PROBLEM SOLVING FRAMEWORK WITHIN DECISION

SUPPORT SYSTEMS

Yan H. Ng

Singapore Polytechnic, 500 Dover Road, Singapore 139651

Keywords: Problem Solving Cycle, Co-operative Planning, Decision Support Systems.

Abstract: In the study of problems solving, cognitive psychologists had generalized it as a process of applying skills

to overcome obstacles and constraints, so as to move from a given state to the desired goal state. In here, we

look at where a line could be drawn to divide the process of problem solving into two sub-processes; a one-

off, but cognitive demanding task for human to define the problem space, in terms of representation, and a

routine search process, that could be undertaken by a machine, to search and select a solution within a

defined solution space, so that to oversee its execution from start to finish to realize the goal.

1 INTRODUCTION

However smart a person might think of himself,

cognitive psychologists, Bransford and alike,

(Bransford & Stein, 1993), had shown that in the

event of problem solving, we still walk through the

process as follows:

1. Problem identification

2. Definition of problem

3. Constructing a strategic for problem solving

4. Organizing information about a problem

5. Allocation of resources

6. Monitoring problem solving

7. Evaluating problem solving

Although many people may not be aware of the

process, and might even miss out on some of these

stages unknowingly, especially the first two stages,

the reality would eventually bring them to

understand that the efforts and resources directed at

a misguided target is unlikely to bring the results as

expected.

Reflecting on the two decision support system

projects currently underway; namely “Hospital Bed

Management System”, and “C

I System for Bus

Lane Management”, we have observed that if the

system developer could identify and define the

problem in terms of some forms of representation, a

decision support system could be put in place to

search for a solution within the defined solution

space, then the human and machine could have

completed the problem solving cycle together to

realize the goal, collectively.

2 WHAT IS A PROBLEM?

Medin and Ross (1997) considered a problem to

have four aspects; goal, givens, means of

transforming conditions, and obstacles.

The goal is some state of knowledge toward

which the problem solving is directed and for which

at least some criterion can be applied to assess

whether the problem has been solved. The givens

include the objects, conditions, and constraints that

are provided with the problem, either explicitly or

implicitly. Problems need to have some means of

transforming conditions, of changing the initial

state to the goal state. Finally, problems usually

associated with some obstacles.

In here, Medin and Ross described problem

solving as a process

a) Starting from an initial state

b) Trying to arrive at the goal state

c) Solution is a path connecting the initial state

to the goal state; along the way, using

whatever mans and resources provided to

transform conditions, and overcoming

obstacles.

Medin and Ross also further divided the types of

problems into Well-defined problems, and Ill-

defined problems; the former as having completely

140

H. Ng Y..

PROBLEM SOLVING FRAMEWORK WITHIN DECISION SUPPORT SYSTEMS.

DOI: 10.5220/0003434301400146

In Proceedings of the 13th International Conference on Enterprise Information Systems (ICEIS-2011), pages 140-146

ISBN: 978-989-8425-54-6

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

specified initial conditions, goals and means of

transforming conditions, while the latter have some

aspects of the problem not completely specified.

Nevertheless, H. A. Simon (1973) has argued that a

crucial part of problem solving is changing an ill-

defined problem into a well-defined problem (or

often, a number of well defined problems).

3 PROBLEM-SOLVING CYCLE

Bransford, and Sternberg (1996) consolidated the

problem solving sequence into a seven step cycle as

stated in the Introduction Section; in which problem

was identified and defined in Step 1 and 2, to

provide the initial state and goal state for the

generation of a path in Step 3, 4 and 5. However, in

a dynamic environment, when all objects are

constantly evolving, the chosen solution is unlike to

run smoothly from start to finish without a hitch,

some form of modifications would likely to take

place so as to steer the course. Therefore, a problem

solver should monitor and guide the solution

towards the goal state, modify and repair it along the

way. Finally, once arrived at the goal state, the

problem solver should evaluate the results, and

hopefully learn from the experience.

4 FRAMEWORK FOR PROBLEM

SOLVING

By studying the process of problem solving by

human; we examine the limitation of problem

solving by machine, with the purpose of introducing

machine intelligence into the process of problem

solving.

Firstly, problem identification and definition are

the two very demanding cognitive tasks that should

be left with human being in the problem solving

cycle, so that the problem solver could study the

problem, focus on the problem, and define it in

terms of goal, and objectives that it could be

decomposed into.

Secondly, although monitoring problem solving

in a dynamic environment is another demanding

task, it is possible to develop a system framework

for a machine to undertake this task to select a

suitable strategic formations and tactical plans from

a set of plans well defined plans kept inside the

solution space, so as to steer the operational plan

from start to finish.

