1 / 34

AN INTERACTIVE DYNAMIC SIMULATION MODEL FOR STRATEGIC UNIVERSITY MANAGEMENT1

Yaman Barlas and Vedat Guclu Diker

Bogazici University, Industrial Engineering Department, Bebek, Istanbul, Turkey 80815 ybarlas@boun.edu.tr diker@boun.edu.tr

ABSTRACT We present an interactive simulation model, on which various problems involved in managing the academic aspects of a university can be analyzed and certain policies for overcoming these problems can be tested. More specifically, the model focuses on long-term, strategic management problems that are dynamic and persistent in nature, such as growing student-faculty ratios, poor teaching quality and low research productivity. The model yields numerous performance measures about the fundamental activities of a university, namely, teaching, research and professional project activities. The modeling process consists of two major phases. The first one is the construction of a system dynamics simulation model (the �engine� behind the interactive game). The second phase consists of converting the simulation model into an interactive game. The simulation model is built first, and validated/verified by standard system dynamics tests, using data from Bogazici University, Istanbul, Turkey. Next, the necessary changes are made on the equations to turn the model into an interactive one and the gaming interface is programmed. The game has been played and tested by faculty members, managers, teaching assistants and students. The game results generated by these players are compared. Differences in the results reveal that players with different orientations focus on different aspects and performance measures of the university in making decisions. Research results obtained suggest that the interactive simulation model promises to be not only a useful technology to support strategic decision making, but also a laboratory for theoretical research on how to best deal with strategic university management problems. Further research is being carried out to add enhancements both to the simulation model and the gaming interface. I. INTRODUCTION Contemporary universities worldwide face the challenge of maintaining the quality of the fundamental functions of a university, namely teaching, research, academic and professional service, under the pressure of limited resources in terms of faculty, facilities and income. [6], [7], [12], [19], [22]. Some typical problems that the university management must deal with are: unbalanced growth in student body in state (public) universities; infrastructures that can not keep up with the enrollment growth; increased student/faculty ratios; concerns about quality of instruction; heavy competition for limited funding for research and heavy competition among private universities for limited student demand. Such problems have been studied both on macro and micro levels by many researchers. While some of these studies have made use of certain quantitative methods (for instance, [13], [17], [18], [20]), a great portion of the existing research on university problems do not have quantitative foundations, primarily because such problems involve qualitative (human) elements that are difficult to quantify and model. Thus, there seems to be a need for research that can handle simultaneously both the quantitative and the qualitative aspects of the university management problems The purpose of this research is to construct a dynamic simulation model - a computerized platform - on which a range of problems (of both qualitative and quantitative nature) concerning the contemporary university management can be analyzed and certain policies for overcoming these problems can be tested. The research utilizes System Dynamics modeling/simulation methodology which is especially suitable for quantitative and realistic representation of variables that are typically perceived to be qualitative - yet important in socio-economic management problems. In particular, the model will focus on those university problems that are dynamic and persistent in nature and as such must be addressed by high level, strategic policy-making mechanisms within the university. These strategic policy making mechanisms are typically the president, the deans, and the major policy-making councils at the university and divisional levels. The ultimate product of the research is the interactive gaming version of the simulation model that can be played by such policy makers. The interactive management simulators (often called �microworlds� or �management flight simulators� [1, 11, 14, 15] constitute a very promising new technology, especially for strategic management. 1 Partial funding for this research was provided by Bogazici University Research Fund.

2 / 34

II. RESEARCH METHODOLOGY The methodology used in this research is System Dynamics methodology. System Dynamics approach is appropriate for analyzing and modeling complex, large scale, non-linear, partly qualitative socio-economic systems. The main assumption of System Dynamics approach is that a problem, with all its physical, ecological, sociological, economic aspects, is a whole. Other important assumptions of System Dynamics are: i) variables that are related have a direct cause and effect relationship, ii) succession of cause and effect relations forms feedback loops, which means that the distinction between cause and effect vanishes through time. Almost every dynamic system includes a number of both negative (goal seeking) and positive (reinforcing) feedback loops, which interact and operate simultaneously through time. Large scale models include very large number of feedback loops. It is not unlikely to encounter variables which are related to several thousand feedback loops within a model with several hundred variables. Thus, the dynamic behavior of a system is a result of complex interactions of numerous loops over time. The strategic university management summarized in the INTRODUCTION section is a typical system dynamics problem. It is a long-term, complex, strategic management problem. The dynamic performance of the system is created as a result of the interaction of numerous feedback loops, as will be discussed below, in the description of the model. System Dynamics methodology typically consists of the following steps: (1) problem description, (2) model conceptualization, (3) model construction (simulation model), (4) verification and validation of the model, (5) simulation experiments [10], [16]. Once the System Dynamics model has been thoroughly verified and validated, it can be converted into an interactive simulation game (’microworld’). In developing an interactive simulator, the proper selection of the interactive decision variables (the player decisions) is crucial. Once these variables are selected the model must be modified accordingly. The next step is the design of the interactive interface (�screen design�). There are many conceptual and technical issues that must be tackled in constructing interactive dynamic games. These are discussed in the relevant literature (see, for instance, [1], [11], [14], [15]). Finally, once the conceptual design is complete, the user interface is programmed by using an appropriate interface development tool.

III. MODEL DESCRIPTION III.1. Model Overview The focus is long-term, academic management problems that are dynamic and persistent in nature, such as growing student-faculty ratios, poor teaching quality and low research productivity As such, the model represents the faculty members� time allocation among the main activity groups, the factors that determine this allocation and various performance indicators obtained as a result of these -often conflicting- allocations. The major activities of a faculty members are: (a) graduate instruction, (b) under-graduate instruction, (c) graduate instruction overhead, (d) under-graduate instruction overhead, (e) unsponsored research, (f) sponsored research, (g) income generating projects, (h) unofficial projects. With respect to these activities, the faculty members are grouped in two: Graduate faculty members, that are primarily involved in graduate instruction and research (but also involved in some under-graduate instruction.) and Undergraduate faculty members, that are involved only in under-graduate instruction and have limited interest in (and background for) research. Thus, graduate instruction and graduate instruction overhead loads apply only to the graduate faculty members. The instruction activities are divided into two groups: (a) in-class instruction and (b) instruction overhead, which includes all out-of-class activities related to instruction. The second main activity group is research activities. Research activities are represented in two categories as (a) unsponsored research activities, which are not sponsored financially except for the university�s own resources and (b) sponsored research activities, which are supported by governmental or private organizations. The last activity group is project (consulting) activities, which are also divided into two: (a) income generating projects, which are activities like seminars, courses or consulting realized through university channels and generate income to the university and (b) unofficial projects, which are activities like seminars or consulting realized through non-university channels and do not generate any income to the university. The model is constructed on sector basis. (Figure 1). The sectors of the model are determined so as to represent the dynamics of major activity groups defined above. The diagrams shown in Figure 2- Figure 5 are called

3 / 34

�stock-flow� diagrams in System Dynamics methodology. They consist of four symbols: Stock variables (denoted by rectangular box) represent the fundamental accumulations over time. (Such as �Number of Graduate Faculty� or �Number of Graduate Students� in Figure 2.) Flow variables (denoted by a double arrow through a valve symbol) represent the rate of change of stock variables. (Such as �New Graduate Faculty� or �Graduate Student Admission�.) Single arrows mean �the variable at the root of the arrow has an influence on the variable at the tip of the arrow�. (For instance, �Number of Graduate Programs� has an effect on �Total Graduate Instruction Hours Needed�.) Variables whose names do not appear in any symbolic shape, are intermediate variables (called Auxiliary or Converter) that represent how variables influence each other, beyond the fundamental stock-flow relationship. (For example, in Figure 2, �Graduate Faculty Hiring Decision� is influenced by canceled lecture hours, number of existing graduate faculty, operating maximum instruction hours per graduate faculty and actual instruction load per graduate faculty. Thus, the mathematical formulation of �Graduate Faculty Hiring Decision� would be a function of these four variables.) The model is constructed using Vensim software [9]. The basic time step used in the simulation is a semester. III. 2. Graduate Instruction Sector In this sector, the graduate faculty FTE that is available for instruction and the need for graduate instruction are calculated and the faculty FTE is assigned to graduate instruction (Figure 2). The main stock variables in this sector are �Number of Graduate Faculty� and �Number of Graduate Students�. �Number of Graduate Faculty� decreases through �GF that Leave� and increases through �New Grad Faculty�. �New Grad Faculty� is determined by �GF Hiring Decision�, under the constraints of �Vacant Faculty Positions and �Indicated GF Supply�. �GF Hirin g Decision � is a user decision variable in the game version, but in the original model, it is computed so as to eliminate the discrepancy between the need for graduate faculty and the existing graduate faculty FTE. �GF Supply� depends on �Instruction Load per GF� and �Historical Average GF Salary�. These variables affect �GF that Leaves�, as well. (Figure 2). The dependence of �GF Supply� on �Instruction Load per GF� and �Historical Average GF Salary� is a typical ’multiplicative effect formulation’ and is given by:

GF Supply = (Max Supply) * Effect of Instruction Load per GF on GF Supply * Effect of Historical Average GF Salary

Each of these ’effects’ is a graphical function of the related variable. For instance, Effect of Instruction Load per GF on GF Supply is given by: Similarly, Effect of Historical Average GF Salary on GF Supply is given by: (This type of typical effect formulation is used in many other parts of the model.)

