If I gave you free reign of the larder, and asked you to prepare a good meal, what would you cook? Something rich and creamy, to help me get through this long week? Or a vegetable laden pie, to help me get my 5-a-day?
It depends on your definition of good, right? Meals are not objectively good or bad. They are just combinations of ingredients. So, ultimately, which ingredients you choose, and how you put them together into a meal, depends only on your specific goals.
It’s the same with algorithms, said Cathy O’Neil, in this talk given at the RSA, last year:
Algorithms are ways of putting data together to achieve a certain goal. They are not objectively good or bad. Which data they draw on, and how they put them together, depends on the goals of the person that built the algorithm (or, rather, of who is paying for it).
So, when someone says that algorithms are objective; or that they are just processing data, remember the ‘good meal’. Algorithms, like meals, are opinions combined in a formula.