Formal Proof—The Four- Color Theorem 论文

摘要

The Tale of a Brainteaser Francis Guthrie certainly did it, when he coined his innocent little coloring puzzle in 1852. He managed to embarrass successively his mathematician brother, his brother’s professor, Augustus de Morgan, and all of de Morgan’s visitors, who couldn’t solve it; the Royal Society, who only realized ten years later that Alfred Kempe’s 1879 solution was wrong; and the three following generations of mathematicians who couldn’t fix it [19]. Even Appel and Haken’s 1976 triumph [2] had a hint of defeat: they’d had a computer do the proof for them! Perhaps the mathematical controversy around the proof died down with their book [3] and with the elegant 1995 revision [13] by Robertson, Saunders, Seymour, and Thomas. However something was still amiss: both proofs combined a textual argument, which could reasonably be checked by inspection, with computer code that could not. Worse, the empirical evidence provided by running code several times with the same input is weak, as it is blind to the most common cause of “computer” error: programmer error. For some thirty years, computer science has been working out a solution to this problem: formal program proofs. The idea is to write code that describes not only what the machine should do, but also why it should be doing it—a formal proof of correctness. The validity of the proof is an objective mathematical fact that can be checked by a different program, whose own validity can be ascertained empirically because it does run on many inputs. The main technical difficulty is that formal proofs are very difficult to produce,