0 10 20

1 .5 0

0 1000 2000

1 .5 0

4 / 34

The variable that represents the graduate faculty FTE that can be assigned to instruction is �Operating Total FTE (Full Time Equivalent) Grad Faculty for Instruction�. �Full time equivalent (FTE)� is a keyword that is used throughout this study and means the equivalent number of faculty members involved full-time in a certain activity. Suppose there are 20 faculty members who spend 1/4 of their time on research activities on the average; so we have 20 * (1/4) = 5 FTE faculty members involved in research. �Operating Total FTE Grad Faculty for Instruction� is the product of �Number of Grad Faculty� and �Operating Maximum Instruction Hours per Grad Faculty�. This latter indicates the maximum weekly instruction hours that can be assigned to a graduate faculty member under normal conditions and is given by:

�Operating Maximum Instruction Hours per Undergraduate Faculty� - �Release Time for Grad Faculty� - 3 which is the normal operating maximum instruction hours per graduate faculty (Thus, graduate faculty�s maximum teaching load per week is at least 3 hours less than that of undergraduate faculty, but the model equates the graduate and undergraduate faculty teaching loads as long as the graduate faculty load is less than operating maximum load.). �Release Time for Grad Faculty� is the amount of decrease made in the weekly instruction load per graduate faculty member on the average, in order to leave them more time for research. It is a user decision variable in the game version. �Operating Maximum Instruction Hours per Undergraduate Faculty� is assumed to be 9 hours per week.. Graduate students are divided into two as �Graduate Students 1�, who take courses and �Grad Students 2�, who have completed their course work and prepare their theses. The need for graduate instruction is represented by �Total Grad Student Hours�, which is given by:

�Number of Graduate Students 1� * �Average Hours per Grad Student� �Total Graduate Instruction Hours Needed� is equal to:

Max {[�Total Grad Student Hours� / �Desired Average Grad Class Size�] , [�Number of Grad Programs� * �Minimum Instruction Hours per Grad Program�]

�Total FTE Needed for Grad Instruction� is given by:

�Total Graduate Instruction Hours Needed� / �Weekly Hours per Faculty�. After �Total FTE Needed for Grad Instruction� and �Operating Total FTE Grad Faculty for Instruction� are calculated, these variables are compared to see whether a discrepancy exists between them. This discrepancy is represented by �Extra FTE Need for Grad Instruction�. This variable takes a positive value if the need is more than available FTE, and it takes the value zero otherwise. (Figure 2). �Extra FTE Need for Grad Instruction� is compensated by �Part Time FTE for Grad Instruction�. �Part Time FTE for Grad Instruction� can not be more than a certain percentage of the FTE graduate faculty assigned to graduate instruction (As a minimum requirement for education quality). So, it is equal to:

Min {�Extra FTE Need for Grad Instruction�, �Operating Total FTE Grad Faculty for Instruction� * �Max Part Time Percentage for Grad Instruction�}.

If �Extra FTE Need for Grad Instruction� is greater than �Part Time FTE for Grad Instruction�, it can not be entirely compensated and a discrepancy, �Grad Instruction Discrepancy�, occurs. The model tries to eliminate this discrepancy by increasing the average class size. �Maximum Grad Class Size Allowed� sets an upper limit for the �Actual Average Grad Class Size� and if increasing the class size up to that limit can not eliminate the discrepancy, the variable �Unsatisfied Need for Grad Instruction� takes a positive value. As a result of this unsatisfied need, the model assigns extra instruction hours to graduate faculty. Weekly instruction hours that can be assigned to a graduate faculty member, in such cases of abnormally high instruction need, is limited by �Absolute Maximum Instruction Hours per GF�. (Assigning extra class hours to full -time faculty members increases the effective full-time FTE for instruction, thus allowing some additional increase in the part-time FTE, since part-time FTE usage is limited by a certain percentage of full-time FTE for instruction). If this extra FTE assignment is not enough to eliminate the unsatisfied need for graduate instruction, the courses that can not be assigned are canceled, so the variable � Canceled Grad Lecture Hours� takes a positive value.

5 / 34

An output of this sector and an input to Undergraduate Instruction sector, �Implied FTE Grad Faculty for UG Instruction� is calculated as the positive difference between �Operating Total FTE Grad Faculty for Instruction� and �Total FTE Needed for Grad Instruction�. Note that if this difference is positive, all graduate instruction need is met by graduate faculty within the limit of operating maximum instruction hours; and the surplus FTE graduate faculty for instruction is transferred to �Undergraduate Instruction Sector� to be used for undergraduate instruction. III. 3. Undergraduate Instruction Sector In Undergraduate Instruction sector, the need for undergraduate instruction is determined and met by the undergraduate faculty time that can be assigned to instruction and the surplus graduate FTE for instruction, if any, from Graduate Instruction sector. If the need is more than the available FTE, some strategies like hiring part-time faculty and increasing class sizes, are used to eliminate the discrepancy. This sector resembles the �Graduate Instruction Sector� in many aspects. The dynamics governing �Number of Undergraduate Faculty� are similar to the dynamics governing �Number of Graduate Faculty�. The tot al available faculty FTE for undergraduate instruction is represented by �Implied Total FTE for UG Instruction� and is equal to the total of �Implied FTE Grad Faculty for UG Instruction� and �Operating Total FTE UG Faculty for Instruction�. (The sector diagram is not shown due to lack of space. See [8]). If the available faculty FTE for undergraduate instruction is more than the need, which is represented by �Total FTE Needed for UG Instruction�, then �Extra FTE Needed for UG Instruction� takes a negative value. This FTE surplus is divided among �Surplus FTE Grad Faculty� and �Surplus FTE UG Faculty�. On the other hand, if the need is more than the available faculty FTE for undergraduate instruction, then the discrepancy is first handled by part-time faculty. Here, again, the amount of part-time FTE is limited by a certain percentage of �Implied Total FTE for UG Instruction�. If part-time FTE fails to compensate the extra need, average class size is increased. The class size is limited by �Maximum UG Class Size Allowed�. If increasing the class size even up to this limit fails to eliminate the discrepancy, the instruction loads on graduate and undergraduate faculty are increased. If increasing these loads up to absolute maximum instruction hour limits and using additional part-time faculty FTE can not compensate the need, then some undergraduate courses are canceled. As all the instruction assignments are completed, average instruction loads per graduate and undergraduate faculty member are calculated in this sector. The dynamics and the role of �Number of Programs� are modeled by similar structures in �Graduate Instruction Sector� and �Undergraduate Instruction Sector�. Number of programs are determined by the historical average number of students in the respective sectors. As historical average number of students increases new programs may be opened and as this average drops below certain threshold levels, some programs are closed. III. 4. Graduate Instruction Quality Sector This is the sector where the graduate instruction quality indicators are calculated. (Figure 3). The main instruction quality indicators are �Graduate Students/Grad Faculty Ratio�, �Actual Graduate Instruction Overhead per Grad Student�, �Actual Average Graduate Class Size� and �Lab Facilities for Graduate Instruction per Graduate Student�. Graduate Instruction Quality Index in a semester is given by:

� ( �Actual Graduate Instruction Overhead per Grad Student�, �Actual Average Graduate Class Size�, �Lab Facilities for Graduate Instruction per Graduate Student�, �Historical Average Research Papers per Grad Faculty� )

The function � (�) is an � additive effect formulation�:

Grad Instr Quality Index = (Eff of Act GI Overhead per Grad Stu on GI Quality +Eff of Grad Class Size on GI Quality+Eff of Research on GI Quality +Eff of Lab Facilities for GI on GI Quality)/4

Each of these effects are graphical functions of the related variables. For instance, �Effect of Actual GI

6 / 34

Overhead per Graduate Student on GI Quality� is a graphical function of �Actual Graduate Instruction Overhead per Graduate Student�. �Assumed Graduate Instruction Quality Index� represents the average graduate instruction quality goal of the faculty members, and �Perceived Grad Instruction Quality Index� is the historical average of Grad Instruction Quality Index which is a typical perception delay formulation that is used in System Dynamics Methodology. �Assumed Undergraduate Instruction Quality Index� and �Perceived Undergraduate Instruction Quality Index� are similar variables that are determined in the Undergraduate Instruction Quality sector. The instruction quality indices are important in determining the Teaching Commitments of faculty members, which are indicators of the long-term attitude of the faculty members about instruction quality. Teaching Commitments are used as inputs to respective Overhead Sectors. III. 5. Undergraduate Instruction Quality Sector The quality indicators about undergraduate instruction are calculated in this sector. This sector resembles to �Graduate Instruction Quality Sector�. �Actual Undergraduate Instruction Overhead per Undergraduate Student�, �Actual Average Undergraduate Class Size�, �Lab Facilities for Undergraduate Instruction per Undergraduate Student� and �Historical Average Research Papers per Faculty� effect the �Period Undergraduate Instruction Quality Index�. The ratio among the historical average of this index, �Perceived Undergraduate Instruction Quality Index� and �Assumed Undergraduate Instruction Quality Index � determine �Indicated UGF Teaching Commitment� together with �UGF Teaching Commitment�. The historical average of �Indicated UGF Teaching Commitment� is �UGF Teaching Commitment� and is used as an input to �Undergraduate Faculty Overhead Sector�. III. 6. Graduate Faculty Instruction Overhead Sector In this sector the instruction overhead loads of graduate faculty members are calculated and the remaining graduate faculty time, after assigning in-class instruction and instruction overhead loads, is divided among research and project activities. (Figure 4). Two different instruction overhead loads are calculated for graduate faculty members: graduate instruction overhead and undergraduate instruction overhead. The main variable that determines the graduate instruction overhead is �Actual FTE Graduate Faculty for Graduate Instruction�. By assuming that a certain amount of in -class instruction implies a certain amount of instruction overhead, �Normal FTE GF for Grad Instruction Overhead� is calculated as:

�Actual FTE GF for Grad Instruction� * �Normal Overhead Ratio for Grad Instruction� �Normal Overhead Ratio for Grad Instruction� represents the instruction overhead load implied by unit in -class instruction load and is determined within certain limits by �Actual Average Grad Class Size�. As average class size increases, instruction overhead implied by unit in-class instruction load increases. A certain portion of �Normal FTE GF for Grad Instruction Overhead� is assigned to teaching assistants through the proc edures explained in the detailed description of �Assistants Sector�. The rest of the implied instruction overhead load is represented by �Indicated FTE GF for Grad Instruction Overhead�. The same approach is used for calculating �Normal FTE GF for Undergraduate Instruction Overhead� and assigning a certain portion of this load to teaching assistants. Thus, �Indicated FTE GF for Undergraduate Instruction Overhead� is determined as:

�Normal FTE GF for Undergraduate Instruction Overhead� - �FTE Assistants for Undergraduate Instruction Overhead�

�Total Indicated Non Instruction FTE Graduate Faculty� is given by:

�Total FTE Graduate Faculty� - [�Total In-Class FTE GF� + �Indicated FTE GF for Graduate Instruction Overhead� + �Indicated FTE GF for Undergraduate Instruction Overhead� + �FTE GF for Other Routine Tasks�

7 / 34

This is the maximum amount of FTE graduate faculty that can be involved in research and project activities by working the normal �Weekly Hours per Faculty�, which is equal to 40 in the base and g ame versions of the model. On the other hand, �Desired FTE GF for Research and Projects� is calculated as a function of �Total Out -of-Classroom FTE GF� which is equal to:

