Disclaimer: My area of expertise is mental health, not animal welfare, so it's possible I'm about to express some really ignorant ideas. Although many uncertainties exist around measuring and improving human well-being, I found it interesting to apply such themes to animal welfare as a thought experiment.

Terminology: Happiness, life satisfaction, and subjective well-being are all different but fairly related concepts. For the purposes of this discussion, I treat them as roughly all the same idea, because the measured differences between them seem trivial (at least in humans) relative to the scale of animal welfare concerns.

We might be overlooking animal life satisfaction by focusing on pain/suffering

Life satisfaction seems to be one of the most promising ways to measure the impact of mental health interventions on people. There is limited research on the effect of pain (e.g., chronic pain and extreme pain) on human life satisfaction, but a reasonable estimate so far seems to be that 0-10 pain scale scores have a 0.1 conversion rate in terms of decreasing life satisfaction measured from 0-10. In other words, being in extreme pain compared to none at all and having a consistently dysregulated nervous system may decrease life satisfaction by one point on average. (If this sounds surprising low, humans tend to overestimate the impact of physical conditions/disabilities on life satisfaction by as much as 10x.)

Informally speaking, the EA research I've seen on animal pain assume that species that fulfill more categories of neural/behavioral evidence of pain probably experience more sentience and pain. Taken literally, this stance assumes that humans experience the widest range experiences of pain and suffering. If extreme pain alone is (usually) not enough to make a human life not worth living, then what does this imply about animals? Do we expect animal life satisfaction to be considerably more dependent on lack of pain?

Although I'm not aware of any EA research on animal life satisfaction, there are some studies out there, such as on chimpanzees. Life satisfaction can't be measured the usual way (we can't ask chimpanzees to rate their life satisfaction out of 10), but for example it can be subjectively estimated by human zoo keepers. Michael Plant suggests that objective measures of wellbeing may be more appropriate for animals and cognitively impaired humans. We probably don't know much about animal life satisfaction and the threshold for net positive vs net positive animal life. And this could be highly significant if it turns out that we are overestimating the conversion rate between extreme pain and life satisfaction, as we intuitively do with other humans.

It might be worth comparing the life satisfaction of farmed vs wild animals

We know surprisingly little about the lives of wild animals and how much pain and suffering they truly experience, often falling into the idyllic assumption that wild animals thrive in nature and less so in captivity. All we have is the visible suffering of farmed animals, but we don't know if wild animals might be far better or far worse off. It seems particularly challenging to measure the life satisfaction of wild animals directly, but we could potentially form a tentative understanding of the life satisfaction of farmed animals first, and then propose objective metrics to form a relative (rather than absolute) comparison with wild animals.

Implications

Combining these two ideas could lead us to four possible conclusions per appropriate class of animals.

1. Farmed and wild animals both have high satisfaction
   If we learn that farmed and wild animals tend to have high satisfaction by default (arguably, this is somewhat true in humans), then some animal welfare projects can be redirected e.g. towards other classes of animals where this is not the case.
2. High satisfaction for farmed animals, low satisfaction for wild
   In this case, we could direct attention towards wild animals or other classes of animals.
3. Low satisfaction for farmed animals, high satisfaction for wild
   Prioritization seems appropriately placed here.
4. Farmed and wild animals both have low satisfaction
   We could focus on whichever category is more tractable.

Who knows, we might find that these conclusions apply to quite broad classes.

There is a particularly challenging implication in cases where wild animals have low satisfaction and extremely high populations. It raises difficult questions about whether extinction is a better option than other interventions, though this could also have unpredictable destabilizing effects on ecosystems. As an extreme example, if an all-knowing oracle told us that insect wellbeing is equally morally as important as human wellbeing, that all insects have extremely poor life satisfaction by default, and that the current population of insects is 10^19, how would EA prioritize this welfare issue, given that the scale is massive but tractability seems low for targeting only insects?

In general, are there any practical ways we could be exploring animal life satisfaction and its connection with pain, instead of just pain alone?