Sounds like you're valuing <mean harm per animal> over <integral of harm over all animals>?

I don't get why that would be a better measure?