�Total FTE GF� - �Total In Class FTE GF� The other variable that effects �Desired FTE GF for Research and Projects� is �Graduate Faculty Research and Projects Commitment�, which is the arithmetic average of �Graduate Faculty Research Commitment� and �Graduate Faculty Financial Pressure�. �Graduate Faculty Research Commitment� is explain ed in the �Graduate Faculty Research Sector� section and �Graduate Faculty Financial Pressure� is explained in the �Graduate Faculty Projects Sector� section. �Desired FTE GF for Research and Projects� does not have to be equal to �Total Indicated Non Instruction FTE GF�. It can be more or less than this value, as well as being equal to it. If it is less, a certain amount of �Inactive FTE GF� occurs. This gives the total idle graduate faculty hours. If �Desired FTE GF for Research and Projects� is more than �Total Indicated Non Instruction FTE GF�, �Indicated GF Extra Work Load takes a positive value. This variable represents the average extra weekly hours per graduate faculty, that must be worked in order to sustain the desired amount of time for research and project activities. If this extra load is more than a certain amount, a reduction need in overhead activities occurs; this is represented by �GF Overhead Reduction Need� and implies �GF Overhead Reduction Coefficient�, which takes values between zero a nd one. This coefficient depends on �GF Teaching Commitment�, as well. (Figure 4). As the teaching commitment increases the reduction coefficient decreases. �Actual FTE GF for Grad Instruction Overhead� is equal to:

Indicated FTE GF for Grad Instruction Overhead� * [1 - �GF Overhead Reduction Coefficient�] and �Actual FTE GF for UG Instruction Overhead� is equal to:

Indicated FTE GF for UG Instruction Overhead� * [1 - �GF Overhead Reduction Coefficient�] �Actual Non Instruction FTE GF� is given by:

�Total FTE GF� - [�Total In-Class FTE GF� + �Actual FTE GF for Overhead Grad Instruction� + �Actual FTE GF for Overhead UG Instruction� + �FTE GF for Other Routine Tasks�]

�Reduced GF Extra Work Load� represents the average extra weekly hours per graduate faculty member after reducing the instruction overhead, and is given by:

�Desired FTE GF for Research and Projects� - �Actual Non Instruction FTE GF� If this extra load is more than �GF Bearable Extra Work Load�, which is the average maximum weekly extra hours that can be accepted by a graduate faculty member, this means that �Desired FTE GF for Research and Projects� can not be realized entirely. �Available FTE GF for Research and Projects� is calculated as:

�Total Non Instruction FTE GF� + Min{�Reduced GF Extra Work Load�, � GF Bearable Extra Work Load�} �Total Non Instruction FTE GF� is the remaining normal weekly graduate faculty FTE for non instruction activities. �Reduced GF Extra Work Load� and � GF Bearable Extra Work Load� represent the desired and bearable extra work loads, respectively. �Actual FTE GF for Research and Projects� is equal to:

Min {�Desired FTE GF for Research and Projects�, �Available FTE GF for Research and Projects. �Actual FTE GF for Research and Projects� is divided among �Actual FTE GF for Projects� and �Actual FTE GF for Research� according to the relative weighs of �GF Financial Pressure� and �GF Research Commitment�.

8 / 34

III. 7. Undergraduate Faculty Instruction Overhead Sector This is the sector where the instruction overhead loads of undergraduate faculty members are calculated and the undergraduate FTE for research activities and for project activities are determined. This sector resembles to �Graduate Faculty Overhead Sector�. The dynamics in both sectors are similar. The main difference is that only undergraduate but no graduate instruction overhead is assigned to undergraduate faculty, because they are not involved in graduate instruction. �Normal FTE UGF for UG Instruction Overhead� is determined as:

�Actual FTE UGF for UG Instruction� * �Normal Overhead Ratio for UG Instruction� and then divided among undergraduate faculty members and teaching assistants. The share of undergraduate faculty is represented by �Indicated FTE UGF for UG Instruction Overhead�. After that, �Total Indicated Non Instruction FTE UGF� is determined as:

�Total FTE UGF� - [�Total In- Class FTE UGF� + �FTE UGF for Other Routine Tasks�] Then, �Indicated UGF Extra Work Load� is calculated as:

Min { [�Total Indicated Non Instruction FTE UGF� - �Desired FTE UGF for Research and Projects�], 0} If there is extra work load and it indicates an overhead reduction, the overhead is reduced and �Actual Non Instruction FTE UGF� is determined. After �Actual FTE UGF for Research and Projects� is calculated, it is divided among �Actual FTE UGF for Projects� and �Actual FTE UGF for Research� according to the relative weighs of �UGF Financial Pressure� and �UGF Research Commitment�. III. 8. Graduate Faculty Research Sector In this sector the graduate faculty FTE dedicated to sponsored and unsponsored research activities are determined together with the motivations behind them and the outcomes of these activities. �Actual FTE GF for Research� is divided among �Actual FTE GF for Unsponsored Research� and �Actual FTE GF for Sponsored Research� according to the relative weights of �GF Unsponsored Research Commitment� and �GF Sponsored Research Commitment�. These two indicators of commitment are the historical averages of �GF Indicated Unsponsored Research Commitment� and �GF Indicated Sponsored Research Commitment� and take values between zero and one. The variables that determine the values of the indicated commitments are �GF Desired/Realized Research Papers Published�, �GF Research Culture� and the corresponding research recognitions. �GF Desired/Realized Research Papers Published� is the ratio between the current historical average of research papers published each semester and the target average papers published. �GF Research Culture� is the long term attitude of graduate faculty members towards research. It takes values between zero and one. Higher research culture causes higher research commitment. Research recognition represents the long term attitude of the administration towards the related research activities and takes values between zero and one. Administration expresses its recognition by rewards and this increases the research commitment. �Indicated Unsponsored Research Recognition� depends on the unsponsored research papers productivity of the faculty members, namely:

�Period Unsponsored Research Papers Published� / �Total FTE for Unsponsored Research� �Indicated Sponsored Research Recognition� depends on the financial productivity, as well as research papers productivity. Thus, �Indicated Sponsored Research Recognition� is given by:

� ( �Funds-Grants Gotten by Sponsored Research�, �Total FTE for Sponsored Research�, �Financial Concern�, �Period Sponsored Research Papers Published�, �Total FTE for Sponsored Research� )

Here, �Financial Concern� is an index of the administration�s will to get funds and grants through sponsored research. After �Actual FTE GF for Unsponsored Research� and �Actual FTE GF for Sponsored Research� are

9 / 34

determined, the outcomes of research activities are calculated. These outcomes depend on the related faculty FTE and productivity indices. III. 9. Undergraduate Faculty Research Sector This sector is similar to �Graduate Faculty Research Sector�. In this sector, the portion of undergraduate faculty FTE dedicated to sponsored and unsponsored research activities, the motivations behind them and the outcomes of these activities are determined. III. 10. Graduate Faculty Projects Sector This is the sector where the portion of the graduate faculty FTE dedicated to official projects and unofficial projects, the motivations behind them and the outcomes of the official projects activities are determined. (Figure 5). �Actual FTE GF for Projects� is divided among �Actual FTE GF for Official Projects� and �Actual FTE GF for Unofficial Projects� according to the relative weighs of �GF OP Motivation� and �GF UP Motivation�. �GF OP Motivation� and �GF UP Motivation� are the historical averages of �Indicated GF OP Motivation� and �Indicated GF UP Motivation�. �Indicated GF OP Motivation� depends on the ratio of incomes realized through official projects and unofficial projects and �OP-UP mentality�. �OP -UP mentality� is an index of the long term attitude of the faculty members towards doing projects through non-university channels, in order to obtain income. OP-UP mentalities, both for graduate and undergraduate faculty members, take values between zero and one. Higher OP-UP mentality indicates lower tendency for doing unofficial projects. Another variable that effects �Indicated GF OP Motivation� is �OP Attractiveness for Administration�. This is an index of the administration�s attitude towards official projects. �Indicated GF UP Motivation� is determined by �OP -UP Income Ratio� and �OP -UP Mentality GF�. (Figure 5). After �Actual FTE GF for OP� is determined �Gross Income Generated by OP� is calculated as a function of graduate and undergraduate faculty FTE and the related OP productivity levels. �Net Funds-Grants Gotten by OP� is calculated by:

�Gross Income Generated by OP� - [�Income Share for GF on OP� + �Income Share for UGF on OP�] �OP Share per FTE per Semester� is calculated by:

�OP Income Level� * �Weekly Hours per Faculty� * �Weeks per Semester� �OP Income Level� represents the amount of money paid per man -hour of faculty workforce for OP. Number of active weeks per faculty member per semester is 23. Total OP share for faculty members for the current semester is calculated by:

�OP Share per FTE per Semester� * �Total FTE for OP� An important sub-system of this sector consists of �GF Financial Pressure� and the variables that effect it. �GF Financial Pressure� is the historical average of �GF Indicated Financial Pressure�, which is an index of the financial concern of the graduate faculty members. It depends on the ratio between the historical average of the actual salary and the average salary desired by faculty members. III. 11. Undergraduate Faculty Projects Sector This sector is similar to �Graduate Faculty Projects Sector�. In this sector, the portion of undergraduate faculty FTE dedicated to official projects and unofficial projects, the motivations behind them and the funds gotten by official projects are determined. III. 12. Laboratory and Facilities Sector In this sector the laboratory facilities are updated and then assigned to instruction, research and project activities. The criteria for assigning the facilities are the relative amounts of faculty FTE allocated to each

