Nerd Pride Radio

One for the old science corner, if you haven't already done it recently, is this study about wasp "reasoning":
https://royalsocietypublishing.org/doi/ ... .2019.0015

In a nutshell: they trained wasps to avoid one color stimulus in a pair, and taught several pairs related in a series. They they tested to see whether the wasps made a kind of transitive inference, where if b is worse than a and c is worse than b, the wasp will correctly predict that c is worse than a. They train 5 in a row, abcde, to ensure the test case is not always an avoid (or prefer) color, and then compare results on both a/e and b/d.

Result: it worked, the wasps learned, and they were the first invertebrates to succeed in such a test.

Difference: this test uses aversion to electric shock rather than positive rewards.

So does this actually test inferential ability? I would say no, but then, such ability evolves from somewhere. Is this the foundation? I'm deep into the rabbit hole on this one.

Here's the quickie CNN version.
Why don't I think it's inferential reasoning? Because the wasps form some sort of mental hierarchy of the five colors, where transitivity doesn't require that kind of relationship.
OK, you might say, but this is still A case of transitivity, right? But I wonder if e.g. B always being associated as proximate to the always-safe-zone A, even if B is 50/50 bad, causes it to be positioned on the hierarchy near A anyway, especially since the learning effect was not huge and the number of test cases small. If so, what we have is more like "relative dangers, distinguished and stored in a sequence that can be reaccessed in different combinations". But that's not inferential. It's a really cool building block of potential inference-making, however.