This is a question that has been at the center of neoclassical economics, and neoclassical economics is the only school of thought that has faced it squarely and given a logically consistent answer to it. I say that the neoclassical answer is consistent, not that it is correct, because there are some problems with it. But since it is the only logically consistent answer in the many traditions of economics, it must be our starting point.
Neoclassical economics begins from the assumption that people are absolutely rational. The idea is that, facing any decision or choice, a human being makes the choice that advances his or her own purposes most efficiently. That is, people maximize their own net benefits. Now, turn that idea around. If we observe the choices a person makes, we can infer what his or her purposes are, what benefits he or she derives from the choice. When a person chooses this rather than that, we can infer that the (absolutely rational) person gets more benefit from this than that. This is the principle of revealed preference: the preferences (purposes, subjective benefits) of absolutely rational people are revealed by the choices they make, according to neoclassical economics.
So, what would a rational Economian pay for a certain quantity of machines? The answer is that she would pay an amount of food that would give her just the same benefit as the quantity of machines! We can measure the total benefit from machine production in terms of food. The measure of benefit is the amount of food that the citizens would be willing to trade for the given quantity of machines.
If we think in terms of utility, the benefits from consuming a machine would be measured as follows: suppose that consumption of one more machine gives the consumer U* "units of utility." Then the benefit, measured in units of food, is the amount of additional food required to give the same U* "units of utility." (In Chapter 5, we measured benefits in units of money. It would be the amount of money necessary to purchase goods that would give U* units of utility -- the same sort of measure, but in dollars rather than bushels of food. But in this chapter we are trying to get "behind the veil of money," so we measure benefits instead in units of food).
We needn't think in terms of utility, though. Much more generally, the measure of the benefits from another machine, in units of food, is the amount of additional food necessary to make the person just as well off as the additional machine would. Utility is just a way of illustrating the method of measurement.