10 / 34

activity. III. 13. Assistants Sector This sector is where the number of assistants, available assistant hours per week and the instruction overhead assigned to these assistants are calculated. �Number of Assistants� is a function of �Number of Graduate Students� and �Assistants/Graduate Students Ratio�. This value is limited by �Assistant Posit ions�, which is a function of the total number of faculty positions and �Faculty/Assistant Positions Ratio�. �FTE Assistants for Instruction Overhead� is first converted to �faculty FTE� units by dividing the total assistant hours by weekly hours per faculty. After the FTE assistants for instruction overhead is computed this way, the total available assistant FTE is distributed among �FTE Assistants for Graduate Instruction Overhead� and �FTE Assistants for Undergraduate Instruction Overhead�. According to these values, indicated instruction overhead loads for graduate and undergraduate faculty members are calculated. These values are used as input in �Graduate Faculty Overhead Sector� and �Undergraduate Faculty Overhead Sector�. IV. BASE RUN OF THE MODEL The �Base Run� of the model is the simulation run made under typical expected set of parameters and input values taken from Bogazici University. The behaviors of the variables, observed in this run are used as reference in interpreting the behaviors of the same variables in sensitivity runs. Base dynamic behaviors of some important variables are shown in Figure 6.1 - 6.6. Figure 6.1 is the dynamic behavior of the variables �Number of Undergraduate Students� and �Number of Graduate Students�. The values of both variables increase through time, but while �Number of Undergraduate Students� increases in a steady pace, the rate of increase in �Number of Graduate Students� decreases as time passes. In Figure 6.2, it is observed that both �Number of Graduate Faculty� and �Number of Undergraduate Faculty� increase in a steady pace. Figure 6.3 is the dynamic behaviors of the instruction loads on the graduate and undergraduate faculty members. The total instruction loads on graduate and undergraduate faculty members are at their operating maximums until period 15. After that period, instruction loads increase gradually towards absolute maximum values, due to high student body. In Figure 6.4 are the weekly hours spent on research and projects by each graduate faculty member. While the weekly hours spent on research activities do not change substantially, weekly hours spent on official projects decrease and weekly hours spent on unofficial projects increase considerably. These behaviors are caused by the increase in �GF OP Motivation� and the decrease in �GF UP Motivation�. The dynamic behavior of �GF OP Motivation� and �GF UP Motivation�, together with that of �GF Research Commitment� are shown in Figure 6.5. The increase in �GF UP Motivation� and the decrease in �G F OP Motivation� are caused by the relative values of income obtained from official projects and unofficial projects, and the mentality of the faculty members. The dynamic behaviors of �Period Research Papers Published�, �Period Unsponsored Research Papers Published�, �Period Sponsored Research Papers Published� are shown in Figure 6.6. Here, all there variables increase steadily. V. VALIDITY OF THE MODEL A crucial step in System Dynamics Methodology is model verification and validation. Verification consists of a series of tests to assure that the computer program is error-free, that the simulation model does what the modeller intends to do [3]. It is basically a question of �correctness�. Validation, on the other hand, has to do with the degree of realism and relevance of the model, with respect to the real problem. Although it is a philosophically deep issue, validation practically means demonstrating that the model is an adequate and useful

11 / 34

description of the real system, with respect to the problem(s) of concern [3]. V.1. Verification First, the twelve sectors that form the model are simulated one by one, isolated from the other sectors. Using these isolated sector simulations, it is easier to apply verification tests and trace any erroneous/illogical results to their origins and correct the potential formulation and/or programming errors. For this purpose, a wide range of short-term test simulations are done, under various test conditions, especially extreme conditions. The built-in unit consistency checking feature of Vensim environment, which checks the equivalence of units on both sides of the equations, was also used in order to eliminate any errors that might be in the equations. After all the sectors were verified, the relationships that link the sectors to each other were verified by running the whole model under different conditions. The units equivalence check was repeated for the model as a whole and the verification tests were completed. V.2. Validation After the verification of the model, validation tests are done. Validation tests are grouped into two: (a) structure validation tests, which are done in order to determine whether the model has an adequate structure, (b) behavior validation tests, which are done in order to determine whether the behavior of the model resembles the behavior exhibited by the real system that was modeled. ([2], [4]). To test the structural validation of the model, extreme condition and sensitivity tests have been applied. V.2.1. Extreme Conditions Tests Extreme condition tests are based on the idea that the behaviors of a given model are far more predictable under extreme conditions than they are under normal conditions. Extreme condition tests are done by simulating the model after setting a certain variable to an extremely high or extremely low value and examining the behavior of key variables. The extreme value of the chosen variable implies certain predictable behaviors by other variables. If the behaviors of one or more key variables are not as they are expected to be, the validity of the model is questioned, the cause of the problematic behavior is traced to the equations which are revised accordingly. Otherwise, i.e. if the behaviors of all the key variables are as they were expected to be, the model passes the extreme condition tests. Numerous extreme condition simulation runs were done on the model . The results of a selection of the results are presented below. No Undergraduate Students The model is simulated for 20 periods (semesters) after setting the number of undergraduate students to zero. The undergraduate instruction and undergraduate instruction overhead loads for both graduate and undergraduate faculty members becomes zero. Some other variables such as �Undergraduate Students/Faculty Ratio�, �Undergraduate Graduation� and �Average Undergraduate Class Size Needed� are also equal to zero, as expected. the model, thus, passes this extreme condition validity test. No Undergraduate Admission Another extreme condition consists of starting with a certain undergraduate student body, but then setting the admission equal to zero. (Figure 7.1-7.5). Under this condition, �Number of UG Students� and �UG Graduation� converge to zero. So do the undergraduate instruction and instructions overhead loads of graduate and undergraduate faculty members and undergraduate students/faculty ratio. All these are logically expected behavior patterns. Extremely High Undergraduate Admission The model is simulated under the condition that undergraduate admission is equal to 10000 students per year. (Figure 8.1-8.5). As expected, under this extreme condition, the instruction loads of the faculty members reach to absolute maximum levels and the average undergraduate class size reaches the maximum value that is allowed, and there is still a big number of canceled undergraduate instruction hours. These performance measures are as expected, hence the model passes this test.

12 / 34

No Undergraduate Faculty The model is simulated under the condition that the number of undergraduate faculty is zero. (Figure 9.1-9.2). Under this condition all undergraduate instruction is assigned to graduate faculty. The instruction load on graduate faculty members rises gradually because of the increasing number of undergraduate students and at period three it reaches the absolute maximum level. At this point �Canceled Undergraduate Lecture Hours� takes a positive value and increases through the rest of the simulation. Undergraduate FTE assigned for research and project activities is equal to zero and this causes all the outputs of these activities, like research papers published, undergraduate faculty members� contribution to official projects to be equal to zero, as expected. No Faculty Another extreme run is made after setting both �Number of Undergraduate Faculty� and �Number of Graduate Faculty� to zero. This makes all faculty workforces and all outputs like research papers and funds gotten by OP an sponsored research equal to zero. All graduate and undergraduate lecture hours are canceled. Extremely Low Salary Another extreme simulation run is made after the salaries of faculty members are set to 50 dollars per month. (The average industry income is around 1300 dollars per month in Turkey.) This increases the number of leaving faculty members and decreases faculty supplies, (candidates for faculty positions). The number of faculty members decreases rapidly and at period eight it reaches 12, its minimum for this run. (Figure 10.1-10.3). (Setting salaries to even lower values causes similar but more steep behaviors.) No Faculty Supply Another extreme condition test is done by setting the faculty supplies (candidates) to zero. (Figure 11.1-11.2). As faculty supplies are zero, �New Graduate Faculty� and �New Undergraduate Faculty� become zero. This gradually decreases the number of faculty members, both graduate and undergraduate, because there are leaving faculty members. In the long run, the number of faculty members becomes zero. Extremely High Instruction Overhead Ratio The model is tested after increasing normal instruction overhead hours per lecture hour ratio to an average of 10 from an average of 2. (Figure 12.1-12.3). As a result, all faculty FTE that is left after graduate and undergraduate lecture hours are assigned are dedicated to instruction overhead. Thus, faculty members can not find time for research and project activities during normal work hours. This leads faculty members to do research and project activities by working extra hours. But as average extra hours per faculty are limited by bearable extra work load limits, the total time a faculty member dedicates to research and project activities also limited. V.2.2. Sensitivity Analyses After the extreme condition tests, a series of sensitivity analyses were made, in order to determine whether the sensitivity of the base model is within acceptable limits. A set of sensitivity runs are made by simulating the model with different values of a certain variable and examining the behaviors of the key variables to determine if the behaviors of these variables differ within acceptable ranges. If the behaviors of one or more key variables differ more than expected, the validity of the model is questioned. If the differences are within acceptable ranges the hypothesis can not be rejected and other sensitivity runs with different values of some other variables are made. Numerous sensitivity tests were done in order to test the validity of the model. Some examples of these sensitivity tests are given below for illustration. Sensitivity Runs with Different Values of �GF OP UP Mentality� The first set of sensitivity runs are done by giving different values to �GF Official Projects-Unofficial Projects Mentality�, which represents the relative strength of preferring Official Projects to Unofficial Projects. (Figure 13.1-13.4). The values are changed from (0.4) to (0.1), (0.2), (0.8) and (1.0). �GF OP Motivation� had the

13 / 34

initial value 0.75 and �GF UP Motivation� had the initial value 0.5 in each run. It was observed that, compared to the base runs �GF OP Motivation� converges to higher values as the value of the variable �GF OP UP Mentality� is set to higher values. �GF UP Motivation�, on the other hand, converges to lower values as �GF OP UP Mentality� increases. The allocation of graduate faculty FTE for projects between Official Projects and UP also change according to the respective motivations, in each run. Sensitivity Runs with Different Values of �Average Hours per Graduate Student� Second set of sensitivity runs are done by giving different values to �Average Hours per Graduate Student�, which indicates the average weekly lecture hours needed by each graduate student. (Figure 14.1-14.3). Four runs are made and the values of �Average Hours per Graduate Student� during these runs were (3), (7), (15), (25). The value of this variable in the base run was (15). As expected, higher values of �Average Hours per Graduate Student� cause �Total FTE Needed for Graduate Instruction� to be higher, which increases the graduate instruction and graduate instruction overhead loads on graduate faculty members. Sensitivity Runs with Different Values of �GF Research Papers Productivity� Another set of sensitivity runs are done by giving different values to the variables �GF Sponsored Research Papers Productivity� and �GF Unsponsored Research Papers Productivity�. (Figure 15.1 -15.3). The values of these variables are calculated by multiplying the maximum possible productivity value by the effects of respective commitments, laboratory facilities allocated to research activities and the number of graduate students. The maximum productivity values in the sensitivity runs are (1), (2), (4) and (7) paper per FTE. This value is equal to (4) in the base run. It is observed that the related variables like �Period GF Sponsored Research Papers Published�, �Period GF Unsponsored Research Papers Published� and �GF Historical Average Research Papers Published per Graduate Faculty� take higher values as the maximum productivity value of �GF Sponsored Research Papers Productivity� and �GF Unsponsored Research Papers Productivity� take higher values. V.2.3. Behavior Validation After the structural validation tests were completed the behavior of the model was compared with the data from Bogazici University. (Figure 16.1-16.3 and Figure 17.1-17.3). Most of the real data were taken from the 1994 edition of the yearly document �Bogazici University in Numbers� [22]. This document includes a wide range of data on many aspects of Bogazici University like students, faculty members, publications and financial figures. Some other data about other aspects of the model, like official projects and available faculty and assistant positions were gathered by interviews and used for calibrating the model. However these data could not be used for behavior validation purposes, because they were mostly single or a few data points. An exact matching between real data and model data points is not required for validating the model, because a System Dynamics model is not designed to include the internal and external details and random factors that are needed in short term forecasting [2]. The purpose of a system dynamics model is to generate the broad dynamic behavior patterns of the system, in the long term. Thus, what is required is the matching of the major patterns of behavior of the model and the real system, rather than individual data points. All the data about Bogazici University that could be used for behavior validation were compared with the behavior of the model and a broad resemblance between the model behavior and the behavior of the real system, was obtained. (Figure 16.1-16.3 and Figure 17.1-17.3). Thus, it was concluded that the model is behaviorally acceptable.

