That seems like saying: "Let's not donate to animal charities because there are people who would donate to the most effective human charities but decide to donate to the less effective human charities when they see people who donate to the most effective human charities switch their donations to animal charities." Probably I'm not following the logic...

Also: if donating to the top-effective animal charities is +100 times as cost-effective as donating to the top-effective human charities, that backfire effect (people donating to the less effective human charities instead of the top effective human charities) should be very strong: more than 100 people should show this backfire effect (i.e. remain non-EA) per effective altruist who donates to top-effective animal charities. That seems very unlikely to me. 

"So it is more important to convince someone to give to e.g. the EA animal welfare fund if they were previously giving to AMF than to convince a non-donor to give that same amount of money to AMF." More generally, I think it is more important to convince an EA human health and development supporter to diversify and donate say 50% of the donation budget to the most effective animal welfare causes, than to convince a non-EA human charity supporter to diversify and donate say 50% of the donation budget to AMF or similar high-impact human-focused charities.

I think there is much room for more funding of alternative protein R&D, and that is very cost-effective to reduce farmed animal suffering

Two important considerations to strongly favor animal welfare 

1. Saving a human life is likely net negative due to increased meat consumption and animal suffering. According to a survey, most people believe the welfare of a farmed chicken is negative and equal in size to the positive welfare of a human. Also most people believe the welfare of birds count almost as much as the welfare of humans (they give animal welfare relative to human welfare an 8 on a scale from 0 to 10). But there are more farmed chickens than humans on earth (3 chickens per human), so total welfare is negative. See also this post: net global welfare is negative and declining. This also means that saving a human (meat eater) most likely negatively contributes to global welfare, as it increases meat consumption and hence farmed animal suffering. This is a strong version of the meat eater problem (strong in the sense that saving humans not only increases animal suffering, but increases it so much that total welfare decreases).
2. Saving a vegetarian human life is likely less cost effective than avoiding farmed animal suffering. There are extremely cost-effective animal welfare interventions. For example development of alternative protein such as cultivated meat saves the suffering of +10 farmed animals per dollar. An average human eats 10 farmed animals per year. So $100 donation to cultivated meat R&D saves the suffering of 1000 farmed animals, which equals the amount of farmed animal suffering caused by 100 years of meat consumption by a human. In size (absolute value), the suffering of 1000 farmed animals is more than a lifetime of human happiness (roughly 3 times as much). In other words: avoiding the suffering of 1000 farmed animals is better than saving a child's life such that the child lives 100 happy years. According to GiveWell's estimate, saving a child's life costs +1000$. Saving 100 healthy human life years easily costs +10.000$. So avoiding farmed animal suffering is +100 times as cost effective as saving a child's life (assuming the child is vegetarian or vegan).

I propose to use new terms: Importance, Implementability, Ignoredness. So three I's, with the I of Impact. 

Equation: 

Impact = Importance x Implementability x Ignoredness

In Dutch these are the three O's: Omvangrijk, Oplosbaar, Onderbelicht

the fact that mind is determined by a physical system not necessarily entail epiphenomenalism. My best analogy is the difference between the object language and the metalanguage. In mathematics (number theory, Godel's theorem), the metalanguage is embedded in the object language. https://en.wikipedia.org/wiki/Metalanguage#Embedded In this sense, the metalanguage supervenes on (and is determined by) the object language, but is not an epiphenomenon (and not eliminated either). 

What I meant was that from all public outreach, this deep questioning is one of the most cost-effective. If deep questioning is much less cost-effective than other strategies, then surely public outreach in general does not belong to the top effective strategies.

Which opportunity costs were not explored?

I think deep questioning should be done as an ancillary activity which people can do in their free time

Some grassroots animal rights organizations that do a lot of public outreach with volunteers, could perhaps switch to deep questioning, or encourage their volunteers to do so.

Please let me know how I could have earned extra money at those times when I do deep questioning ;-) I think it's really difficult, if you have some spare time today, to say: I'm going to earn some extra money. But I can imagine spending that spare time to learn some new skills and then use those acquired skill to switch career to a higher paying job. But to me, that's a difficult strategy. So to me, deep questioning public outreach seems to have a low opportunity cost.

That 0.6$ seems to me very low. The annual revenue of THL is almost $20M, so THL neutralizes the harm caused by 30 million people?

One conservative estimate of the effectiveness of deep questioning

-10% of people reduce meat consumption after the conversation

-for those reducers: 10% reduction of animal products (especially from the small animals)

-reduction fades out after 10 years.

So 10 conversations reduces the harm caused to farmed animals by the annual food consumption of a random person.

I did a small follow up study in my early days of deep questioning, around 2016. I mailed the people I spoke with on the street, a few months after our conversation. 50% responded to my mail, 20% of those respondents said they reduced their meat consumption since our conversation. That could mean 10% of people change their behavior. That seems pretty effective for the difference making risk averse person: it requires only 5 conversations with couples to make some difference. In my recent conversations, at the end of the conversation, roughly half of people say they really intend to reduce their meat consumption, try new meat alternatives. The other half is not interested that much. 

I often mention fish, eggs and shrimp, and especially mention that there is more suffering involved with smaller animals. I often mention that it is not good to replace chicken meat with eggs or fish.

Thanks for the questions, David. Insects are not really ignored. When I refer to small animals, that includes insects. The transition can happen in many ways, both legally (regulations that decrease chicken farming), economically (taxing chicken meat). Farmers can be asked to sell the land to the government, who turns it into grassland habitat. Social norms could be the major obstacle. Individual consumers can always resist social norms and decrease their consumption of products from small animals, even if that goes against the social norms. And we could of course change social norms. Also, there may be social norms about meat consumption in general, but not about chicken meat consumption. The ask is to reduce chicken meat consumption, which is more feasible than going vegan.