'Virtue-signalling' has only been around a few years. It's usually traced to James Bartholomew's 2015 The awful rise of 'virtue-signalling' at The Spectator (which Levy can't even be bothered to link). Despite Levy's claim that he did not in fact invent it, this is not obvious; while you can find very occasional uses of phrases like 'virtue-signalling' prior to Bartholomew, it's not always clear that they are using the phrase in the same sense as Bartholomew, and they are in any case sporadic enough that there's not really any reason to deny that he invented, or at least independently invented, the term as it is used today. What is certainly true is that Bartholomew is a major dispersion point for the term as it is currently used, since his article was widely shared across social media and he wrote a book on it.
There are a number of potential pitfalls we have to be careful about from the beginning. Given that it has so widely been used as an insult, we have to be very careful, because people in general tend not to be very accurate in the use of insult phrases. And just as I've noted that popular arguments tend to be summaries of argument-families rather than precise arguments, so too the details of terms arising in popular usage will often be fuzzy. In particular, the meaning is better seen by judicious consideration of common or paradigmatic cases rather than any strict definition laying down a precise boundary in a classification. It would not, I think, surprise anyone that there is a lot of gray area with this term -- things that could be considered virtue-signalling that are not necessarily bad, things that may or may not be virtue-signalling, etc. It's the central and obvious cases that are determinative. Bartholomew treats virtue-signalling as a kind of advertising:
Go to a branch of Whole Foods, the American-owned grocery shop, and you will see huge posters advertising Whole Foods, of course, but — more precisely — advertising how virtuous Whole Foods is. A big sign in the window shows a mother with a little child on her shoulders (aaaah!) and declares: ‘values matter.’
Notably Levy does not mention advertising once. Contrary to what Levy claims, Bartholomew does not say that virtue signalling is driven by "vanity and self-aggrandisement"; what he actually says is that in certain cases in which you virtue-signal by expressing what you hate, using your hating the right people or things to show how good you are, if you were to say directly what you meant, "your vanity and self-aggrandisment would be obvious", just as with certain blatant kinds of value advertisement like in the Whole Foods example.
There is a more important potential pitfall, and it is unfortunately one into which Levy directly jumps. You can't assume, as you can with technical and semi-technical terms in academic fields, that a term in popular parlance is being used literally. It is quite clear from the tone of Bartholomew's article, and this is also clear from common usage, that it is at least broadly ironic. The label may be 'virtue-signalling', but you can't assume that every sense of 'signalling' is relevant, and you can't assume that it literally has much to do with virtue itself. What Levy does in the course of his article is unfortunately an easily identifiable mistake: he argues that an insulting label actually indicates a good thing because if you read the label hyper-literally rather than sarcastically or ironically, it would describe something that could be considered good in some cases. Because of this mistake, we get drivel like this in an argument that signalling is important to moral life, as if it were relevant to the subject:
Signalling is very common in nature. The peacock’s tail, for instance, is a signal of evolutionary fitness. It’s what biologists call an honest signal, because it’s hard to fake. It takes a lot of resources to build a tail like that, and the better the signal – the bigger and brighter the tail – the more resources must have been devoted to it.
*Gasp* Who would have thought that useful signals of some form or other exist? That changes everything! It is also, of course, not obviously relevant; Levy nowhere shows that virtue-signalling is a way to signal evolutionary fitness, or that it is even signalling in a relevantly similar sense. Now, of course, some analogy of some kind is entirely reasonable to expect, but what one needs is exactly what Levy never bothers to give: the account of what kind of analogy, on what grounds, is in play. For instance, if we simply took Levy's argument as it stands, it would imply that advertisements fulfill a "central function" of moral discourse; this commercial signals to you that State Farm cares, caring is a moral notion, therefore selling State Farm insurance and the like is fulfilling a central function of moral discourse. It's not clear what "central function" could even mean in this kind of context, and it seems unlikely that anyone not thoroughly infected by a consumerist culture would even imagine such a thing. But there's no real difference between this kind of argument and Levy's; an original and recurring context for applying the term 'virtue-signalling' is advertisement, as in, for example, trying to convince people that virtuous people are the kind of people who buy and sell expensive all-natural artisanal cheese.
