- The origin of Keynes' ideas on uncertainty
- Shackle and Shafer (Evidence Theory)
- Deciding not to decide
- Recognition of innovations by means of neural networks
- The Garbage Can model of organizational decision-making
- Coordination, routines and learning in organizations
- On the maximum size of human groups
- Industrial clusters and networks
The origin of Keynes' ideas on uncertainty
In the second half of the XIX century Johannes von Kries,
a physiologist who was applying probability theory to the
evaluation of the effectiveness of new drugs, realised that the
computation of probability distributions depends on the
classification of symptoms and pathologies into diseases.
Confronted with a setting where the crucial uncertainty was the
very definition of "events" by the experimenter, von Kries
developed the logical foundations of a probability theory where the
ubjectivity of mental representations may impair the
possibility of assigning numerical values to probabilities.
With a series of distortions and misunderstandings, von Kries's
ideas passed on to Keynes and formed the core of his
economics.
Shackle and Shafer
In the 1950s the economist George Shackle outlined the features of
a decision theory that would account for human behavior in
the face of unforeseen contingencies. In the 1970s the
mathematician Glenn Shafer initiated Evidence Theory, that
extends and formalizes many of Shackle's intuitions. The
prototypical situation of Evidence Theory is not a gambler
throwing dice, but a judge or detective evaluating testimonies. In
businesses like in detective stories, it is crucial to take account
of unforeseen contingencies. My work on this subject consists of
bridging between Shackle and Shafer, illustrating the principles of
Evidence Theory to social scientists.
Deciding not to decide
Liquidity preference, so relevant for investment decision making
and credit rationing, is an instance of deciding not to make any
decision. This is not an option to be evaluated with respect
to other alternatives, but rather stems from recognizing that novel
contingencies disrupted any confidence in previously held
mental models. Thus, deciding not to decide can be seen as
originating from too intricate cognitive maps caused by unexpected
causal relations. By means of a computational model of the
intricacy of cognitive maps it is possible to simulate visionary
investment decisions, wait-and-see attitudes, the arousal of
confidence and its disruption.
Recognition of innovations by means of neural
networks
Investing in novel fields requires the ability of recognizing the
potentiality of innovations. Indeed, much of the difference between
successful amd unsuccessful firms depends on this.
Recognition of innovation is an instance of pattern recognition,
which can be reproduced by neural networks. In particular,
Kohonen's self-organizing maps are able to reproduce the formation
of mental categories for classifying novel items.
The Garbage Can model of organizational
decision-making
The Garbage Can model by Cohen, March and Olsen is by far
the most influential model of organizational decision-making. It
was presented as a simulation model implemented on procedural code,
but it describes decision-making as resulting from the interactions
of four kinds of agents: 'participants', 'opportunities',
'solutions' and 'problems'.
By implementing the Garbage Can model as an agent-based model we
have been able to derive its most interesting properties from first
principles, rather than encoding them explicitely as in the
original version. Furthermore, the greater clarity imposed by the
agent-based representation suggested a deeper understanding of the
model, its limits and its implications.
Coordination, routines and learning in
organizations
Organizations can be seen as connectionist systems amenable to be
studied by means of agent-based models and related formalisms. This
way of looking at organizations yields interesting insights so far
it concerns the emergence of routines, the features of
organizational learning and the boundaries between markets and
organizations.
On the size of human groups
By evaluating the cognitive stress of the members of
groups of different size and interaction structure, I arrive at
maximum thresholds of 5-6 persons for plain groups, 15 persons for
centered groups and 50-100 persons for federative groups,
respectively. Beyond each threshold, a group either dissolves or
modifies its structure according to the requirements of the
subsequent class. My theory is in accord with empirical findings
from psychology, anthropology and organization science.
Essentially, this is an instance of the concept of bounded
rationality. It has a number of consequences for the management of
teams, committees and communities of practice. Furthermore, it
explains several puzzles in sociology and macroeconomics.
Industrial clusters and networks
Agent-based simulations and statistical network analysis are very
powerful tools to study the emergence and evolution of local
networks in terms of the structure of relationships between
component firms, entrepreneurs, researchers and other economic
agents. Among else, the impact of information and communication
technologies and the formation of virtual clusters can be
investigated.