threaten, have only opportunity cost of them is model after firing, one princle - no rematching, is learning relational specific or --- ⭐️ONLY OF SUBSET OF ACTION IS AVAILABLE each eaction is productive even if the agent does the wrong thing, agent should be fired common except, elicit information (agent knows all) - not about experimentation the moment agent choose productive action, then on the prinipal label as productive (knowledge) provent to be successful, - choose an action and the env. has probability with pro - ⭐️relation is becoming more valuable - choose productive actions more often - as long as we experiment, the relation is fragile (to get the agent to experimnt) if you have money (use money to incentivize) conflict to arise in equilibrium path () - need private knwoledge principal - availability of the action is where 🍀luck comes in - different - principal (productive, unproductive) - secon bullet produce ppd, - relation and routines are all - pitched itself as models of ppd - intuition is very specific (firing business) - NOT experiment but elicitation of knowledge - as agent already knows which is productive --- - sum of the player payoff from tomorrow (continuation value given ) # multiple