Thirdly, if the construction of strategic

formation could be combined with its tactical

realization, and package together as one co-operative

plan, then selection of the best possible plan could

be based on not only the quality of strategic

formation, but also the cost of realization as the

criteria for plan selection.

Based on Sternberg (1996) 7 steps sequence, we

put forward a problem solving framework that

consolidated the problem solving cycle into a 4

stages operational procedure:

1. Deliberation;

2. Planning;

3. Plan Selection; and

4. Plan Realization.

Deliberation:

At the Deliberation stage,

• The system architect studies the current

situation and identifies the desired situation

by analyzing the situation map.

• The system architect performs mapping on

the desired situation (qualitative goal) into

a list of achievable (quantitative)

objectives, which must be accomplished in

order to realize the goal.

• The system architect evaluates the current

situation and its feasibility in reaching all

the intermediate states required in the

desired formation either by case-based

reasoning or planning.

Figure 1: The process of translating goal into a strategic

formation.

Planning:

During the Planning stage, the task is to develop

strategic formations that connect all objectives

identified together, as well as looking into the details

of each objective to study the feasibility of their

realization.

The main task for planning is to produce at least one

fully connected solution paths in the form of a two

stage co-operative plan

• Strategic Planning establishes the

operational flows of the entire business

process.

• Tactical planning to work out the physical

PROBLEM SOLVING FRAMEWORK WITHIN DECISION SUPPORT SYSTEMS

141

paths between states

• The distinction between strategic and

tactical planning is essential here, as the

former provides the overview of the plan

from start to finish, while the latter would

only generate paths between two states.

• Strategic Planning provides the formation

based on the list of intermediate states, or

achievable objectives (partial plan),

sandwich between the initial and goal

states.

• The eventual outcome of a strategic plan

should be at least one operational plan that

flows from the initial state to the desired

goal state, stating the constraints and

preconditions in each state, as well as the

interlocking relationship between states

(objectives), but without physical paths

joining states together. As shown in Figure

4 below.

Figure 2: Strategic Planning is a process of decomposition.

• Tactical Planning is responsible for

generating detailed paths to achieve the

objectives in the formations.

• its approaches the problem from the lowest

level of abstraction, formulating solution in

the form of course of actions to move from

one state to another, often without an

overall picture (strategy) that includes the

initial and desired goal states.

• It decision is based on short-term

consideration to connect the two

neighboring states under the guidance of

some heuristics.

The framework for Co-operative Planning

brings together the overview of the strategic

formation (strategic planning), as well as the detail

course-of-action for the realization of each objective.

As shown in Figure 3.

Figure 3: Framework for Co-operative Planning.

This two level Co-operative Planning process

enhances the chances of the solution (work

procedure) survive in a dynamic environment during

plan realization.

• By focusing initially on strategic issue as a list

of objectives, which have only pre- and post-

condition, but without rigid physical paths

between states. This enable changes in the

strategic formation by simply adding and

deleting objectives, without constraint of

physical connection.

• Nevertheless, the plan will become rigid and

inflexible once the paths between intermediate

states have been at the tactically planning stage.

Plan Selection

The third stage is a decision-making procedure to

select an operational plan, where all proposed

solutions are compared based on some forms of

selection mechanisms to resolve conflict among

different plans.

• Each strategic plan once being merged and

consolidated into a workable solution

(workflow), its quality can be measured in terms

– cost, and

– uncertainty,

• The most preferential solution path would be

given priority as the operational plan, with other

workable solutions being labeled as stand-by

plans.

Plan Realization

At the final stage of the operational procedure, an

operational manager will be put in place within the

decision support system (DSS) to check and assess

the progress of each process to ensure that the pre-

programmed work procedure (workflow) remains

relevant and cost-effective in the light of changes in

the environment. With the framework of Co-

operative planning, the Operational Manager could

modify and repair at either the strategic or tactical

level.

The roles of the DSS Operational Manager is

• to coordinate all required resources to achieve

the ultimate goal.

• to repair the paths between intermediate states,

or perform local re-planning, when the survival

of operational plan is threaten

• to trigger stand-by plans, when operational plan

becomes impossible

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

142

5 HOSPITAL BEDS

MANAGEMENT SYSTEM

In a modern hospital, the management of hospital

beds is an essential part of the in-patient workflow

system, as it indicates not only the flows of in-

coming patients into the wards, but also the

resources and manpower required to support these

patients using the beds. Therefore, the pattern of

hospital bed usage is often used as a proxy to

measure the number of patients treated in the

hospital.