VI. EXPERIMENTS WITH THE MODEL Some demonstrative simulation experiments are made with the presented model in order to show its simulation capabilities. Two of these experiments and their results are explained below. Graduate Study Orientation versus Undergraduate Study Orientation This simulation experiment is designed to determine the effects of different ratios of undergraduate

14 / 34

students/graduate students. (Figure 18.1-18.6). Two simulation runs are made to compare the results of increasing and decreasing undergraduate students/graduate students ratios. The number of undergraduate students is increased slower in the second run than it is in the first run. (Figure 18.1). On the other hand, number of graduate students is increased faster in the second run. (Figure 18.2). The number of undergraduate students and the number of graduate students have the same initial values in both runs. Both simulation runs begin with a undergraduate students/graduate students ratio that is around 5. In the first run, which represents a high undergraduate study orientation, the ratio increases gradually and reaches to 12. In the second run, which represents a high graduate study orientation, this ratio decreases and finally reaches to 2.5. (Figure 18.3). The first effect of the change in student profile is the change in instruction load allocations. The instruction load per undergraduate faculty member takes lower values in the second run because of the slower increase in the number of undergraduate students. (Figure 18.4). The total instruction load per graduate faculty also takes lower values in the second run. Though the faster increase in the number of graduate students causes the graduate instruction load per graduate faculty member to take higher values, the undergraduate instruction load per graduate faculty take substantially lower values because of the slower increase in the number of undergraduate students. (Figure 18.5). Lower instruction loads cause the faculty FTE dedicated to research and project activities to be higher in the second run and this brings higher research and project outputs, in terms of both research papers and funds obtained from research and project activities, namely project income. As observed from the results of the simulation runs, an important effect of higher graduate study orientation is higher research productivity per faculty member in the long run. (Figure 18.6). Experiments with Different Undergraduate Class Sizes Another set of simulation experiments is done in order to determine the effects of different undergraduate class sizes. (Figure 19.1-19.6). The base values for �Desired Undergraduate Class Size� and �Maximum Allowable Undergraduate Class Size� are 50 and 100, respectively. These values are used in run 2, as well. In the first run the values for �Desired Undergraduate Class Size� and �Maximum Allo wable Undergraduate Class Size� are 20 and 40, respectively and in the third run they are 80 and 160. The direct effect of different class sizes are observed on total undergraduate instruction hours needed (Figure 19.1) and total FTE for undergraduate instruction (Figure 19.2). This in turn affects the instruction loads per graduate and undergraduate faculty member. Lower undergraduate class sizes bring higher undergraduate instruction loads for both graduate (Figure 19.3) and undergraduate faculty members (Figure 19.4). Extremely high instruction loads cause more faculty members to leave and less new faculty members to apply. This causes the number of faculty members to decrease considerably. (Figure 19.5). Low undergraduate class sizes bring up a higher need for part-time faculty workforce. But the need for part-time faculty workforce is limited by a certain percentage of full-time faculty workforce. As full-time faculty workforce decreases because of the high instruction loads, the upper limit for part-time faculty workforce usage decreases, as well. The discrepancy between needed and delivered undergraduate instruction is tried to be eliminated by increasing the average class size, but if the maximum limit for class size is low (as in the first simulation run), the cancellation of some undergraduate lecture hours is inevitable. (Figure 19.6).

VII. THE UNIVERSITY GAME In the next step of the research, the model is converted into an interactive dynamic simulation game. Venapp feature of Vensim software is used in building the game version [9]. In the game, the player plays the role of a university policy-maker, who is trying to seek a delicate balance among the main academic functions of the university, in order to get better output from these activities, both in terms of quality and quantity. The player does not have too many decision opportunities, because most of the factors are imposed by the environment the university exists in. The objective of the player is to make six decisions, so as to improve the performance indicators of the university, within the limitations imposed by outside factors. These decisions are New Graduate Students, New Undergraduate Students, Graduate Faculty Hiring Decision, Undergraduate Faculty Hiring Decision, Share on Income-generating-projects per Faculty Member and Weekly Release Time per Graduate Faculty Member (Figure 20). Sixty different performance indicators are displayed after each decision period. There is also detailed information option that the player can use in order to carry out more detailed causal analysis of the dynamics of the model (Figure 20).

15 / 34

The game consists of a series of screens which are linked between themselves. Only one screen can be observed at a time. Some of these screens are just query screens that ask the player which name he wants to give to the file that the current game data will be kept in, whether he wants to end the game, whether he wants to exit the simulation environment, etc. The main game screen is divided into four display boxes (Figure 20). �Game Control� box includes buttons to be used in order to end the game, exit the simulation environment or see the game conditions. There are also two display objects in this box, that show final time and current time. �Decisions� box includes the decision entries and the �Advance� button. The user enters the values he decided on �New Graduate Students�, �New Undergraduate Students�, �Graduate Faculty Hiring Decision� and �Under -graduate Faculty Hiring Decision�, which represent the number of new graduate and under-graduate faculty members he wants to hire, for the current semester; �Share on Income-generating-projects per Faculty Member�, which indicates the amount of money that will be paid to each man-hour of faculty workforce from income generating projects and Weekly Release Time per Graduate Faculty Member, which indicates the amount of reduction in maximum weekly instruction load of graduate faculty members. When �Advance� button is pressed the simulation proceeds one period (semester) and the new values are calculated and displayed. The �Help� button calls a small frame that includes information about the decision variables. �Main Indicators� box displays the values of the 30 variables at the current time period. The buttons which have the names of the decision and indicator variables display the behavior patterns of the related variables, when pressed. The button �More Indicators� calls another set of 30 variables. The button in the �Detailed Analysis� box executes the link to �Detailed Analysis Screen�, which include certain analysis tools. These are �Causes Tree�, which displays the tree of variables that effect a certain (active) variable, �Uses Tree�, which displays a tree of variables that are effected by the active variable, �Loops�, which displays the causal loops that include the active variable, �Graph�, which displays the time graph for the active variable, �Causes Graph�, which displays the graphs for the active variable and the variables that effect it, �Causes Table�, which displays a table that includes the values of the active variable and the variables that effect it. A series of verification and validation tests were done on the interactive simulation game. The behavior of the variables in case of extreme player decisions were tested and several sensitivity tests were done. (These test results can not be shown due to space restrictions. See [8]). The necessary improvements both in the model and the game were made according to the results of these tests. The resulting game is believed to be a robust and valid interactive simulation environment. A group of players with different academic degrees and different orientations were invited to play the game. Among the players were graduate students, teaching and research assistants, faculty members and managers. We are currently in the process of obtaining and analyzing diverse game results. Some initial results are presented below for illustration. Game Results of a Research Oriented Faculty Member The first player is a faculty member who has high research interests and little interest in income-generating projects, even if they are realized through official, university channels. During the game, except for period 6, he hires more graduate faculty than undergraduate faculty. He gives considerably release time to graduate faculty (2 h./week on the average). He decreases Official Projects share for faculty members gradually. (Figure 21.1) As a result of his emphasis on graduate study and research, he obtains a remarkable increase in the number of research papers per semester. The average research papers per faculty member has increased, as well (Figure 21.2). On the other hand, decreasing the Official Projects share for faculty members causes Official Projects motivations of the faculty members to decrease and unofficial project (UP) motivations to increase. These, in turn, causes the weekly hours dedicated to Official Projects by the faculty members to decrease and the hours dedicated to UP to increase. In Figure 21.3 the behaviors of project loads of graduate faculty members and Total FTE for Official Projects are shown; the patterns for undergraduate faculty members are similar. Game Results of a Balanced Faculty Member Another player is a faculty member who tries to strike a balance between instruction, research and project activities. He puts some emphasis on research by giving release time for graduate faculty, but also encourages official projects by giving a considerable official projects income share to faculty members. (Figure 22.1).

16 / 34

He obtains an increase in the number of the research papers per semester and average research papers per faculty member. (Figure 22.2). Although these figures are somewhat less than those obtained by the first player, the differences are not very significant. The important difference between the decisions of the Research Oriented Faculty Member and the Balanced Faculty Member is that the latter put the necessary emphasis on official projects activities, which the first neglects. That causes a difference between the number of FTE faculty that deal with official projects, as well as differences in the distribution of the project loads among official and unofficial projects. (Figure 21.3 and Figure 22.3).

VIII. CONCLUSIONS We present an interactive simulation model, on which various problems involved in managing the academic aspects of a university can be analyzed and certain policies for overcoming these problems can be tested. The research consists of two phases: A system dynamics model (the � engine� b ehind the game) of the academic aspects of a university system is built first. Numerous verification, validation and sensitivity tests are carried out on the model. The model is calibrated using data from Bogazici University and the dynamic behavior patterns of the model are found to be consistent with the major historical time patterns obtained from Bogazici University. The interactive simulation game, based on the system dynamics model was constructed using Venapp facility of Vensim software. This game version of the model was validated by a series of tests and the necessary improvements were made in the model and the gaming interface. A group of players with different academic degrees, backgrounds and orientations played the game. The comparison of the game results of the players reveals that players with different orientations emphasize different performance measures of the university. Research results obtained so far suggest that the interactive simulation model promises to be not only a useful technology to support strategic management, but also a laboratory for theoretical research on how to best deal with strategic university problems. We are currently in the process of initiating further research on the existing model and the gaming interface. The model will be extended to include more aspects of the university system, such as budget considerations, support staff and in general more detailed representations of variables such as facilities, infrastructure and projects. Also, the gaming interface will be enhanced to include various additional user-friendly features.

