You are on page 1of 4

National University of Colombia

Facultad de minas
Department of computing science and decision
MSc - Systems Engineering
Decisiones bajo incertidumbre en las organizaciones

Medelln, marzo de 2014

Analysis of 50 years probabilistic decision


A brief review

Alejandro Gallegoa
Abstract

This document is a brief review of the 50 years probabilistic decision UK. We will see that
the science of Probabilistic decision analysis has a dierent views, combining probabilistic, llings,
preference, consequences, e.t.c.

Key words : probabilistic decision analysis, preferences, science, belief nets, trees.

1. Introduction
Probability decision analysis (PDA) should been see like a way to make better decision when we have
uncertainty of our problems, this uncertainty occurs when for instance we don't know all the variables that
models some reality. Always when we construct a model, we make assumption that reduce the problem
itself, in this way we can't assume that our solution or decision is the best solution. Probabilistic decision
analysis has his foundation in probability theory, this is a branch of mathematics that models the world
with variables that has a function associate, this function allow us to give a value for the variables. In
this sense we can say something like if we ip a coin, we have a probability p of obtain head, and (1-p)
of obtain tail. In other words the function take the variable X = side of a coin, and make a function

PX (x = head00 ),

this function achieves some properties that is well study with the Kolmogorov axioms.

When we deals with inferences of how is the behavior of the certain system, statisticians constructs
statistics in order to capture some information relevant to the person who make the decisions about the
system. This information has a probability model, with this in mind we can say that the person who
make a decision is biased in some sense by this statistics. Mean while the real problem is what to do
when we have a result. Thus, in the paper Fifty years of probabilistic decision analysis: a view from the
UK (Morton & Phillips 2009), they tried to recapitulate the main ideas of the PDA, begins with the
ideas of Leonard J. Savage who is considerate the person who break the history of PDA, consider also
the ideas of Frank P. Ramsey who believes that the strengths of a thought models the decision that we
make in dierent circumstances.

2. Who is rst
Often when we think if acts of a person are rational or senseless, we imagine a person with a huge
brain that consider all of the possibles variables and by the use of this variable he makes a decision. This
is true if the behaviors of a person were rational, but this is not the reality. Leonard Savage says starts
from a situation in which the world is in a number of possible states and the statistician has to make
one among a number of possible decisions, this introduce the notion of probability, where the decision is

a Cdigo: 1022369610. E-mail: jagallegome@unal.edu.co

A brief review

taking in situation that has uncertainty, and then says Savage proposes certain rules and then says that
a man is rational if he obeys these rules in his decision making. Here Savage assume that people think
in a rational way, but many of the recently works shows that the humans has biased and it's clear that
we need dierent approach to dealing with uncertainty and decisions. (Morton & Phillips 2009)
In the paper (Morton & Phillips 2009) propose the following theorem There exist unique probabilities
associated with each state and unique utilities associated with each consequence, this assumption makes
a lot of confusion because if we can break a problem in dierent states such that the intersection of the
states is the set empty, we can associate a probability function to every states, but in reality those states
are overlapping and it's impossible to take for sure that the states of a system is one in particular, hence
the state of a system can model like a superposition of dierent states breaking the Kolmogorov Axioms.
Because of that the utility function loose his meaning as a result of the superposition. In this theorem
also have  such that an action a is preferred to an action b if and only if the expected utility (the sum
of the products of the utilities associated with the consequence in each state and the probability of that
state) of action a is greater than that of action b., in this part of the theorem we can see how statistician
thinking in the early twentieth century, with the advanced of the theorems and studies of the human
behavior we nd that people don't think with a probability function in mind furthermore people is more
condent with alternatives that has a stronger connection with his belief as was thought by Ramsey.

