ABSTRACT
Just as all real-world rationality is bounded rationality, all real-world agency is bounded agency: resource constraints make performance tradeoffs inevitable, and consequently any agential architecture will perform poorly in some environments. Ideal agency could be exhibited only by an agent that shifted its architecture to match its current environment whatever that was. Success concepts for agency in general, and the determination that agents exhibit in particular, must be formulated with a concrete agential architecture in view. Consequently, we are unable to formulate success concepts for ideal agency; thus, theory of agency should not take the ideal agent as its reference point.
