An introduction to signalling theory

Signalling theory as a formal concept originated in the field of animal behaviour (ethology) in the Dutch ethologist Tinbergen’s book “Inleiding tot de diersociologie“ (1946) and later found purchase in economics, where it was introduced in the context of job markets by Michael Spence (1973), and finally into political culture, i.e. in the form of “virtue signalling.”

This is partially a pre-amble to Holly Elmore’s Virtue signaling is sometimes the best or the only metric we have, because I felt some of the readers could have benefited from more background knowledge.

What is a signal?

A signal is simply a unit of communication. Although it’s often used in humans to talk about non-verbal forms of communication, this is likely because the classic examples of signalling theory in non-human animals are often, though not always, non-verbal.

A peacock’s attractive tail is intended to impress peahens. Analogously, one might think if a man wants to signal (communicate) that he’s rich to potential mates, he might purchase an expensive car or watch to impress women. But signals don’t have to be non-verbal.

In animal behaviour there are a variety of definitions of a signal, but essentially a signal is anything evolved to change the behaviour of the receiver^[1]. Signals can be an action, like waving your arm, or chemical, like the pheromones an ant leaves for others to follow. They can also be verbal, like the mating call of a bird. Or they can be a colour or design of body morphology, like our peacock’s tail, or in humans, height or facial symmetry.

Signals don’t have to be cognitively intentional

As we step away from animal behaviour, the definition of a signal as something that specifically has to evolve to change the behaviour of the receiver can sometimes be a tricky subject when we’re talking about humans.

A peacock, of course, does not intend to grow a beautiful tail to attract a mate; it simply happens without him thinking about it. In humans, signals might be deliberate—there are plenty of times when humans are conscious of what they are trying to communicate—but they can also be unconscious, such as an instinctive shriek of fear if someone surprises you, or as in the peacock, not even involving the brain at all.

However, it is sometimes convenient to talk about a signal as something which is intended to change the behaviour of the receiver. This is a verbal red herring. Human speech evolved in order to signal, but when we talk about verbal signals, sometimes there is confusion about to what extent intentionality matters for something to count as signal. From the perspective of animal behaviour. The sender of the signal need not be conscious of it, though they very well may be. Receivers can equally make use of both unconscious and conscious signals.

Senders, receivers, and eavesdroppers

In communication, there is usually an intended receiver of the information. However, information may be accessed by a third party - this is termed eavesdropping. This can affect how signals evolve, because third parties who overhear a signal might use it to their advantage, and to the disadvantage of the sender.

In fireflies, some females of some species have evolved to mimic the flashing light pattern of other species, attracting males of other species- whereupon she promptly eats them.

In politics, a “dog whistle” is an example of a kind of signal designed so that only the intended recipient is supposed to understand it.

Honest and dishonest signalling

There is also the issue of “lying”. A signal is an honest signal if it correctly signals the information—but as dishonest signalling can be adaptive, this is a challenge for the evolution of communication.

If a bird can simply lie and say, in birdsong, “I’m high quality”, and this gains him extra mating opportunities, then all the birds would do this and the signal becomes useless; any females selecting males on the basis of that particular birdsong would do no better than birds that didn’t.

This is partially resolved by Zahavi’s 1975 paper in which he introduces the concept of costly signalling.^[2] The idea is that signals are more likely to be honest if they incur a cost, and this drives the evolution of costly signals, because they are less likely to be faked.

Driving a Porsche usually does indeed indicate wealth, in the same way that peacocks with nice tails are probably of genuine high quality. You can’t fake the good foraging ability it takes to energetically support the tail, in the same way it’s somewhat hard to fake having enough money to own a Porsche (though humans especially excel at such deception).

Much of the study of signalling theory in animals involves figuring out the costs of various signals, to figure out how honesty of the signal has been maintained. In humans, which have a much more complex communication system than anything in the animal kingdom, this presents a challenge. How is honesty mediated in human communication?

Evolution of communication in humans

Probably most of you are familiar with the Prisoner’s dilemma, in which two rational agents will defect against each other, resulting in a poorer outcome for them both if they’d only both cooperated. Humans have obviously solved this problem, but how?

One way humans solve this is with a combination of skills—individual recognition- the ability of us to recognise and remember the faces of other humans—and reputation. This allows people to do iterated Prisoner’s dilemma—the same game, but played repeatedly. If a person has a tendency to cooperate, then the other party can use this information to solve the game and both can benefit.

It is therefore typically advantageous for humans to be virtuous, because that can engender mutually beneficial, reciprocal relationships that improve the standing of both parties. However, it’s also beneficial to merely be seen as virtuous, and then use this reputation to “get one over” the other person. These are wolves in sheep’s clothing, so to speak.

Virtue signalling

In literal terms, a cooperative human society essentially rests on signalling virtue. But I don’t think an accusation of virtue signalling is meant to apply to any visibly virtuous person. Rather I think a number of assumptions are rolled into an accusation of virtue signalling.

Firstly, is the signalling honest? In many cases, an accusation of virtue signalling is an accusation of dishonest or false signalling of virtue. An example might be a “fake male feminist” who either intentionally misrepresents his beliefs- or even genuinely believes them, but whose actions are no more cooperative than a person who doesn’t hold those beliefs.

Despite the relatively recent appearance of the term “virtue signalling”, the problem of questioning someone’s motives for appearing virtuous is an old idea. Matthew writes 2 millennia ago, “So when you give to the needy, do not announce it with trumpets, as the hypocrites do in the synagogues and on the streets, to be honored by others. Truly I tell you, they have received their reward in full.”

Why would signalling virtue if you’re actually virtuous be bad? This starts to get more speculative, but it may mean they may be motivated by the benefits of appearing virtuous instead of actually being virtuous.

To a certain extent we seem to have a preference to cooperate with individuals that are genuinely cooperative and virtuous (authentic altruists)^[1]. We have less reason to worry they’ll defect. If a person makes it too obvious they’re instead after the benefits, we become suspicious they’re false altruists who will defect the moment it benefits them. It’s rational to be wary of these Raskolnikov-esque Übermensch.^[3]

(By the way, this may be part of the reason a few people find EA “creepy”—helping someone because it “feels like the right thing to do” indicates you might be a good person to do reciprocity with.^[1] Some EAs might inadvertently be giving off the impression they’ll defect, take the higher payoff, and donate it all to a good cause.)

^
Michael D. Breed, Janice Moore,
Chapter 7 - Communication,
Animal Behavior,
Academic Press,
2012,
Pages 183-217,
ISBN 9780123725813
https://sci-hub.se/10.1016/B978-0-12-372581-3.00007-6
^
Zahavi, A. (1975). Mate selection – a selection for a handicap. Journal of Theoretical Biology 53, 205–214. https://sci-hub.se/10.1016/0022-5193(75)90111-3
^
Raskolnikov is the main character in Crime and Punishment, who thinks of himself as an ubermensch (superhuman) and rationalises himself into committing murder on a utilitarian basis, only to find out he’s human after all (i.e. wracked with guilt). Whoops.