3. The problem of take a decision


Ramsey propose a problem where a person have to make a decision to which way to go, and he decides
if he is taking the right way if he is condent with his beliefs. On account of the fact that we are more
condent in our belief we can walk more deep onto the eld without ask if we are in the right way, thus
Ramsey denes the utility of two consequences: arriving at the right destination which is given utility
r; and arriving at the wrong destination, which is given utility w (Morton & Phillips 2009, Pag 2), He
also denes a disutility function, this function obtain his value from the distances x to be traveled back
again.
Whenever

f (x) r w

because if f(x) is greater than the dierences of r-w, we don't want to come

back. If we assume that exist a probability function that models if we take the right way, we can assume
that the probability of taking the right way is p, and the probability of taking the wrong way is 1-p. If
we calculate the expected utility of continuing, we can do the following

w + p(r w),

EX (x) = p (r) + (1 p)(w) =


EX (x) = r f (x). Now we

and the expected utility of turning aside to ask directions is

can calculate a point of break


same, in terms of variables is

x0 where the expected utility of continuing


w + p(r w) = r f (x), if we solving for p
p=1

walk, and turning aside is the


obtain

f (x0 )
rw

(1)

Once we have dene a formula for obtain p, this change the structure of the decision, before we have
a subjective problem that we don't know how to solve, now we have a formula for obtain this strange
parameter p. But this parameter continue to have a subjective implication, because is the decision maker
who have to make a decision, and need to dene what is the value for w, r and the function f(x). This is
the way that Frank Plumpton Ramsey a british philosopher dealing with the relation between probability
and decisions.
In the paper (Morton & Phillips 2009, Pag 2) write the following theorem There exist unique utility
scores associated with each consequence such that a lottery a is preferred to a lottery b, if and only, if
the expected utility of a lottery a is greater than the expected utility of a lottery b.
This theorem indicate that the utility function helps us to obtain a way to dealing with preferences,
because each person has a preference of certain consequence, due to this we have that certain people
prefer to take the car without thinking in the future generation and we have a people prefer to take
the public transportation thinking in the future generation. This is more obvious when we ask a person

Decisiones bajo incertidumbre en las organizaciones (2014)

A brief review

why he choose some product over other product, and he answer that he prefer some product because he
expected more utility over other.
In the last decades, surge a new way to dealing with probabilities, with the help of Bayesian networks,
this networks allow us to simplify the writing of a model, in a way that every one who sees the model
understand the same. We rst need to dene what are the principal nodes that start the Bayesian network,
for example we can have a network that says what is the probability that given the grass is wet, in the
past has been raining or doing a sunny day. This can models as follows.

Raining

Sunnyday

W etgrass

4. Why the issue to learn PDA


In the beginning we can think to a PDA a procedure with certain steps that allows us to make
clear what is the problem itself, making explicit the variables that can become complex in the system.
What is more the decision-maker can make decision without fell bad, because he didn't make an analysis
exhaustive. However recently denition makes PDA a language allow us to capture on a model all the
complex thinking of the decision-maker. If we think that PDA is a language we can express all the ideas
that spring in natural language and translate to the new language for PDA.
(Morton & Phillips 2009, Pag 2) quotes a Tocher's critique, who said that we need to remember that
DA is a way of thinking but not a automatic procedure. In addition when we model a system always the
person who has the general idea of the system who take the decision, and over him is the responsibility
of taking a good or bad decision, is not the tool who is responsible.
It's common to think that people make to do decision in a probabilistic way, but (Morton & Phillips
2009, Pag 2) quotes that recently studies show us that this is no entirely true, if we search in psychology
we can nd that such person doesn't think with a utility or probability when he makes a decision, is more
like what is the preference of the person, with his biased thinking.

5. Conclusion
Probabilistic decision analysis bring us a framework or language that allow us to make better decision,
when we have a complex system, but is also important to remember that the responsibilities is for the
decision-maker and not for the tool. Exists a dierent tools with which we might come to think that we
don't need to make a decision, but is important to remember the system always is complex and the tools
mostly the time couldn't take to account all the complex relationships that exist.


Decisiones bajo incertidumbre en las organizaciones (2014)

A brief review

Referencias
Morton, a. & Phillips, L. D. (2009), `Fifty years of probabilistic decision analysis: a view from the UK',

Journal of the Operational Research Society 60,

S33S40.

*http://www.palgrave-journals.com/doinder/10.1057/jors.2008.175

Decisiones bajo incertidumbre en las organizaciones (2014)

You might also like