Deliberation:

Problem identification:

There is no central control to overseeing the release

of beds for the admission of new patients.

Definition of problem:

Although a bed management system is installed at

Khoo Teck Puat Hospital (KTPH), it is only a

situation map to display the status of hospital beds

usage in terms of R (Ready to use), or P (Prepare for

discharge), but does not monitor the transition from

“P” to “R”. Hence, we define the goal as follows:

Goal: To monitor the realization and execution of

the operational plan to transform from “P” to

“R”.

Objectives: Discharge (A, F, T, L) Æ C.

The goal could be decomposed into a list of

objectives shown above. Patient discharges are often

decided upon during the physician visit to the ward

every morning. A list of activities followed

thereafter must be completed before patients leave

hospital; scheduling appointment (A), collecting

prescriptions (F), arranging family member to take

them home (T), collecting letter from the doctor who

decide to discharge them (L). These activities are not

sequential, and therefore could be executed in

parallel. Once the bed is freed, the bed should be

cleaned (C) before the status changes from “P” to

“R”. In theory, cleaning process could take place in

parallel with Discharge (A, F, T, L) provided the

discharged patient could be moved off the bed into a

discharge lounge, where patients could rest and wait

in the lounge for the completion of “A”, “F”, “T”

and “L”.

Planning:

Show in the situation map, the existing bed

management system indicates the number of “P” and

the number of “R” on the screen, to indicate the

status of patient admission into the wards. The

execution of the course of actions for Discharge (A,

F, T, L) are undertaken outside the system, only

when “P” is transformed into “R” status would the

status be displayed onto the system.

To speed up the transformation from “P” to “R”,

tactical planning for “A”, “F”, “T” and “L” will be

formulated as part of the co-cooperative plan.

Co-operative plan 1

Strategic Plan 1: Discharge (D Lounge) Æ C Æ R

Tactical Pan 1: If patient is well enough to be send

home, discharge to the Discharge

Lounge, would enable the bed to be

cleaned, and proceed to turning the

status of bed from “P” to “R”.

Co-operative plan 2

Strategic Plan 2: Discharge (A, F, T, L) Æ C Æ R

Tactical Plan 2: However, if the patient is not well

enough to be moved off the bed, the

completion of “A”, “F”, “T”, and

“L” will be the precondition to

proceed to cleaning the bed “C”.

This will become the task for DSS

operational manage to monitor and

execution during the plan

realization stage.

Co-operative plan 1 and Co-operation plan 2 are the

alternative realization for the same goal to transform

from “P” to “R”.

Plan Selection:

Different sets of Co-operative plan will be studied

and compared for the selection of the best plan as

the operational plans, leaving the rest of the plans as

stand-by plans.

Plan Realization:

One of the strength of Co-operative planning is in its

ability to repair or change plan during execution in

respond to changes in environment; either

strategically, or tactically. For instance, if status “F”

have not completed within a predetermined time,

alert will be sent to pharmacy to look into the matter.

Same is applicable to “L” status, if letter from doctor

is unable to be completed in time, then alternatively

solution such as sending it by email, or post, could

be activated. The role of DSS operational manager is

to study and monitor to ensure the operational plan

is successfully executed.

In summary, the operational procedure of the KTPH

bed management system consists of 4 stages:

1. Deliberation:

A Dashboard or Situation Map will

consolidate information about bed status to

be displayed onto the nurse station in the

wards, and the A&E department.

PROBLEM SOLVING FRAMEWORK WITHIN DECISION SUPPORT SYSTEMS

143

2. Planning:

A number of Co-operative plans would be

studied and put forward for consideration

3. Plan Selection:

Assume that Co-operative Plan 1 had been

selected as operational plan. Co-operative

Plan 2 will become the stand-by plan.

4. Plan Realization:

In the event of operational plan become

impossible to move patient into the

Discharge Lounge, the DSS operational

manager could switch over to one of the

stand-by plan, such as Co-operative Plan 2

to complete the Discharge (A, F, T, L)

before cleaning the bed.

In reflection, we consider the hospital beds

management system as a well-defined problem with

very little uncertainty, therefore, we could walked

through the process of problem solving cycle, by

dedicating the first two stages to the system

architect, and leaving the plan selection and

realization phases to the decision support system to

take care of.

1. Deliberation; untaken by system architect

2. Planning: untaken by system architect

3. Plan Selection: untaken by the system

4. Plan Realization: untaken by the system

In here, only case-based reasoning approach is

required to search through the solution space stored

in the case-based library for managing the bed usage

in the hospital.

