1. Your model of the world is incomplete or inconsistent (eg, forgetting to include negotiation as a strategy, forgetting long term consequences).
2. You've hit on a cognitive bias and can probably do a huge amount of good that other people will miss (eg, lending with interest -> modern industrial society, rule of law rather than letting high status people do whatever -> mass prosperity). The cognitive bias is probably well known and understood and the evidence that the emotional response leads to poor outcomes should be obvious.
I'm completely befuddled by this. My experience is that people make most decisions emotionally and then harness their brain to rationalize those irrational decisions after the fact. It's the reason that science (the process) has to exist. Because people will fool themselves to reach a conclusion they want if there are no procedural guardrails.
Surely 800,000 mosquito nets is "more effective" than 800,000 robo calls and shitty campaign ads with a moron at the end saying "I'm Joe Blow, and I endorse this bullshit"
Then you will be delighted/horrified to find out that the effective altruism community doesn't feel the need to make quantifiable justifications for how they spend their own donations, as evidenced by this $20k grant they gave someone to learn to ride a bike and think about AI[1].
That's about 3853 to 4090 mosquito nets' worth of donations, using GiveWell's numbers for the Against Malaria Foundation[2].
[1] https://forum.effectivealtruism.org/posts/CJJDwgyqT4gXktq6g/...
[2] https://www.givewell.org/charities/amf