c
 



  








ip { 

 


Herbert Simon used bounded rationality to explain the disparity between the assumed and
observed attributes of economic man. He used the science of artificial as a link between nature
and its artifacts. He also supported computer intelligence as a source of knowledge. The author
supports Simon¶s view and argues that markets need not be built from individual behavior if one
considers them as human artifacts. He also states that computer simulations have already shown
that allocative efficiency ± a key characteristic of market outcomes ± is largely independent of
individual behavior.

  



   Substantive rationality assumes that an agent chooses
the most preferred option from its opportunity set. However, substantive rationality does not take
into account the efforts that an agent has to put in order to build its opportunity set. Simon
argued that an agent continues searching for an option until it finds one that satisfies its level of
aspiration about its welfare. Simon referred to this as procedural rationality. Economics needs
substantive rationality in order to come up with predictive models. However, such predictive
models do not help understand the phenomena since they are based on invalid assumptions.



 
 
  The author states that we synthesize artifacts from elements in
order to perform functions and attain goals. The intent of its creator (inner environment)
distinguishes an artifact from nature (outer environment). The performance of the artifact
depends on the interface between the two environments. A twig is a natural artifact used by
chimpanzees to extract food out of a termite hill, whereas titanium alloy is a manufactured
artifact used to build supersonic aircrafts. A chimpanzee could also use a straw or a thin bone to
extract food. On the other hand, a twig could also be used to extract honey from a honey-comb.
The assumptions of economics do not take into account the roles of inner and outer environments
of an artifact.

c
  

 Considering markets as social artifacts helps to delineate the boundary
between their inner and outer environments. Market structures or rules lie inside, while the
agents, defined by their endowments, preferences and decision rules lie outside. The inner
environment distributes messages and allocates resources based on the messages sent by the
participants. This is similar to a supermarket in which a seller sends messages in the form of
price labels and buyers send messages by adding appropriate quantities to their shopping carts.
The interaction between the inner and the outer environment leads to the desirable outcome of
transfer of money from the buyer to the seller. The author states that only the critical features of
the environment should be taken into consideration while designing the artifact. While designing
a car seat for infants, it is important to look into the safety aspect rather than worry about the
texture of the seat material. The inner environment remains largely unnoticed by most users,
usually attracting attention only when it is stretched beyond its limits and the outcome fails to
stand in the desired relationship with the outer environment. Neoclassical economics, while
focusing on allocative efficiency, price and distribution of gains from trade, assumes simple,
idealized forms for decision-making processes in the outer environment.

º!"!
  c
 Artificial intelligence and computer simulation

models can be used for modeling the behavior of systems for the outer environments of markets
that are unobservable in the field or difficult to compute for neoclassical economists. In case of
markets, even if the behavior of traders was well-defined, their interactions can be quite complex
and only a laboratory simulation can help characterize the market outcomes of a variety of
existing and new market designs. Simulations can also help focus only on the relevant part while
ignoring the part that is not of interest to us. Thus, they can help reveal the organization of the
inner environment of a market independent of the decision-making behavior of individuals who
constitute the outer environment. This can be particularly useful for Walrasian demand function
that is theoretically derived and is never observed in practice.

 #$
!


 In order to test whether stock market crashes could be
blamed on program trading; the author designed a course on double auction markets and the
structure of trading strategies used in them. In double auctions, buyers compete by raising their
bids and sellers compete by lowering their prices. The author wrote the robot software for trading
whereas students could trade themselves or write their own software rules in order to trade in the
market. Students were rewarded in the form of good grades in case of a profit. When students
traded amongst themselves, after some initial variability, double auction prices and allocations in
classical market environments quickly settled down in the neighborhood of the predictions
theory. When the human traders were replaced by the programs written by the students, volatility
persisted for longer time and equilibrium was attained after several time periods. Over a period
of time, although students¶ programming skills improved, price and allocative efficiency in the
markets surprisingly did not attain 100 percent at equilibrium. Finally when the author wrote the
software program for trading, he simply used a uniformly distributed random number between 0
and the actual willingness to pay in order to make a bid. Also, for a seller the software program
used a similar uniformly distributed random number between the minimum and maximum price
for selling the product. The preceding two conditions ensured that the ³Zero Intelligence´ trader
of the author did not trade at a loss but kept spewing new proposals. Although variability was
high, surprisingly the prices converged at equilibrium over a long period of time. This shows the
surprising result that equilibrium can be attained even if the economic agent isn¶t trying to
maximize profits. Also, the allocative efficiency of the market was almost the same as the
software programs that the students had written despite no real attempts to maximize the profits.
The no-loss constraint leads to the same allocative efficiency as the profit-maximizing
assumption. Simon had developed and validated the theory of bounded rationality for individual-
decision making. The author¶s experiments showed that as long as the traders do not make losses
by making stupid trades, even the assumption of bounded rationality is too strong and not
entirely necessary for the efficiency attained by the market. A weak form of ³pursuit of self-
interest´ with random choices in one¶s opportunity set is sufficient for attaining allocative
efficiency.

Markets can attain efficiency even if the individuals do not know how to. This is because
efficiency of markets is primarily a function of their rules. As the market demand and supply
conditions change, the expected loss of efficiency has an upper bound. This bound is generated
by a trade-off between the magnitude and the probability of efficiency loss associated with the
displacement of intra- by extra-marginal traders. This market-level trade-off is independent of
the individual trade-off between a proposal¶s profit and its probability of being accepted. Double
auctions are more efficient than one-sided auctions such as sealed-bid auctions because the
former require more conditions to be fulfilled for an inefficient trade to occur. Such auctions
have lower probability of allowing the extra-marginal traders to displace the intra-marginal
traders; other things being the same, call markets are favored over continuous auctions. On the
other hand, efficiency is higher if traders can observe market data (e.g., call auctions in which the
bids and asks are made public in real time, as compared to call markets in which they are not
made public). Single-market findings about double auctions can easily be generalized to a set of
multiple interlinked markets. The author believes that a combination of the Zero Intelligence
model and the Walrasian model would do a better job than either can do alone in helping us
understand the markets.