History [of concept of mathematics]
Babylonian and Egyptian mathematics emphasized arithmetic and the idea of explicit calculation. But Greek mathematics tended to focus on geometry, and increasingly relied on getting results by formal deduction. For being unable to draw geometrical figures with infinite accuracy this seemed the only way to establish anything with certainty. And when Euclid around 330 BC did his work on geometry he started from 10 axioms (5 "common notions" and 5 "postulates") and derived 465 theorems. Euclid's work was widely studied for more than two millennia and viewed as a quintessential example of deductive thinking. But in arithmetic and algebra—which in effect dealt mostly with discrete entities—a largely calculational approach was still used. In the 1600s and 1700s, however, the development of calculus and notions of continuous functions made use of more deductive methods. Often the basic concepts were somewhat vague, and by the mid-1800s, as mathematics became more elaborate and abstract, it became clear that to get systematically correct results a more rigid formal structure would be needed.
The introduction of non-Euclidean geometry in the 1820s, followed by various forms of abstract algebra in the mid-1800s, and transfinite numbers in the 1880s, indicated that mathematics could be done with abstract structures that had no obvious connection to everyday intuition. Set theory and predicate logic were proposed as ultimate foundations for all of mathematics (see note below). But at the very end of the 1800s paradoxes were discovered in these approaches. And there followed an increasing effort—notably by David Hilbert—to show that everything in mathematics could consistently be derived just by starting from axioms and then using formal processes of proof.
Gödel's Theorem showed in 1931 that at some level this approach was flawed. But by the 1930s pure mathematics had already firmly defined itself to be based on the notion of doing proofs—and indeed for the most part continues to do so even today (see page 859). In recent years, however, the increasing use of explicit computation has made proof less important, at least in most applications of mathematics.