Levy's argument is in fact a mighty game of leap-frog: we start with signalling in nature (for reasons unknown, but my guess is that it is supposed to make more plausible the "central" in the "central function"), then to the more limited technical meaning in cognitive science of religion (overlooking, as far as I can see, that 'signalling' in this sense is an after-the-fact analysis of effect, not something that is always deliberately or even half-consciously done the way virtue-signalling would have to be for the term to serve the function it does -- the claim is not that hermits necessarily live the eremitic life because it signals that they are holy, but rather that people find hermits impressive and memorable because they take the costliness and apparent sincerity of hermits as a sign or signal of holiness), then to saying that religious signalling of this sort is a kind of moral signalling (in a broad sense of the term), then from 'moral signalling' to 'virtue signalling'. Every hop in this chain of lilypads is thoroughly problematic. Going backwards, we have no reason to think from the behavior of the terms that every kind of moral signalling in any sense of the term is virtue-signalling; while the capacity of religious activities to be taken as a sign of goodness can be called 'moral signalling' in some sense of the term, this is in a very broad sense, and it seems to require a different action-structure than virtue-signalling, since it is primarily about how others interpret your actions rather than what you yourself are trying to signal by them; the use of the term 'signalling' in the cognitive science of religion does not have any straightforward association with the kind of explicit moral advertising associated with the term 'virtue-signalling'; and the fact that signals are common in nature tells us nothing about whether virtue-signalling is itself good, or fulfills any central function of moral discourse, or indeed is signalling in anything more than a very loosely overlapping sense of the term.
One of the notable problems with Levy's hyper-literal reading is that he takes 'virtue-signalling' to mean that there is some particular kind of independently recognizable virtue that the virtue-signaler is trying to display. As far as I can see, this is an illegitimate assumption. Doritos advertising that it is really committed to gay marriage by selling special Pride bags is not identifying some particular virtue and then displaying that it has it; it is a corporation, it doesn't have that level of coherent self-reflection. Doritos executives are signalling support of gay marriage because they are trying to get in on a (potentially profitable) thing by a symbolic action that Doritos can treat, and that Doritos executives hope others treat, as itself a virtue stand-in. As Bartholomew puts it (and notably Levy again completely ignores it, despite the fact that structurally it is the core of Bartholomew's article):
No one actually has to do anything. Virtue comes from mere words or even from silently held beliefs. There was a time in the distant past when people thought you could only be virtuous by doing things: by helping the blind man across the road; looking after your elderly parents instead of dumping them in a home; staying in a not-wholly-perfect marriage for the sake of the children. These things involve effort and self-sacrifice. That sounds hard! Much more convenient to achieve virtue by expressing hatred of those who think the health service could be improved by introducing competition.
The 'virtue' in 'virtue-signalling' is not about virtues; virtue-signalers aren't displaying their virtues, nor even necessarily faking the display of virtues like an ordinary hypocrite (although it seems common enough that people take virtue-signalers to be doing so because they are hypocrites). The whole point of the term (and actual usage seems to continue Bartholomew's point here) is that the 'virtue' in 'virtue-signalling' is in fact nothing more than the signalling. It is not costly; it is not credibility-enhancing. It is simply assuring yourself or others, directly or indirectly, that you are on the virtuous side as if self-identification as virtuous were the same as being virtuous, treating the advertising of values as proof that you live by those values. It is, as I said, an ironic label.
[ADDED LATER (almost immediately afterward, in fact): As a minor side note, here's a sign of just how much this term has struck a chord. When I started writing this post, I opened Twitter and searched 'virtue-signalling' to refresh my sense of the common usage. (Unsurprisingly for Twitter, most uses were cases of it being used as an insult without further explanation.) In the little more than half an hour it took to write this post, Twitter added over eighty new tweets with the phrase.]