There is no need to deploy planning approach to

construct solutions backward from the goal state, or

forward from the current state. Nevertheless, in

order to demonstrate the potential of our problem

solving framework, we have included a different set

of Co-operative plan in the event of national

emergency when the hospital was overwhelmed by

hundred of patients coming through the A&E

department.

Deliberation:

Problem identification:

There is no central control to overseeing the release

of beds for the admission of new patients.

Definition of problem:

In order to increase the number of patients admitted,

we now defined the problem by means of beds

available within the whole KTPH, and not by

number of beds inside the wards. In other words,

beds could be placed anywhere along the hospital

corridors, so that to increase the intake of patients.

The Goal: to search for spaces to accommodate the

extra beds outside the wards.

Objectives: to pack maximum number of beds in the

designated areas (DA).

Planning:

Assuming the hospital has sufficient space to

accommodate all the extra beds needed to cope with

the situation. One of the strategic plans could be

Co-operative plan N

Strategic Plan n:

Extra capacity = DA

+ DA

+ …… DA

Tactical Pan m:

Searching extra beds for DA

; +

::::::::::::::::::::::::::::::::::::::::::::::

Searching extra beds for DA

Logistically, in order to realize the tracking of

moving objects (beds) moving around the hospital,

each bed would be fitted with a wireless tag, so that

its movement could be tracked and status displayed

on the dashboard or situation map.

6 BUS LANE MANAGEMENT

SYSTEMS

The bus lane scheme in Singapore was launched in

1974, to give buses priority to travel along special

destined bus lane on the roads, so that to enter and

exit bus stops without the obstruction.

Subsequently, the Land Transport Authority

(LTA) of Singapore launched a full-day bus lane

scheme in 2005 along Orchard Road to improve bus

travelling speeds, with the purpose of encouraging

shoppers to use public transport, rather than taking

onto their cars. In December 2008, it was Estimated

that there were a total of 23km of full-day bus lanes

and 155km of normal bus lanes in Singapore.

This project is funded by the National Research

Foundation of Singapore.

Deliberation:

Problem identification:

Vehicles other than buses are obstructing buses

travelling on the roads.

Definition of problem:

To detect and identify the vehicle intruding into the

designated bus lanes; and record down the evidence

of violation legally admissible in court.

The Goal: to detect and identify the vehicle

intruding into designated bus lanes.

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

144

Objectives: Obj1 is to detect the vehicle into the bus

lanes; (Using scanners and overview

camera)

Obj2 is to identify the type of vehicle,

and its number plate; (Using a front and

a rear camera)

Obj3 is to record down the sequence of

violation with time stamp, which could

be legally admissible in court. (Using an

overview camera)

Obj4 is to transmit the data to the

backend system (LTA office) by means

of wireless technology.

Planning:

One of the strategic plans could be

Strategic Plan 1: Obj1 + Obj2 + Obj3 + Obj4

Tactical Pan for Obj1 Detect vehicle into bus lane

within the 50m range before

the bus stop, trigger the

activation of the front and

rear cameras, plus nIR

lighting if needed.

Tactical Pan for Obj2: capture and read the vehicle

licence plates (including

front and rear views) and

vehicle classification.

Tactical Pan for Obj3: record sequence of vehicle(s)

violation footage inside the

bus lanes.

Tactical Pan for Obj4: processed all the information

required on-site, before

transmitting them to the back

office by wireless.

Plan Selection:

There will be different sets of strategic plan for

different scenarios; in a normal scenario, the vehicle

licence plates are clearly recognisable, but in other

situations where vehicle licence plates could not be

positively identified. In such cases, a different Co-

operative plan might have to be activated to identify

the vehicle involved, such as checking inside the

database for profile of the vehicles resemble the one

in question that might have been captured by the

system earlier. In this case, features of this vehicle in

question, other than licence plates could be used to

help to identify the vehicle. Based on the assumption

that people always making the same journey when

travelling from one place to the other.

Different tactical plans are also kept n place to

support classification of vehicle, by means of colour

detection, license plate colour detection, type of

vehicles.

Plan Realization:

In order to cope with every kind of situations on the

roads, the role of the DSS operational manager is to

steer the operation to achieve the objectives set out

in the strategic plan, and realize the goal by detect,

identify, capture, and transmit the violation of bus

land back to the back office.