REFERENCES [1] D.F. Andersen et al., �Issues In Designing Interactive Games Based on System Dynamics Models�, Proceedings of the 1990 International System Dynamics Conference, Vol. 1, pp. 31-45, Chestnut Hill-Massachusetts, 1990. [2] Y. Barlas, �Multiple Tests for Validation of System Dynamics Type of Simulation Models�, European Journal of Operational Research, Vol. 42, No. 1, pp. 59-87, 1989. [3] Y. Barlas and S. Carpenter, �Philoso phical Roots of Model Validation: Two Paradigms�, System Dynamics Review, Vol. 6, No. 2, pp. 148-166, 1990. [4] Y. Barlas, �Formal Aspects of Model Validity and Validation in System Dynamics�, System Dynamics Review, Vol. 12, No. 3, pp. 183-210, 1996. [5] Y. Barlas and V.G. Diker, �An Interactive Dynamic Simulation Model of a University Management System�, Proceedings of Symposium on Applied Computing ’96, pp. 120-128, Philadelphia, Pennsylvania, 1996. [6] E. Benjamin, �A Faculty Response to the Fiscal Cris is: From Defense to Offense�, Higher Education Under Fire, (Ed: BØrubØ, M. and Nelson, C.), Routledge, NY, 1995. [7] E.F. Cheit, The New Depression in Higher Education, Mc-Graw Hill, New York, 1971. [8] V.G. Diker, An Interactive Dynamic Simulation Model of a University Management System, M.S. Thesis, Institute for Graduate Studies in Science and Engineering, Bogazici University, 1995. [9] R.L. Eberlein and D.W. Peterson, �Understanding Models with Vensim�, Modeling for Learning Organizations, (Ed: J.D.W. Morecroft and J.D. Sterman), Productivity Press, Portland, OR, 1994. [10] J.W. Forrester, Industrial Dynamics, Productivity Press, Portland, OR, 1961. [11] A.K. Graham and P.M. Senge, �Computer -Based Case Studies and Learning Laboratory Projects�,

17 / 34

System Dynamics Review, Vol. 6, No. 1, pp. 100-106, 1990. [12] K. Gürüz et al., Higher Education, Science and Technology in Turkey and Abroad (in Turkish), TÜSÝAD Publications, Istanbul, 1994. [13] M. Mahmoud and P.Genta, �Microworld of an Open University: A Strate gic Management Learning Laboratory�, Proceedings of the 1993 International System Dynamics Conference, Cancun-Mexico, pp. 318-327, 1993. [14] D. Meadows, �Gaming to Implement System Dynamics Models�, Computer-Based Management of Complex Systems, (Ed: P. Milling and E.O.K. Zahn), Springer Verlag, Berlin, 1989. [15] J.D.W. Morecroft, �System Dynamics and Microworlds for Policymakers�, European Journal of Operational Research, Vol. 35, No. 1, pp. 301-320, 1989. [16] G. Richardson and A.L. Pugh, Introduction to System Dynamics Modeling, Productivity Press, Portland, OR, 1991. [17] K. Saeed, �The Dynamics of Collegial Systems in the Developing Countries�, Proceedings of the 1993 International System Dynamics Conference, Cancun-Mexico, pp. 444-453, 1993. [18] Z. Sinuany-Stern, �A Financial Planning Model for a Multi -campus College�, Socio Economic Planning Sciences, Vol. 18, No. 2, pp. 135-142, 1984. [19] G. Ulusoy (ed.), ATAS-Bogazici University Science Conference, Bogazici University Publications, Istanbul, 1994. [20] S.R. Vemuri, �A Simulation Based Methodology for Modeling a University Research Support Service System�, Socio Economic Planning Sciences, Vol. 16, No. 3, pp. 107-120, 1982. [21] -, (OECD Report), Financing Higher Education, OECD, Paris, 1990. [22] -, Bogazici University in Numbers (in Turkish), Bogazici University Publications, Istanbul, 1994.

18 / 34

Assistants Sector

LaboratoryFacilities Sector

UndergraduateFaculty Projects

Sector

Graduate FacultyProjects Sector

UndergraduateFaculty Research

Sector

Graduate FacultyResearch Sector

UndergraduateFaculty InstructionOverhead Sector

Graduate FacultyInstruction

Overhead Sector

UndergraduateInstruction Quality

Sector

GraduateInstruction Quality

Sector

UndergraduateInstruction Sector

GraduateInstruction Sector

Figure 1. Global Sector Diagram

19 / 34

Number of Grad Students

Grad Stu 1 to ThesisGrad Admission Grad Graduation

Number ofThesis

Students

Number ofGrad Students

1

<Indicated UGF Hiring>

Total Indic Faculty Hiring

<Number of Grad Fac>

Act UG Instr Load per GF

Eff of Instr Load on Leaving GF

Eff of Salary on Leaving GF

Eff of Instr Load on GF Supply

Eff of Salary on GF Supply Maximum Grad Programs Possible

Act Grad Instr Load per GF

Min Instr Hours per Grad Program

Number ofGrad Programs

Net Cha in No of Grad Students

Hist Avr Number ofGrad Students

Normal Students per Grad ProgramMin Students per Grad Program

Closed Grad ProgramsNew Grad Programs

Grad Programs DecreaseGrad Programs Increase

New Grad FacultyGF that Leave

Vacant FacultyPositions

Indicated GF Supply

Indicated GF Hiring

Part Time FTE for Grad Instr

GF Hiring Decision

Act Instr Load per GF

New Faculty Positions

Additional Part Time FTE for Grad Instr

Act FTE GF for UG Instr

Act FTE GF For Instr

Number ofGrad Fac

Act FTE GF For Grad Instr

Absolute Max Inst Hours per GF

Absolute Total FTE GF for Inst

Additional FTE GF for Grad Inst

Act Average Grad Class Size

Unsatisfied Need for Grad Instr

Average Grad Class Size Needed

Extra FTE Need for Grad Inst

Operating Part Time FTE for Grad Instr

Grad Inst Discrepancy

Total Grad Student HoursTotal Grad Instr Hours Needed

Total FTE Needed for Grad Instr

Operating Max Inst Hours per Grad Fac

Cancelled Grad Lecture Hours

Operating Total FTE Grad Faculty for Instr

Figure 2. Stock-Flow Diagram of the Graduate Instruction Sector

20 / 34

GF Hist Avg Res Papers Pub<Desired Average Grad Class Size>

Lab Facilities for Grad Instr per Grad Stu

Grad Students GF Ratio FTE Assist for Grad Instr Overhead

Act FTE GF for Overhead G Instr

Act GI Overhead per Grad StuLab Facilities for Grad Instr

Eff of Lab Facilities for GI on GI Quality

Act Average Grad Class Size

Number of Grad Fac

Assumed Perceived UG Instr Quality

<Number of Grad Students>

Cha in Perceived Grad Instr Qua Index

Eff of Reseach on GI Quality

GF Hist Avg Res Papers Pub per GF

Eff of Grad Class Size on GI Quality

Eff of Act GI Overhead per Grad Stu on GI Quality

Perc Grad Instr Qua Index ATPeriod Grad Instr Quality Index

Perceived GradInstr Quality Index

Assumed Grad InstrQuality Index

GF Teaching Culture

Cha in GF Teach Comm

Indicated GF Teach Comm

GF Teach Comm AT

Assumed Perceived Grad Instr QualityGF TeachingCommitment

Figure 3. Graduate Instruction Quality Sector

21 / 34

<Total In Class FTE GF>

<FTE GF for Other Routine Tasks>

<Act Non Instr FTE GF>

<Act FTE GF for UG Instr>

Desired FTE GF for ProjectsDesired FTE GF for Research

GF Research Commitment

Weekly Inactive Hours per GF

Act Overhead Load per GF

Act UGI Overhead Load per GF

Act Average Grad Class Size

<Total FTE GF>

Act GI Overhead Load per GF

GF Project vs Research Orientation

Normal FTE GF for UG Instr OverheadNormal FTE GF for G Instr Overhead

Inactive FTE GFAct FTE GF for Research and Projects

<Number of Grad Students>

Min FTE GF Required for Research

Total FTE UGF

FTE GF for Other Routine Tasks

<Number of UG Students>

<Number of Grad Students>

FTE Needed for Other Routine Tasks

Act FTE GF for Grad Instr

Act FTE GF for UG Instr

Normal Overhead Ratio for UG Instr

Normal Overhead Ratio for Grad Instr

GF Financial Pressure

Total FTE GF

Act FTE GF for Research

Act FTE GF for Projects

Desired FTE GF for Research and Projects

<Total FTE GF>

Available FTE GF for Research and Projects

GF Bearable Extra Work Load

Act Non Instr FTE GF

Reduced GF Extra Work Load

GF Overhead Reduction Need

Act FTE GF for Overhead UG Instr

Act FTE GF for Overhead G Instr

GF Overhead Reduction Coef

Indicated FTE GF for UG Instr OverheadIndicated FTE GF for G Instr Overhead

Total In Class FTE GF

Total Indicated Non Instr FTE GF

Indicated GF Extra Work Load

GF Teaching Commitment

Act Average UG Class Size

Figure 4. Graduate Faculty Overhead Sector

22 / 34

<Act FTE GF for Official Projects>

<Act FTE GF for Official Projects>

<Act FTE GF for Projects>

<Official Projects Share for Faculty>

Cha in Avg GF Monthly Income

UP Income LevelDirect UP Load per GFDirect Official Projects Load per GF

Monthly UP Income per GFMonthly Official Projects Income per GF

Average GF Monthly Income AT

<UP Income Level>

Official Projects UP Income Ratio

Direct FTE GF for Official Projects

Hours Spent for UP per GFAct Official Projects Load per GF

<GF Project vs Research Orientation>

<Act FTE GF for Research and Projects>

<Total FTE for Official Projects>

<UGF Official Projects Productivity>

<Act FTE UGF for Official Projects>

Indicated Official Projects Attr for Admin

Official Projects Attr for Admin AT

<Lab Facilities for Official Projects>Official Projects

Attractiveness forAdministration

Eff of Lab Faci for Official Projects

Cha in Official Projects Attr for Admin

Income Share for GF on Official Projects

Gross Income Generated by Official Projects

Act FTE GF for Projects

Act FTE GF for UP

Act FTE GF for Official Projects

GF Official Projects Productivity

Eff of GF Official Projects Mot

Eff of Mentality GF

OfficialProjects UPMentality GF

Official Projects Share for Faculty

Expected Income

Eff of Income Ratio GF

Indicated GF UP Mot

GF UP Mot AT

Cha in GF UP MotGF UP

Motivation

<Income Share for UGF on Official Projects>

Net Income for University by Official Projects

<Act FTE UGF for Official Projects>

Indicated GF Official Projects Mot

GF Official Projects Mot AT

Cha in GF Official Projects Mot

GF OfficialProjects

Motivation

GF Period Salary

GF Expected Monthly Income

GF Act vs Exp Ratio of Monthly Income

GF Research Commitment

Average SalaryHist AverageGF Monthly

Income

GF Fin Press AT

Cha in GF Financial Pressure

GF FinancialPressure

GF Indicated Fin Press

Figure 5. Graduate Faculty Projects Sector

23 / 34

Number of Students

8,0001,400

4,000700

00

0 2 4 6 8 10 12 14 16 18 20Time (semester)

