I think there is a subtle tradeoff between the benefits of impartiality of algorithms and the risks of their potential catastrophic failure, either for one participant (simple example: a person losing their private keys and thus access to their cryptocurrency), or the system as a whole if someone finds a clever way to hack it or to bring it down.

Automated mechanisms and systems might fail in the face of some black swan, a dramatic and unexpected change in the environment.

At least until we have AGI (or, perhaps, especially after that) I would be very uneasy with totally automated platforms and much rather prefer platforms with weakly-defined, "libertarian" protocols for agent cooperation. For example, I would much prefer a reputation to be computed by agent themselves based on some open history data (e. g. stored in blockchain) rather than reputation stored in the blockchain itself.