In comparison with the hospital bed management

system, the bus lane management system is also a

well-defined problem with changes in environment,

and quality of pattern recognition on the licence

plates being the uncertainty, we, therefore, also

dedicated the first two stages to the system architect,

and the plan selection and realization tasks to the

DSS operational manager.

1. Deliberation; untaken by system architect

2. Planning: untaken by system architect

3. Plan Selection: untaken by the system

4. Plan Realization: untaken by the system

Once identified, and defined, all possible scenarios

are studied to develop solutions to cope with the

situations, with representation of each solution kept

inside the case-based library, so that, DSS

operational manager could select, and activate the

operational plan to detect and identify the vehicle

intruding into designated bus lanes. If successful,

LTA of Singapore will consider doing away with

civil contractors to police bus lane to enforce

violation.

7 CONCLUSIONS

The framework for problem solving put forward in

this paper has been applied to the two DSS projects

presented above with encouraging results; the two

level co-operative planning has provided the

flexibility for the decision support system to repair

and modify the operational plan during execution,

supported with stand-by plans available in the

solution space.

The difficulty with searching for problem space

is that the number of possibility is often too large to

try them all. The problem solving framework we

proposed here, provides a sound partition of work

between the system developer, and the DSS

operational manager to enable them to complete the

problem solving cycle co-operatively and

collectively; each focus on what they do best; human

developers take on the most cognitive demanding

task in deliberation and planning so as to define a

solution space, focus on only a limited sets of

solutions for the machine to select, while the DSS

PROBLEM SOLVING FRAMEWORK WITHIN DECISION SUPPORT SYSTEMS

145

operational manager would perform the routine tasks

of plan selection and execution; selecting the most

appropriate plan as operational plan, and steer it to

achieve the objective, and realize the goal. Hence,

the framework for problem solving we put forward

here reinforces Newell and Simon’s view that

problem solving as representation and search. The

steps for deliberation and planning focus on issues

of identifying problem, defining problem into a set

of representations that is sufficient to cover the

problem space, yet small enough for machine to

search for solution, and realize it. (Newell & Simon,

1972)

Representation:

1. Deliberation; untaken by system architect

2. Planning: untaken by system architect

Search:

3. Plan Selection: untaken by the system

4. Plan Realization: untaken by the system

In conclusion, the two decision support systems

undertaken are domain dependent and well defined,

with only a limited set of scenarios to cover.

Therefore, at this moment in time, we could idealize

the solution space as a closed solution space with a

set of well defined solutions of problem stored in the

knowledge base. However, as H. A. Simon (1973)

put it “a crucial part of problem solving is changing

an ill-defined problem into a well-defined problem

(or often, a number of well defined problems).”

Given time, when we are able to build or granted

access to sufficient number of well-defined

problems, we would be in the position to develop

methodology to solve ill-defined problems.

The development of Evidence-based medicine

(EBM) is one step along this direction.

“The conscious, explicit and judicious use of the

best current evidence in making decisions

about the individual patients.”

(Sackett et al., 1996)

The so called best current evidences are patient

case notes of certain disease kept in the databases of

some EBM centres. One day, in a not too distance

future, we, as doctors, could use a list of symptoms

associated with a particular disease as search keys to

locate a summary of these patient case note stored in

databases around the world, to help us in our

diagnosis for some unknown or ill-defined disease.

With the connectivity of internet today, it should not

be too far in the future to realize this dream.

ACKNOWLEDGEMENTS

We would like to thank the support of National

Research Foundation of Singapore for funding the

“C

I System for Bus Lane Management”, project,

SciMed (Asia) Pte Ltd, Khoo Teck Puat Hospital

and Singapore Polytechnic for funding the “Hospital

Bed Management System” project.

REFERENCES

Bransford, J. D., & Stein, B. S. (1993). The IDEAL

problem solver: A guide for impoving thinking,

learning and creativity (2nd ed.). New York: W. H.

Freeman.

Medin, D. L., Ross B. H., 1997. Cognitive Psychology,

Harcourt Brage College Publishers. United States of

America

Newell, A., & Simon, H. A. 1973. Human problem

solving. Englewood Cliffs, NJ: Prentice.

Sackett, D., Rosenberg, W., Muir Gray, J., Haynes, R.

Richardson, W. (1996). Evidence based medicine:

what it is and what it isn't. British Medical Journal,

312, 71-72. http://cebm.jr2.ox.ac.uk/ebmisisnt.html.

Simon, H. A. 1973. The structure if ill-structured

problems. Artificial Intelligence, 4, 181-201.

Sternberg, R., 1996. Cognitive Psychology Second

Edition, Harcourt Brage College Publishers. United

States of America.

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

146