Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio.
This is: What Is Signaling, Really? , published by Scott Alexander on the AI Alignment Forum.
The most commonly used introduction to signaling, promoted both by Robin Hanson and in The Art of Strategy, starts with college degrees. Suppose, there are two kinds of people, smart people and stupid people; and suppose, with wild starry-eyed optimism, that the populace is split 50-50 between them. Smart people would add enough value to a company to be worth a $100,000 salary each year, but stupid people would only be worth $40,000. And employers, no matter how hard they try to come up with silly lateral-thinking interview questions like “How many ping-pong balls could fit in the Sistine Chapel?”, can't tell the difference between them.
Now suppose a certain college course, which costs $50,000, passes all smart people but flunks half the stupid people. A strategic employer might declare a policy of hiring (for a one year job; let's keep this model simple) graduates at $100,000 and non-graduates at $40,000.
Why? Consider the thought process of a smart person when deciding whether or not to take the course. She thinks “I am smart, so if I take the course, I will certainly pass. Then I will make an extra $60,000 at this job. So my costs are $50,000, and my benefits are $60,000. Sounds like a good deal.”
The stupid person, on the other hand, thinks: “As a stupid person, if I take the course, I have a 50% chance of passing and making $60,000 extra, and a 50% chance of failing and making $0 extra. My expected benefit is $30,000, but my expected cost is $50,000. I'll stay out of school and take the $40,000 salary for non-graduates.”
...assuming that stupid people all know they're stupid, and that they're all perfectly rational experts at game theory, to name two of several dubious premises here. Yet despite its flaws, this model does give some interesting results. For example, it suggests that rational employers will base decisions upon - and rational employees enroll in - college courses, even if those courses teach nothing of any value. So an investment bank might reject someone who had no college education, even while hiring someone who studied Art History, not known for its relevance to derivative trading.
We'll return to the specific example of education later, but for now it is more important to focus on the general definition that X signals Y if X is more likely to be true when Y is true than when Y is false. Amoral self-interested agents after the $60,000 salary bonus for intelligence, whether they are smart or stupid, will always say “Yes, I'm smart” if you ask them. So saying “I am smart” is not a signal of intelligence. Having a college degree is a signal of intelligence, because a smart person is more likely to get one than a stupid person.
Life frequently throws us into situations where we want to convince other people of something. If we are employees, we want to convince bosses we are skillful, honest, and hard-working. If we run the company, we want to convince customers we have superior products. If we are on the dating scene, we want to show potential mates that we are charming, funny, wealthy, interesting, you name it.
In some of these cases, mere assertion goes a long way. If I tell my employer at a job interview that I speak fluent Spanish, I'll probably get asked to talk to a Spanish-speaker at my job, will either succeed or fail, and if I fail will have a lot of questions to answer and probably get fired - or at the very least be in more trouble than if I'd just admitted I didn't speak Spanish to begin with. Here society and its system of reputational penalties help turn mere assertion into a credible signal: asserting I speak Spanish is costlier if I don't speak Spanish than if I do, and so is believable.
In other cases, mere assertion doesn't work. If I'm at a seed...
view more