users (like in regards to chat interaction).
You also have how aware are you of the alternatives (e.g: do you watch a restreamer who is giving DSP an extra view, or the one who is restreaming RawPhil? That way you'd deincentivize the one giving him the extra view, and unify the DSP restream process into the way that gives him the least amount of money, like if all detractors agreed to restream RawPhil (as an example) to not give him even an extra view, but at the same time they have the right to restream him as well, not just RawPhil).
So assuming chat interaction is a variable
amongst others (like viewership retention) in order for him to earn money, you now have a gradient of "how bad this action is" that you can build and morph according to your own criteria.
This stuff could be made into an equation/one of those test online that determine your score for autism.
Variables for the equation (some)
F = (boolean) do they watch DSP or not?
f = (boolean) do they watch DSP directly or indirectly?
M = (boolean) do they donate/tip to DSP at all?
m = (float) calculate how much they donate/tip.
T = (boolean) do they donate/tip to "troll" or not?
E
0 = (boolean) do they enjoy DSP's content or not?
E
1 = (boolean) do they enjoy the restreamer's content or not?
B = (float) quantify the bias (how bad/good it is) based on the quality of the restreamer.
R = (boolean) do they donate/tip to the restreamer or not?
r
0 = (float) calculate how much they donate/tip to the restreamer.
r
1 = (float) calculate bias (how bad/good it is) of amount given, based on the quality of the restreamer.
A = (boolean) do they use some form of ad-blocker on sites where it matters (e.g: YouTube) or not?
I = (boolean) do they interact in DSP's chat or not?
i = (float) quantify the level of positive/negative interaction/messages.
C = (boolean) do they interact with DSP's content, directly & outside of streams or not?
c = (float) quantify the level of positive/negative interaction/messages.
K = (boolean) do they talk about DSP in detractor circles or not?
k = (float) calculate the bias (how bad/good it is) of that attention they give DSP.
Q = (boolean) do they incentivize others to donate/tip to DSP?
q = (float) calculate the bias (how bad/good it is) of those comments.
...

And many more! You'd have to calibrate the range of that bias according to your criteria, and the scoring system, given at the end/how wide the gradient is.