>Let me explain a specific example: If a person A cheats person B out of his money, it affects me negatively. It poses a risk in the future that person A, being

>Let me explain a specific example: If a person A cheats person B out of his money, it affects me negatively. It poses a risk in the future that person A, being a cheater, will repeat this many times and eventually will have enough money to cheat me. (Both B and me will agree to be cheated by A at the time when we will be cheated, yet it won't be our preference. Kind of like getting a forced move by check in chess.) It depends what you mean by "cheat". If A lies, then A potentially committed fraud or the like, and can be persecuted. If A just took advantage of B's gullibility (sold him some great swampland or the like), this wouldn't pose any risk to you unless you were gullible too. >Which is a bullshit excuse. We have mathematical capabilities to deal with these kinds of problems - probability, fuzzy logic, whatever. It's just laziness of economists to show that the free market assumptions hold under these new assumptions. It's a not about a lack of mathematics, it's an empistemological issue: how can these assumptions be justified? Any qualitative model necessarily requires more assumptions. As all moral assumptions/statements are ultimately arbitrary due to the regress problem I mentioned previously, fewer assumptions means less arbitrarity. >In particular, we can model things like this in computers. You can't model things like this in computers until you can model human brains in computers. Until then, you have to make arbitrary assumptions that some kind of agent in particular model morally approximates an actual human, which is yet another assumption that must be justified. >It's a nice curiosity to have free markets work under some weird, unrealistic assumptions, but it's rather useless in the real world. You're missing the point that it's about how "work" is defined. If you define it by allowing people to maximise their ordinal utility at each given moment in time, then they "work" by default. If you define it by some other criteria, then how much they work depends on how similar those criteria are to maximising momentary ordinal utility.