What's really interesting here is the problem of just a "naive threat estimation", which usually is always just estimating the purely mechanical danger the body of a foe could pose. We have a healthy respect for large animals like horses and cattle that could crush us. We've got an open fear of predators with obvious sharp teeth and such (wolves, big cats, crocodilians). But bugs? Bugs are really weird because they're not mechanically dangerous, and as, say, a child, sizing them up, there's no sensible reason to be scared.
I find it really interesting how - even without necessarily understanding that they've got poison, we as a species do a really good job of being scared of "things that aren't visually obvious threats" like this - especially in kids, who really struggle to "get it" when it comes to what's actually dangerous about, say, a spider or a jellyfish.
I think you really hit on something there - I think a little bit of it is actually a built-in instinctual fear, but there's also a huge part that's an imitative fear kids pick up from anything that the body language of a parent conveys as dangerous.
Think about how you feel approaching a very large animal, like a horse you don't know. You don't need to know how or even if it can hurt you, but it's worth being careful. If you know the animal has sharp points (like a large dog), even more so. I think sparrows avoid cars by similar logic.
I really think non-verbal, subconscious intelligence is capable of a lot more abstraction than people think. We need language to describe it, but not to do it. "Threat" is an abstraction that lots of simple animals handle just fine. It's not so much that we need to give animals more credit than that we give ourselves too much. Our actual motivations are usually much simpler than the stories we tell ourselves.