Exercise description:

Exercise: making a knowledge model for an office allocation task (Ch. 5)

This exercise assumes that you have studied Ch. 5 of the KE&M book on Knowledge Model Components.

Carefully read the following description of the office allocation problem, and answer the following questions.

  1. Make a diagram containing the major information types, such as office, employee, project, topic.
  2. Make a list of the types of knowledge that are employed by the expert. A key role is played by various "constraints", "requirements" and "preferences". Model these knowledge types by means of the "rule type" construct, or by suitable alternative constructs, if you wish.
  3. Give an argument whether or not the task template for assignment (KE&M book Ch. 6) would be suitable for this office allocation task. To show this, construct an "annotated inference structure".
  4. Indicate the necessary steps needed to complete the knowledge model.

Hints for the exercise:

Ad 1: For these information types the standard constructs are usually adequate: concept, relation, property and subtypes.

Ad 2: Consider the order of allocation, the requirements imposed on certain offices, and the knowledge needed for selecting two persons for a single office.

Ad 3: Often, the knowledge types already suggest a certain inference step. However, not every row in the table below corresponds to a single inference step. For example, determining the order of the allocation and the allocation decision itself may be done together.

Ad 4: Make the simplifying assumption that no "backtracking" is necessary.

Tools

To make the diagrams you may use the ModelDraw program (runs under Win95/98/NT) that can be freely downloaded from the CommonKADS website http://www.commonkads.uva.nl.

The office allocation problem

All departments of a large research institution move from an old location in the center of the city to a new building in a business parc outside the city. Every department gets a number of nearby rooms in the new building. Your task is to develop a system for the allocation of these offices within each department. Many constraints and preferences have to be taken into account in order to come to an acceptable solution. The data for one department are given by way of example. This example department has one department head, one deputy head, three project leaders, eight researchers and two secretaries. The floorplan of this department with all its offices is shown below. The offices 117 and 119 upto and including 123 are offices large enough for two persons. The offices 113 upto and including 116 are single-person offices. The offices 118 and 124 are in use as general-purpose offices.
Information on the allocation process is found in the Table below. It contains data from an interview session with a human expert who solved the allocation problem for this example department. The given solution is not the only possible one. The left column of the Table shows the allocation decisions made by the expert, including their time ordering. The right column shows the motivations given by the expert in the knowledge-elicitation session.


 

Action by the expert Motivation by the expert
1.  Put the department head into office 117  1.a.  The department head needs a large and central office. 

1.b.  This allocation is done first, because it strongly constrains the alternative options.

1.c. Office 119 would have been a good choice, too. The choice between 117 and 119 is arbitrary.

2.  Put the secretaries together into office 119 2.a.  The secretaries should be located as closely as possible to the department head.

2.b.  They must have a centrally located office.

3.  The deputy head gets office 116  3.a.  The deputy head must be located close to both the department head and the secretaries.

3.b.  A small office suffices.

4.  The three project leaders are put in offices 113, 114, and 115 4.a.  Project leaders are entitled to have their own office. 

4.b.  Who actually gets what single-person office, is an arbitrary choice.
5. Researchers A and B are put together in office 123 5.a.  Preferably, researchers share an office.  

5.b.  A and B are allocated an office first because they both smoke. Putting smokers and non-smokers together in an office is not acceptable. This is a "hard" constraint.

5.c. The choice of the office is arbitrary as long as it is a two-person office.

6.  Researchers A and B are put together in office 122  6.a.  These researchers are working on the same topic, but in different projects.

6.b.  The institute's approach is to put researchers working on the same project in different offices, in order to stimulate discussion and cross-fertilization between projects. However, this is not a strict rule but a preference, a "soft" constraint.

6.c. The institute further prefers to put together researchers working on the same topic, in order to create synergies. This is also a "soft" constraint, that easily conflicts with the "same-project" rule.

7.   (et cetera)