Constraints can be built into the activity or experience:
"If the escape key is defined as a self-destruct mechanism, for instance, the constrint against pressing it in the course of flying a mimetic spaceship is intrinsic to the action." 10

"Constraints should be applied without shrinking our perceived range of freedom of action: Constraints ahould limit, not what we can do, but what we are likely to think of doing. Such implicit constrainsts, when successful, eliminate the need for explicit limitastions on our behavior. Context is the most effective medium for establishing implicit constraints."11

And within context, the character of the agents we put in the environment can help within the constraining process - for eample a grumpy or capricious agent that throws a tantrum covering something the computer doesn't understand - thus little by little we train the user in how to manipulate the agents successfully, gaining control of them! (Given that it's much easier to train the fuzzy human than program the fuzzy computer.)

Laurel suggests that basically people are very willing to "suspend disbelief" for the sake of an art/entertainment experience. Especially in cases where the potentially "dangerous" can be experienced in safety.

"People respect the limits of the mimetic world by refraining from introducing new potential into it (for instance, avoiding words or actions that the system is unlikely to know about). In exchange for this complicity, people experience increased potential for effective agency, in worlds in which the causal relations among events are not obscured by the randomness and noise characteristic of open systems (like real life)."12