Here - Agents Lab - University of Nottingham

Fig. 3. The broker.buy protocolThe ordering of the core agents in the list above reflects the order in whichthey should be interacted with, thus creating a clearly defined implementationsequence. The tasks are designed so that successfully completing a task will bedependent on previous tasks being completed first. Although there are no technicalrestrictions on the order in which participants may choose to implementthe tasks, these dependencies discourage them from writing their implementationsin a different order. For example, selling items to bidders is impossiblebefore interacting with the Auctioneer to buy properties. However, these cannotbe bought prior to interacting with the stock market, as the minimum propertyprice is deliberately set to be higher than the Player’s initial capital. Similarly,the advice of the Guru is useless unless an agent can use it when interactingwith the Stockbroker, and it is impossible to buy or sell stocks without havingpreviously opened an account with the banker.To reward active agents, the stock price calculation mechanism is intentionallyartificial, in that the price of stocks always rises. This rewards developers forimplementing features, to the detriment of idle agents. If stock prices can fall, an91