I think standard difference-making risk aversion, understood to not at all discount the modestly rarer cases where you cause vast amounts of net harm to invertebrates (and they turn out to be sentient and count substantially morally individually), will just count a lot against doing almost anything relative to inaction. Even modest levels of difference-making risk aversion?
Animal agriculture has huge effects on wild animals (land use, climate change) but with unclear sign, so farmed animal work could backfire spectacularly. Confidence either way on the sign of the net effects of farmed animal work on wild invertebrates (or in a small net effect) isn’t warranted, and a modest bias against backfire could rule out almost all farmed animal work, maybe other than slaughter reform or extremely careful farmed (and/or wild-caught) invertebrate welfare. Even chicken welfare reforms have effects on land use.
Research on wild animal welfare, sentience and interventions could be misused and cause net harm to vast numbers of invertebrates. And I mean “misuse” here pretty broadly: further research might not solve normative/moral cruxes,[1] and then it could take a course that you would judge to be net negative with modest probability (e.g. 5%, considering also moral uncertainty you could still have). Similarly, moral advocacy could end up leading to a course of work affecting wild animals that’s worse than otherwise by your judgement with modest probability.
Practically anything that affects humans affects farmed animals and wild animals, and could affect wild animals far more and be net negative.
Some responses:
Diversification (or more importantly, hedging) with a portfolio, as you point out, could help. I’m not sure how much.
Give less weight to both extreme upsides and extreme downsides, which means partially discounting net harm to invertebrates. (You could still discount upsides more.)
Use some other account of difference-making risk aversion, e.g. pairwise comparisons without privileging doing nothing as an option, or compare pointwise using corresponding quantiles instead.
(Minor edits for clarity.)
I think standard difference-making risk aversion, understood to not at all discount the modestly rarer cases where you cause vast amounts of net harm to invertebrates (and they turn out to be sentient and count substantially morally individually), will just count a lot against doing almost anything relative to inaction. Even modest levels of difference-making risk aversion?
Animal agriculture has huge effects on wild animals (land use, climate change) but with unclear sign, so farmed animal work could backfire spectacularly. Confidence either way on the sign of the net effects of farmed animal work on wild invertebrates (or in a small net effect) isn’t warranted, and a modest bias against backfire could rule out almost all farmed animal work, maybe other than slaughter reform or extremely careful farmed (and/or wild-caught) invertebrate welfare. Even chicken welfare reforms have effects on land use.
Research on wild animal welfare, sentience and interventions could be misused and cause net harm to vast numbers of invertebrates. And I mean “misuse” here pretty broadly: further research might not solve normative/moral cruxes,[1] and then it could take a course that you would judge to be net negative with modest probability (e.g. 5%, considering also moral uncertainty you could still have). Similarly, moral advocacy could end up leading to a course of work affecting wild animals that’s worse than otherwise by your judgement with modest probability.
Practically anything that affects humans affects farmed animals and wild animals, and could affect wild animals far more and be net negative.
Some responses:
Diversification (or more importantly, hedging) with a portfolio, as you point out, could help. I’m not sure how much.
Give less weight to both extreme upsides and extreme downsides, which means partially discounting net harm to invertebrates. (You could still discount upsides more.)
Use some other account of difference-making risk aversion, e.g. pairwise comparisons without privileging doing nothing as an option, or compare pointwise using corresponding quantiles instead.
And even if it does, people might ignore the solutions.