In the same way that a chef judges a large pot of soup by tasting just one spoonful. Bearing in mind that the soup is well mixed, and that one spoonful is “representative”, then it is quite enough to try only that amount.
Public opinion polls are conducted on the same principle: achieving a representative sample is very similar to stirring a soup. Unscientific research is like a bowl of soup that has not been stirred.
The cook could drink a large amount of liquid from the top and get a wrong picture if some of the ingredients have sunk to the bottom. Just as the trick in tasting a soup is to stir it well rather than trying a large amount, so the essence of scientific research is to ensure a representative sample, not as large as possible.