Number of UG Students - BASE_RUNNumber of Grad Students - BASE_RUN

Figure 6.1 Number of Students in the Base Run

Number of Faculty Members

400400400

200200200

000

0 2 4 6 8 10 12 14 16 18 20Time (semester)

Number of Grad Fac - BASE_RUNNumber of UG Fac - BASE_RUNNumber of Faculty - BASE_RUN

Figure 6.2 Number of Faculty Members in the Base Run

Instruction Loads

10101010

0000

0 2 4 6 8 10 12 14 16 18 20Time (semester)

Act Grad Instr Load per GF - BASE_RUNAct UG Instr Load per GF - BASE_RUNAct Instr Load per GF - BASE_RUNAct Instr Load per UGF - BASE_RUN

Figure 6.3 The Dynamic Behaviors of Instruction Loads

in the Base Run

Research and Project Loads per Graduate Faculty Member

121212

666

000

0 2 4 6 8 10 12 14 16 18 20Time (semester)

Act Research Load per GF - BASE_RUNAct Official Projects Load per GF - BASE_RUNHours Spent for UP per GF - BASE_RUN

Figure 6.4 GF Research and Project Loads in the Base

Run

Graduate Faculty Motivations

111

.5

.5

.5

000

0 2 4 6 8 10 12 14 16 18 20Time (semester)

GF Research Commitment - BASE_RUNGF Official Projects Motivation - BASE_RUNGF UP Motivation - BASE_RUN

Figure 6.5 Graduate Faculty Motivations in the Base Run

Research Papers

650650650

325325325

000

0 2 4 6 8 10 12 14 16 18 20Time (semester)

Period Res Pap Pub - BASE_RUNPeriod Spons Res Pap Pub - BASE_RUNPeriod Unsponsored Res Pap Pub - BASE_RUN

Figure 6.6 Research Papers per Semester in the Base Run

FIGURES 6.1-6.6 Dynamic Behavior of the Model in the Base Run (Reference Run)

24 / 34

Graph for Number of UG Students

4,000

3,000

2,000

1,000

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Number of UG Students - EXT_632 students

Figure 7.1. Number of UG Students in Case of No UG Admission

Graph for UG Graduation

400

300

200

100

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

UG Graduation - EXT_632 students/semester

Figure 7.2. UG Graduation in Case of No UG Admission

Undergraduate Instruction Load per Faculty Member

49

24.5

00

0 2 4 6 8 10 12 14 16 18 20Time (semester)

Act UG Instr Load per GF - EXT_632Act Instr Load per UGF - EXT_632

Figure 7.3. UG Instruction Loads in Case of No UG Admission

Undergraduate Instruction Load per Faculty Member

612

36

00

0 2 4 6 8 10 12 14 16 18 20Time (semester)

Act UGI Overhead Load per GF - EXT_632Act Overhead Load per UGF - EXT_632

Figure 7.4. Undergraduate Instruction Overhead Loads in Case of No UG Admission

Graph for UG Student Faculty Ratio

20

15

10

5

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

UG Student Faculty Ratio - EXT_632 students/faculty

Figure 7.5. UG Student/Faculty Ratio in Case of No UG Admission

Graph for Number of UG Students

40,000

30,000

20,000

10,000

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Number of UG Students - EXT_633 students

Figure 8.1. Number of UG Students under Extreme Undergraduate Admission

FIGURES 7.1-7.5 Results of Extreme Condition Test: No Undergraduate Admission

FIGURE 8.1 Result of Extreme Condition Test: Extreme Undergraduate Admission (10,000/year)Admission

25 / 34

Graduate Faculty Instruction Loads

101010

555

000

0 2 4 6 8 10 12 14 16 18 20Time (semester)

Act Grad Instr Load per GF - EXT_633Act UG Instr Load per GF - EXT_633Act Instr Load per GF - EXT_633

Figure 8.2. Instruction Loads per Graduate Faculty under Extreme Undergraduate Admission

Graph for Act Instr Load per UGF

20

17

14

11

80 2 4 6 8 10 12 14 16 18 20

Time (semester)

Act Instr Load per UGF - EXT_633

Figure 8.3. UG Instruction Load per UG Faculty under Extreme Undergraduate Admission

Graph for Act Average UG Class Size

200

150

100

50

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Act Average UG Class Size - EXT_633 students

Figure 8.4. Average UG Class Size under Extreme Undergraduate Admission

Graph for Cancelled UG Lecture Hours

8,000

6,000

4,000

2,000

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Cancelled UG Lecture Hours - EXT_633 hours/week

Figure 8.5. Canceled UG Lecture Hours under Extreme Undergraduate Admission

Graduate Faculty Instruction Loads

101010

555

000

0 2 4 6 8 10 12 14 16 18 20Time (semester)

Act Grad Instr Load per GF - EXT_634Act UG Instr Load per GF - EXT_634Act Instr Load per GF - EXT_634

Figure 9.1. GF Instruction Loads in Case of No Undergraduate Faculty

Graph for Cancelled UG Lecture Hours

2,000

1,500

1,000

500

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Cancelled UG Lecture Hours - EXT_634 hours/week

Figure 9.2. Canceled UG Lecture Hours in Case of No Undergraduate Faculty

FIGURES 8.2-8.5 Results of Extreme Condition Test: Extreme Undergraduate Admission (10,000/year)Admission

FIGURES 9.1-9.2 Results of Extreme Condition Test: No Undergraduate Faculty

26 / 34

Faculty Members that Leave

10060

5030

00

0 2 4 6 8 10 12 14 16 18 20Time

GF that Leave - EXT_636UGF that Leave - EXT_636

Figure 10.1. Faculty Members that Leave in Case of Extremely Low Salary

Potential Faculty Supply

2020

1010

00

0 2 4 6 8 10 12 14 16 18 20Time

Indicated GF Supply - EXT_636Indicated UGF Supply - EXT_636

Figure 10.2. Faculty Supply in Case of Extremely Low Salary

Number of Faculty Members

350350350

175175175

000

0 2 4 6 8 10 12 14 16 18 20Time

Number of Grad Fac - EXT_636Number of UG Fac - EXT_636Number of Faculty - EXT_636

Figure 10.3. Number of Faculty in Case of Extremely Low Salary

New Faculty Members

.2

.2

.1

.1

00

0 2 4 6 8 10 12 14 16 18 20Time

New Grad Faculty - EXT_637New UG Faculty - EXT_637

Figure 11.1. New Faculty Members in Case of Zero Faculty Supply

Number of Faculty

400400400

200200200

000

0 2 4 6 8 10 12 14 16 18 20Time (semester)

Number of Grad Fac - EXT_637Number of UG Fac - EXT_637Number of Faculty - EXT_637

Figure 11.2. Number of Faculty in Case of Zero Faculty Supply

Overhead Loads per Faculty Member

40404040

0000

0 2 4 6 8 10 12 14 16 18 20Time

Act GI Overhead Load per GF - EXT_638Act UGI Overhead Load per GF - EXT_638Act Overhead Load per GF - EXT_638Act Overhead Load per UGF - EXT_638

Figure 12.1. Instruction Overhead Loads in Case of Extremely High Instruction Overhead Ratio

FIGURES 10.1-10.3 Results of Extreme Condition Test: Extremely Low Salary ($50/m.)

FIGURES 11.1-11.2 Results of Extreme Condition Test: Zero Faculty Supply

FIGURE 12.1 Result of Extreme Condition Test: Extremely High Instruction Overhead Ratio(10 times the base value)

27 / 34

Research and Project Loads per Graduate Faculty Member

101010

555

000

0 2 4 6 8 10 12 14 16 18 20Time

Act Research Load per GF - EXT_638Act IGP Load per GF - EXT_638Act OUP Load per GF - EXT_638

Figure 12.2. GF Research and Project Loads in Case of Extremely High Instruction Overhead Ratio

Research and Project Loads per Under-graduate Faculty Member

101010

555

000

0 2 4 6 8 10 12 14 16 18 20Time

Act Research Load per UGF - EXT_638Act IGP Load per UGF - EXT_638Act OUP Load per UGF - EXT_638

Figure 12.3. UGF Research and Project Loads in Case of Extremely High Instruction Overhead Ratio

Graph for GF Official Projects Motivation

.8

.6

.4

.2

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

GF Official Projects Motivation - SEN1_01 dimensionlessGF Official Projects Motivation - SEN1_02 dimensionlessGF Official Projects Motivation - SEN1_03 dimensionlessGF Official Projects Motivation - SEN1_04 dimensionless

Figure 13.1. GF OP Motivation for Different Values of GF OP UP Mentality

Graph for GF UP Motivation

1

.75

.5

.25

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

GF UP Motivation - SEN1_01 dimensionlessGF UP Motivation - SEN1_02 dimensionlessGF UP Motivation - SEN1_03 dimensionlessGF UP Motivation - SEN1_04 dimensionless

Figure 13.2. GF UP Motivation for Different Values of GF OP UP Mentality

Graph for Act FTE GF for Official Projects

80

60

40

20

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Act FTE GF for Official Projects - SEN1_01 FTEAct FTE GF for Official Projects - SEN1_02 FTEAct FTE GF for Official Projects - SEN1_03 FTEAct FTE GF for Official Projects - SEN1_04 FTE

Figure 13.3. Actual FTE GF for OP for Different Values of GF OP UP Mentality

Graph for Act FTE GF for UP

60

45

30

15

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Act FTE GF for UP - SEN1_01 FTEAct FTE GF for UP - SEN1_02 FTEAct FTE GF for UP - SEN1_03 FTEAct FTE GF for UP - SEN1_04 FTE

Figure 13.4. Actual FTE GF for UP for Different Values of GF OP UP Mentality

FIGURES 12.2-12.3. Results of Extreme Condition Test: Extremely High Instruction Overhead Ratio(10 times the base value)

FIGURES 13.1-13.4. Results of Sensitivity Test: Different Values of GF OP UP Mentality; (decreased to 0.1 in run 01, to 0.2 in run 02 and increased to 0.8 in run 03, 1.0 in run 04 from 0.4 in the base run.)

