Picture a postage chart: letters up to 100 g cost $1.50, and heavier letters up to 250 g cost $2.80. Weigh any letter and the chart gives exactly one price, so price is a function of weight. Many weights share a price, which is allowed. What would break the dependency is one weight with two possible prices. Tables of pairs are tested the same way: in {(1,3),(2,5),(3,7),(2,9)} the input 2 appears with both 5 and 9, so y is not a function of x.
Direction matters. Price is a function of weight, but weight is not a function of price: a $2.80 letter could weigh 120 g or 240 g. Equations behave the same way. In x=y2, x is a function of y, because each y gives one x. But y is not a function of x: at x=9, both y=3 and y=−3 work. On a graph, a vertical line that crosses the curve twice shows an input with two outputs.
Dependency is not causation. A model such as "sales =12T−100 cones" makes sales a function of the temperature T, and the same equation rearranged makes temperature a function of sales. Both directions are valid functions, yet selling ice cream does not heat the weather. Whether one quantity causes another is a question about the world, answered by experiments, not by the form of an equation.
Think of it like this. A functional dependency is like a library catalogue: each call number points to exactly one spot on the shelves, so knowing the number tells you where to look. Many books can share a subject, but a call number leading to two places would make the catalogue useless. The catalogue says nothing about why a book sits where it does.
Core idea. Functional dependency means fixing the input forces a single determined output.