That's kinda my job! :-)

Thing is, many of these algos assume an agent-ignorance, a lack of total info, so that agents (sometimes human) act with incomplete knowledge, eg. the Prisoner's Dilemma is only a dilemma because the two people cannot communicate, so each has partial knowledge.

But on a blockchain, one can extract a lot of data so that, in theory, everybody acts from a POV of full-knowledge, not partial. This means decisions can be taken with global consequences in mind. This makes it easier to form allegiances on-chain that can break some of the purity of the algorithms.

I've said this many times, and perhaps takes some time to really sink in to the public: that after writing an algorithm, when you try to optimise its effects, it can end up doing something very different to the original intention.

Just as the 2-person cake sharing is easy, but many-people is hard - not just the process but also potential collusion among cake-lovers.