I had a great deal of fun last summer with an A To Z glossary of mathematics terms. To repeat a trick with some variation, I called for requests a couple weeks back. I think the requests have settled down so let me start. (However, if you’ve got a request for one of the latter alphabet letters, please let me know. There’s ten letters not yet committed.) I’m going to call this a Leap Day 2016 Mathematics A To Z to mark when it sets off. This way I’m not committed to wrapping things up before a particular season ends. On, now, to the start and the first request, this one from Elke Stangl:
Mathematics is built of arguments. Ideally, these are all grounded in deductive logic. These would be arguments that start from things we know to be true, and use the laws of logical inference to conclude other things that are true. We want valid arguments, ones in which every implication is based on true premises and correct inferences. In practice we accept some looseness about this, because it would just take forever to justify every single little step. But the structure is there. From some things we know to be true, deduce something we hadn’t before proven was true.
But where do we get things we know to be true? Well, we could ask the philosophy department. The question’s one of their specialties. But we might be scared of them, and they of us. After all, the mathematics department and the philosophy department are only usually both put in the College of Arts and Sciences. Sometimes philosophy is put in the College of Humanities instead. Let’s stay where we were instead.
We know to be true stuff we’ve already proved to be true. So we can use the results of arguments we’ve already finished. That’s comforting. Whatever work we, or our forerunners, have done was not in vain. But how did we know those results were true? Maybe they were the consequences of earlier stuff we knew to be true. Maybe they came from earlier valid arguments.
You see the regression problem. We don’t have anything we know to be true except the results of arguments, and the arguments depended on having something true to build from. We need to start somewhere.
The real world turns out to be a poor starting point, by the way. Oh, it’s got some good sides. Reality is useful in many ways, but it has a lot of problems to be resolved. Most things we could say about the real world are transitory: they were once untrue, became true, and will someday be false again. It’s hard to see how you can build a universal truth on a transitory foundation. And that’s even if we know what’s true in the real world. We have senses that seem to tell us things about the real world. But the philosophy department, if we eavesdrop on them, would remind us of some dreadful implications. The concept of “the real world” is hard to make precise. Even if we suppose we’ve done that, we don’t know that what we could perceive has anything to do with the real world. The folks in the psychology department and the people who study physiology reinforce the direness of the situation. Even if perceptions can tell us something relevant, and even if our senses aren’t deliberately deceived, they’re still bad at perceiving stuff. We need to start somewhere else if we want certainty.
That somewhere is the axiom. We declare some things to be a kind of basic law. Here are some thing we need not prove true; they simply are.
(Sometimes mathematicians say “postulate” instead of “axiom”. This is because some things sound better called “postulates”. Meanwhile other things sound better called “axioms”. There is no functional difference.)
Most axioms tend to be straightforward things. We tend to like having uncontroversial foundations for our arguments. It may hardly seem necessary to say “all right angles are congruent”, but how would you prove that? It may seem obvious that, given a collection of sets of things, it’s possible to select exactly one thing from each of those sets. How do you know you can?
Well, they might follow from some other axioms, by some clever enough argument. This is possible. Mathematicians consider it elegant to have as few axioms as necessary for their work. (They’re not alone, or rare, in that preference.) I think that reflects a cultural desire to say as much as possible with as little work as possible. The more things we have to assume to show a thing is true, the more likely that in a new application one of those assumptions won’t hold. And that would spoil our knowledge of that conclusion. Sometimes we can show the interesting point of one axiom could be derived from some other axiom or axioms. We might replace an axiom with these alternates if that gives us more enlightening arguments.
Sometimes people seize on this whole axiom business to argue that mathematics (and science, dragged along behind) is a kind of religion. After all, you need to have faith that some things are true. This strikes me as bad theology and poor mathematics. The most obvious difference between an article of faith and an axiom must be that axioms are voluntary. They are things you assume to be true because you expect them to enlighten something you wish to study. If they don’t, you’re free to try other axioms.
The axiom I mentioned three paragraphs back, about selecting exactly one thing from each of a collection of sets? That’s known as the Axiom of Choice. It’s used in the theory of sets. But you don’t have to assume it’s true. Much of set theory stands independent of it. Many set theorists go about their work neither committing to the idea that it’s true or that it’s false.
What makes a good set of axioms is rather like what makes a good set of rules for a sport. You do want to have a set that’s reasonably clear. You want them to provide for many interesting consequences. You want them to not have any contradictions. (You settle for them having no contradictions anyone’s found or suspects.) You want them to have as few ambiguities as possible. What makes up that set may evolve as the field, or as the sport, evolves. People do things that weren’t originally thought about. People get more experience and more perspective on the way the rules are laid out. People notice they had been assuming something without stating it. We revise and, we hope, improve the foundations with time.
There’s no guarantee that every set of axioms will produce something interesting. Well, you wouldn’t expect to necessarily get a playable game by throwing together some random collection of rules from several different sports, either. Most mathematicians stick to familiar groups of axioms, for the same reason most athletes stick to sports they didn’t make up. We know from long experience that this set will give us an interesting geometry, or calculus, or topology, or so on.
There’ll never be a standard universal set of axioms covering all mathematics. There are different sets of axioms that directly contradict each other but that are, to the best of our knowledge, internally self-consistent. The axioms that describe geometry on a flat surface, like a map, are inconsistent with those that describe geometry on a curved surface, like a globe. We need both maps and globes. So we have both flat and curved geometries, and we decide what kind fits the work we want to do.
And there’ll never be a complete list of axioms for any interesting field, either. One of the unsettling discoveries of 20th Century logic was of incompleteness. Any set of axioms interesting enough to cover the ability to do arithmetic will have statements that would be meaningful, but that can’t be proven true or false. We might add some of these undecidable things to the set of axioms, if they seem useful. But we’ll always have other things not provably true or provably false.