THE SAD MODEL
According to my S.A.D. model of agent characteristics, every agent has 3 characteristics with respect to an event:
Sensor characteristic- ability or cost to get information from their environment about the probability of an outcome of an event.
Actuator characteristic- ability or cost to influence the probability of an outcome.
Demander characteristic- utility or payoff associated with each outcome.
Here is a copy of a paper of mine I brought with me and shared with some individuals at The Institute For Humane Studies' Workshop in Experimental Economics at Bryn Mawr College in July, 2006.
The S.A.D. Model of Agent Characteristics
The particular choice of mathematical model can be for example, a penny mechanism, a vote mechanism, an urn mechanism, a markov mechanism. Other mechanisms are possible, these are only examples of how the S.A.D. model could be applied mathematically.
What I like about the SAD model is that firstly, it is a positive not a normative economic model. It does not take sides. It can be used to run experinomic laboratory games that incorporate prediction markets. The agents are given private information about their own SAD characteristics. This substitutes for example in place of the double oral auction reservation prices given to each player, as famously developed by Nobel Prize winner Vernon Smith. But unlike commodity markets for pounds of apples or quantity of porkbellies, event markets like political markets in the IEM cannot be tackled experimentally using reservation prices. That is why the SAD model is useful. I will again repeat what I said at Bryn Mawr: To the best of my knowledge at this point in time, this is the first and only available model/framework that allows an experimental approach to the study of prediction markets. The SAD model allows the study of a controlled, repeatable artificial event with a probability that is known to the researcher.
What the SAD model does not do is define strategies for the agents. In order to perform a pure simulation using the SAD model, the simulated agents will be defined in terms of their SAD characteristics, but also must be endowed with behavioral strategies.
© 2006 Jonathan Moses Katcher, All Rights Reserved.
