Vladimir Savchuk

Making
Making decision
decision on
on applicant's
applicant's
suitability
suitability
Bayesian
Bayesian approach
approach

Kyiv, 2015
 The Essence of Bayesian approach
1. Every manager has his own subjective judgment about a particular event
in the future (the prior judgment).
– Whether an applicant possesses all necessary skills to suit definite position
– Whether a new project will be successful
2. Is it possible to trust this judgment unequivocally?
– One hardly can, because the manager's intuition may fail
– The world is changing - past experience may not be quite acceptable and
adequate
3. What one need to do in order to improve the reliability of the output? One
should conduct an experiment (in the broad sense) and get actual data.
– One should test the applicant
4. There is little time to conduct numerous experiments - it complicates the
decision-making
5. Conclusion - one should make a decision by combining subjective the
prior information and a few results of the experiment!

V. Savchuk. Bayesian Making Decision

 The Essence of the Bayesian approach
Subjective
Data Posterior judgment
prior judgment

+
Usually Usually The amount of
not reliable little data information increased
V. Savchuk. Bayesian Making Decision
Foundations of Bayesian approach: mathematics
Let events H1, H2, … , Hn form a full
  group of events (hypothesis).
P(Hk) – is the prior probability of hypothesis H k.
Event А can occur together with one of the hypotheses.
P(A/Hk) – the conditional probability of event A, given H k
The General probability of event A is
P

Let event А occurred. Then the posterior

  probability of hypothesis
is estimated to be

V. Savchuk. Bayesian Making Decision

 The procedure of Bayesian decision-making
HR-manager considers hiring a certain candidate. Interviewing,
testing and Bayesian technology are used for making decision.
Step 1. Evaluation of the prior probabilities
After standard interview with the applicant, HR-manager evaluates
his suitability. The HR manager has the two hypotheses:
Hypothesis Н1 – applicant is suitable
Hypothesis Н2 – applicant is not suitable.
The HR-manager can assess probabilities of these hypotheses
Р(Н1) and Р(Н2).
This assessment is subjective and based on the experience of the
HR-manager.

Let the probabilities be Р(Н1)=0,80 and Р(Н2)=0,20.

V. Savchuk. Bayesian Making Decision

 The procedure of Bayesian decision-making
Step 2. Evaluation of the conditional probability
HR-manager instructs a field expert to examine the applicant by
asking him to complete a certain task. According to his preliminary
assessment:
• If the applicant corresponds to the position requirements, he will
perform the task successfully (Event A) with probability P(A/H2).
• If the applicant does not possess the necessary skills, he is able
to perform the work by P(A/H2).

Let for example P(A/H1)=0,6 and P(A/H2)=0,1.

Than the general probability of completing the task will be

P(A)=P(H1)* P(A/H1)+ P(H2)* P(A/H2)=
0,8*0,6+0,2*0,1=0,50.
V. Savchuk. Bayesian Making Decision
 The procedure of Bayesian decision-making
Step 3. Evaluation of the posterior probability
There are two options of further scenario:
Option 1. Event А occurred – the applicant completed the task
By means of Bayes’ theorem the posterior probability of
applicant’s suitability (hypothesis Н1) is evaluated to be:
= =  0,96

Option 2. Event А didn’t occurred – the applicant failed the task

According to Bayes’ theorem the posterior probability of the
applicant’s suitability (hypothesis H1) is
= =  0,64

V. Savchuk. Bayesian Making Decision

 The Bayesian MS Excel algorithm

Applicant succeeded the assessment Applicant FAILED the assessment

Step 1. Assessment А Step 1. Assessment А

Hypotheses Н1 и Н2 Hypotheses Н1 и Н2
A priori probability of P(H1) 0,8 A priori probability of P(H1) 0.8
hypotheses P(H2) 0,2 hypotheses P(H2) 0.2

Assessment А Assessment А
P(A/H1) 0,6 P(A/H1) 0.6
Conditional probability P(A/H2) 0,1 Conditional probability P(A/H2) 0.1
Probability of event P(A) 0,50 Probability of event P(A) 0.50

Was assessment succesful? Yes Was assessment succesful? No

Hypotheses Н1 и Н2 Hypotheses Н1 и Н2
A posteriori probability H1 0,64 A posteriori probability H1 0.64
of hypotheses H2 0,36 of hypotheses H2 0.36
V. Savchuk. Bayesian Making Decision
 Bayesian technology of decision-making
3-step procedure: probabilities of applicant’s suitability
Step А Step В Step C
P(A/H1)= 0.6 P(B/H1)= 0.9 P(C/H1)= 0.6
P(A/H2)= 0.1 P(B/H2)= 0.2 P(C/H2)= 0.3

Yes 1.00
Yes 0.99
No 0.98
0.96
Yes Yes 0.86
No 0.75
P(H1)= 0.8 No 0.63
P(H2)= 0.2
Yes 0.94
No Yes 0.89
No 0.82
0.64
Yes 0.31
No 0.18
No 0.11

V. Savchuk. Bayesian Making Decision

 Bayesian approach: summary
1. The Bayesian approach to decision-making makes
it possible to combine HR-manager's prior judgment
with the results of applicants' testing.
2. HR-manager obtains rationale for more grounded
decision on the applicants’ suitability.
3. HR-manager is required to be able to operate with
probabilistic categories and evaluate his opinions in
the form of subjective probabilities.

V. Savchuk. Bayesian Making Decision