For today’s entry in the Leap Day 2016 Mathematics A To Z I have an actual request from from Elke Stangl. I’d had another ‘c’ request, for ‘continued fractions’. I’ve decided to address that by putting ‘Fractions, continued’ on the roster. If you have other requests, for letters not already committed, please let me know. I’ve got some letters I can use yet.
An old joke says a mathematician’s job is to turn coffee into theorems. I prefer tea, which may be why I’m not employed as a mathematician. A theorem is a logical argument that starts from something known to be true. Or we might start from something assumed to be true, if we think the setup interesting and plausible. And it uses laws of logical inference to draw a conclusion that’s also true and, hopefully, interesting. If it isn’t interesting, maybe it’s useful. If it isn’t either, maybe at least the argument is clever.
How does a mathematician know what theorems to try proving? We could assemble any combination of premises as the setup to a possible theorem. And we could imagine all sorts of possible conclusions. Most of them will be syntactically gibberish, the equivalent of our friends the monkeys banging away on keyboards. Of those that aren’t, most will be untrue, or at least impossible to argue. Of the rest, potential theorems that could be argued, many will be too long or too unfocused to follow. Only a tiny few potential combinations of premises and conclusions could form theorems of any value. How does a mathematician get a good idea where to spend her time?
She gets it from experience. In learning what theorems, what arguments, have been true in the past she develops a feeling for things that would plausibly be true. In playing with mathematical constructs she notices patterns that seem to be true. As she gains expertise she gets a sense for things that feel right. And she gets a feel for what would be a reasonable set of premises to bundle together. And what kinds of conclusions probably follow from an argument that people can follow.
This potential theorem, this thing that feels like it should be true, a conjecture.
Properly, we don’t know whether a conjecture is true or false. The most we can say is that we don’t have evidence that it’s false. New information might show that we’re wrong and we would have to give up the conjecture. Finding new examples that it’s true might reinforce our idea that it’s true, but that doesn’t prove it’s true.
For example, we have the Goldbach Conjecture. According to it every even number greater than two can be written as the sum of exactly two prime numbers. The evidence for it is very good: every even number we’ve tied has worked out, up through at least 4,000,000,000,000,000,000. But it isn’t proven. It’s possible that it’s impossible from the standard rules of arithmetic.
That’s a famous conjecture. It’s frustrated mathematicians for centuries. It’s easy to understand and nobody’s found a proof. Famous conjectures, the ones that get names, tend to do that. They looked nice and simple and had hidden depths.
Most conjectures aren’t so storied. They instead appear as notes at the end of a section in a journal article or a book chapter. Or they’re put on slides meant to refresh the audience’s interest where it’s needed. They are needed at the fifteen-minute park of a presentation, just after four slides full of dense equations. They are also needed at the 35-minute mark, in the middle of a field of plots with too many symbols and not enough labels. And one’s needed just before the summary of the talk, so that the audience can try to remember what the presentation was about and why they thought they could understand it. If the deadline were not so tight, if the conference were a month or so later, perhaps the mathematician would find a proof for these conjectures.
Perhaps. As above, some conjectures turn out to be hard. Fermat’s Last Theorem stood for four centuries as a conjecture. Its first proof turned out to be nothing like anything Fermat could have had in mind. Mathematics popularizers lost an easy hook when that was proven. We used to be able to start an essay on Fermat’s Last Theorem by huffing about how it was properly a conjecture but the wrong term stuck to it because English is a perverse language. Now we have to start by saying how it used to be a conjecture instead.
But few are like that. Most conjectures are ideas that feel like they ought to be true. They appear because a curious mind will look for new ideas that resemble old ones, or will notice patterns that seem to resemble old patterns.
And sometimes conjectures turn out to be false. Something can look like it ought to be true, or maybe would be true, and yet be false. Often we can prove something isn’t true by finding an example, just as you might expect. But that doesn’t mean it’s easy. Here’s a false conjecture, one that was put forth by Goldbach. All odd numbers are either prime, or can be written as the sum of a prime and twice a square number. (He considered 1 to be a prime number.) It’s not true, but it took over a century to show that. If you want to find a counterexample go ahead and have fun trying.
Still, if a mathematician turns coffee into theorems, it is through the step of finding conjectures, promising little paths in the forest of what is not yet known.
12 thoughts on “A Leap Day 2016 Mathematics A To Z: Conjecture”
Thanks :-) So you say that experts’ intuition that might look like magic to laymen is actually pattern recognition, correct? (I think I have read about this in pop-sci psychology books) And if an unproven theorem passes the pattern recognition filter it is promoted to conjecture.
I think that there is a large aspect of it that’s pattern recognition, yes. But some of that may be that we look for things that resemble what’s already worked. So, like, if we already have a theorem about how a sequence of real-valued functions converges to a new real-valued function, then it’s natural to think about variants. Can we say something about sequences of complex-valued functions? If the original theorem demanded functions that were continuous and had infinitely many derivatives, can we loosen that to a function that’s continuous and has only finitely many derivatives? Can we lose the requirement that there be derivatives and still say something?
I realized at one point while taking real analysis in grad school that many of the theorems we were moving into looked a lot like what we already had with one or two variations, and could sometimes write out the next theorem almost by rote. There is certainly a kind of pattern recognition at work here, though sometimes it can feel like playing with the variations on a theme.
LikeLiked by 1 person
Yes, I agree – I meant pattern recognition in exactly this way, in a very broad way … searching for a similar pattern in your own experiences, among things you have encountered and that worked. I was thinking in general terms and comparing to other skills and expertise, like what makes you successful in any kind of tech troubleshooting. It seems that you have an intuitive feeling about what may work but actually you draw on related scenarios or aspects of scenarios we had solved.
I understood all that! I definitely deserve a prize … I am no mathematician … And I enjoyed every word! I love your use of English.
You’re most kind, thank you. I’m glad you enjoyed.
LikeLiked by 1 person
If you’ve nothing for Q, what about Quadratic Equations … though I start twitching whenever I think about them!
I’m sorry to say Q already got claimed, by ‘quaternion’. But P got ‘polynomial’, which should be close enough to quadratic equations that there’s at least some help there.
LikeLiked by 1 person