Let’s start with a thought experiment; these are common in science. Einstein used them to devise relativity, and James Maxwell used them to devise Maxwell’s demon in thermodynamics. Let’s say somebody replaced your mother with an exact replica. Almost exactly the same, except it didn’t feel right. Try as you might, you can’t tell any physical difference between your supposed mother and your real one. Maybe they are the same, maybe not. Some people claim this is what happens with some politicians. They think Joe Biden is a clone replacing the aging politician. The real question is can you tell the difference between the clone and your real mother?
I doubt most people would; the gorilla test shows that most people don’t even pay attention. A clone might go unnoticed by even the most ardent observer. What about a unicorn? What if I saw a horse with a horn on its head in the middle of the forest? Is it fake or real? The odds are that unless we can show something is fake about it, the unicorn is real if we randomly see one in the wild. This question plagues me now, although not about unicorns, and about something related but different. Could what we see not really be what we see? There are videos going on alleging CIA masks that look like a face; maybe there’s something to it.
The duck test is a test that asks exactly this question. It says it probably is if something walks like a duck and quacks like a duck. This applies in most circumstances but not all. Often paranoia is the way to determine whether the duck is really a duck; looking at it and observing it closely can determine the difference between duckness and non-duckness. A simple flaw or difference reveals the subject as not a duck. We can think of this in terms of probability theory, where the probability of it being a duck is a level of confidence; this is the Bayesian approach to probability vs. the frequentist approach. In this approach, the probability goes from 0, meaning we have no confidence, to 1, meaning certainty.
We can use Bayes’ rule to determine confidence given some evidence. Don’t worry the math won’t be too hard, just multiplication and some basic probability. We need some kind of evidence that we’re looking at a duck. This evidence could be visual, like the color, the beak, or the quacking sound. Then the equation for whether it is a duck given the evidence is this.
If our prior probability of it being a duck is high, then the evidence doesn’t really matter. However, if the probability that the evidence is the evidence we would observe, given that it is a duck, is high, then the evidence does matter. That sounds like a lot. It simply means if we’re pretty sure that given that the animal in front of us is a duck, then what is the confidence we have that it would be a certain color with feathers, waddling, and quacking? For a duck, this is high; there aren’t many other animals that quack, waddle, and have feathers, maybe a penguin, but not really.
Why all this math, though? It’s not really necessary, and often, it gets in the way. The P(evidence) term is difficult to estimate and the duck test is more than just a simple equation; it’s relevant to everything we do today. If an AI system looks intelligent and behaves intelligently, is it really intelligent? This is the Turing test; if a politician looks crooked and acts crooked, are they really crooked? Everyone is corrupt in some way. The Nuremberg incident looked like a space battle; should we say it is otherwise? Again, if it looks and sounds like it, it probably is. I find most extraterrestrial possibilities disheartening and outright evil; I don’t believe in benevolent aliens, and experience has taught me otherwise. I believe everything is evil and contains the seed of evil within it.
Yes, there are exceptions, the show Face-Off depicts contestants creating alien-looking costumes. They look strange and act strange, yet they are human. What appears to be may not be so, yet the duck test often passes for reality. When something behaves and looks like it, we can assume it is at least at a base level. This runs into problems in quantum physics because what appears to be so can change; a particle can be both a particle and a wave. If we apply this to the larger world, a duck can be a duck and something else entirely. The duck has a wave property, too; it can be multiple things in a way.
Why do I discuss the duck test? Because it’s the most important test, we can use it to determine the truth, at least in part. If we wonder whether it’s something or not that at all, we can test whether it looks the part, plays the part, and acts like the part. In this sense, it matters more whether we believe it’s a believable act than anything. The duck test fails sometimes, especially when you’re being deceived by those who want you to fail the duck test, however often it’s correct. I currently use this principle to determine fact from fiction, even if it takes me to strange and terrifying conclusions.