28 / 34

Graph for Total FTE Needed for Grad Instr

60

45

30

15

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Total FTE Needed for Grad Instr - SEN2_01 FTETotal FTE Needed for Grad Instr - SEN2_02 FTETotal FTE Needed for Grad Instr - SEN2_03 FTETotal FTE Needed for Grad Instr - SEN2_04 FTE

Figure 14.1. The Behavior of Total FTE Needed for Grad Instruction for Different Values of Average Hours per Graduate Student

Graph for Act Grad Instr Load per GF

8

6

4

2

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Act Grad Instr Load per GF - SEN2_01 hours/(week*faculty)Act Grad Instr Load per GF - SEN2_02 hours/(week*faculty)Act Grad Instr Load per GF - SEN2_03 hours/(week*faculty)Act Grad Instr Load per GF - SEN2_04 hours/(week*faculty)

Figure 14.2. Graduate Instruction Load per GF for Different Values of Average Hours per Graduate Student

Graph for Act GI Overhead Load per GF

4

3

2

1

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Act GI Overhead Load per GF - SEN2_01Act GI Overhead Load per GF - SEN2_02Act GI Overhead Load per GF - SEN2_03Act GI Overhead Load per GF - SEN2_04

Figure 14.3. Grad Instruction Overhead Load per GF for Different Values of Average Hours per Graduate Student

Graph for Period GF Spons Res Pap Pub

600

450

300

150

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Period GF Spons Res Pap Pub - SEN3_01 papers/semesterPeriod GF Spons Res Pap Pub - SEN3_02 papers/semesterPeriod GF Spons Res Pap Pub - SEN3_03 papers/semesterPeriod GF Spons Res Pap Pub - SEN3_04 papers/semester

Figure 15.1. Period GF Sponsored Research Papers for Different Values of GF Research Papers Productivity

Graph for Period GF Unsponsored Res Pap Pub

600

450

300

150

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Period GF Unsponsored Res Pap Pub - SEN3_01 papers/semesterPeriod GF Unsponsored Res Pap Pub - SEN3_02 papers/semesterPeriod GF Unsponsored Res Pap Pub - SEN3_03 papers/semesterPeriod GF Unsponsored Res Pap Pub - SEN3_04 papers/semester

Figure 15.2. Period GF Unsponsored Research Papers for Different Values of GF Research Papers Productivity

Graph for GF Hist Avg Res Papers Pub per GF

4

3

2

1

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

GF Hist Avg Res Papers Pub per GF - SEN3_01GF Hist Avg Res Papers Pub per GF - SEN3_02GF Hist Avg Res Papers Pub per GF - SEN3_03GF Hist Avg Res Papers Pub per GF - SEN3_04

Figure 15.3. GF Historical Average Research Papers per GF for Different Values of GF Research Papers Productivity

FIGURES 14.1-14.3. Results of Sensitivity Test: Different Values of Average Hours per Graduate Student; (decreased to 3 in run 01, to 7 in run 02 and increased to 20 in run 04 from 15 in the base run and run 03.)

FIGURES 15.1-15.3. Results of Sensitivity Test: Different Values of GF Research Papers Productivity; (decreased to 1 in run 01, to 2 in run 02 and increased to 7 in run 04 from 4 in the base run and run 03.)

29 / 34

Number of Graduate and Undergraduate Students

1,4008,000

7004,000

00

0 2 4 6 8 10 12 14 16 18 20Time (semester)

Number of Grad Students - BASE_RUNNumber of UG Students - BASE_RUN

Figure 16.1. Dynamic Behavior of Number of Students in the Base Run

Number of Faculty

400

300

200

100

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Number of Faculty - BASE_RUN

Figure 16.2. Dynamic Behavior of Number of Faculty Members in the Base Run

Research Papers per Faculty Member

1.6

1.2

.8

.4

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Hist Avg Res Pap per Faculty - BASE_RUN

Figure 16.3. Dynamic Behavior of Number of Research Papers per Faculty Member in the Base Run

30 / 34

Number of Graduate and Undergraduate Students

8,0001,400

4,000700

00

1984 1986 1988 1990 1992Year

Number of Undergraduate Students - REALDATANumber of Graduate Students - REALDATA

Figure 17.1. Real System Behavior According to Data from Bogazici University

Number of Faculty Members

400

300

200

100

01984 1986 1988 1990 1992

Year

Number of Faculty Members - REALDATA

Figure 17.2. Real System Behavior According to Data from Bogazici University

Research Papers per Faculty Member

1.6

1.2

.8

.4

01984 1986 1988 1990 1992

Year

Research Papers per Faculty Member - REALDATA

Figure 17.3. Real System Behavior According to Data from Bogazici University

31 / 34

Graph for Number of UG Students

20,000

15,000

10,000

5,000

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Number of UG Students - EXP1_01 studentsNumber of UG Students - EXP1_02 students

Figure 18.1. Number of Undergraduate Students in Experiments with Graduate versus Under-graduate Study Orientation

Graph for Number of Grad Students

4,000

3,000

2,000

1,000

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Number of Grad Students - EXP1_01 studentsNumber of Grad Students - EXP1_02 students

Figure 18.2. Number of Graduate Students in Experiments with Graduate versus Under-graduate Study Orientation

Graph for UG Students Grad Students Ratio

20

15

10

5

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

UG Students Grad Students Ratio - EXP1_01UG Students Grad Students Ratio - EXP1_02

Figure 18.3. Undergraduate Students / Grad Students Ratio in Experiments with Graduate versus Under-graduate Study Orientation

Act Instr Load per UGF

1414

77

00

0 2 4 6 8 10 12 14 16 18 20Time

Act Instr Load per UGF - EXP_71_1Act Instr Load per UGF - EXP_71_2

Figure 18.4. Instruction Loads per Undergraduate Faculty in Experiments with Graduate versus Under-graduate Study Orientation

Act Instr Load per GF

1414

77

00

0 2 4 6 8 10 12 14 16 18 20Time

Act Instr Load per GF - EXP_71_1Act Instr Load per GF - EXP_71_2

Figure 18.5. Instruction Loads per Graduate Faculty in Experiments with Graduate versus Under-graduate Study Orientation

Graph for Period Res Pap Pub

600

450

300

150

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Period Res Pap Pub - EXP1_01Period Res Pap Pub - EXP1_02

Figure 18.6. Research Papers Published in Experiments with Graduate versus Under-graduate Study Orientation

FIGURES 18.1-18.6. Results of Experiments with Graduate versus Under-graduate Study Orientation (Higher Under-graduate and Lower Graduate Admission in run 1 versus Lower Under-graduate and Higher Graduate Admission in run 2)

32 / 34

Graph for Total UG Instr Hours Needed

10,000

7,500

5,000

2,500

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Total UG Instr Hours Needed - EXP2_01 hours/weekTotal UG Instr Hours Needed - EXP2_02 hours/weekTotal UG Instr Hours Needed - EXP2_03 hours/week

Figure 19.1. Total UG Instruction Hours Needed in Experiments with Different �Desired� and �Maximum� UG Class Sizes

Graph for Total FTE Needed for UG Instr

400

300

200

100

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Total FTE Needed for UG Instr - EXP2_01 FTETotal FTE Needed for UG Instr - EXP2_02 FTETotal FTE Needed for UG Instr - EXP2_03 FTE

Figure 19.2. Total FTE Needed for UG Instruction in Experiments with Different �Desired� and �Maximum� UG Class Sizes

Graph for Act UG Instr Load per GF

4

3

2

1

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Act UG Instr Load per GF - EXP2_01 hours/(week*faculty)Act UG Instr Load per GF - EXP2_02 hours/(week*faculty)Act UG Instr Load per GF - EXP2_03 hours/(week*faculty)

Figure 19.3. Instruction Loads per Graduate Faculty in Experiments with Different �Desired� and �Maximum� UG Class Sizes

Graph for Act Instr Load per UGF

20

16.5

13

9.5

60 2 4 6 8 10 12 14 16 18 20

Time (semester)

Act Instr Load per UGF - EXP2_01 hours/(week*faculty)Act Instr Load per UGF - EXP2_02 hours/(week*faculty)Act Instr Load per UGF - EXP2_03 hours/(week*faculty)

Figure 19.4. Instruction Loads per Undergraduate Faculty in Experiments with Different �Desired� and �Maximum� UG Class Sizes

Graph for Number of Faculty

400

350

300

250

2000 2 4 6 8 10 12 14 16 18 20

Time (semester)

Number of Faculty - EXP2_01 facultyNumber of Faculty - EXP2_02 facultyNumber of Faculty - EXP2_03 faculty

Figure 19.5. Number of Faculty in Experiments with Different �Desired� and �Maximum� UG Class Sizes

Graph for Cancelled UG Lecture Hours

400

300

200

100

00 2 4 6 8 10 12 14 16 18 20

Time (semester)

Cancelled UG Lecture Hours - EXP2_01 hours/weekCancelled UG Lecture Hours - EXP2_02 hours/weekCancelled UG Lecture Hours - EXP2_03 hours/week

Figure 19.6. Canceled UG Lecture Hours in Experiments with Different �Desired� and �Maximum� UG Class Sizes

FIGURES 19.1-19.6. Results of Experiments with Different �Desired� and �Maximum� UG Class Sizes (Desired=20, Maximum=40 in run 1; Desired=80, Maximum=160 in run 3 Desired=50, Maximum=100 in the base run and run 2)

33 / 34

Figure 20. Main Game Screen of the University Game

34 / 34

Player’s Decisions

707065

5

0000

0 2 4 6 8 10 12 14 16 18 20Time (semester)

GF Hiring Decision - RESORFACUGF Hiring Decision - RESORFACOfficial Projects Share for Faculty - RESORFACRelease Time for GF - RESORFAC

Figure 21.1. Decisions of the Research Oriented Faculty Member

Research Papers

1,0001,200

2

500600

1

000

0 2 4 6 8 10 12 14 16 18 20Time (semester)

Hist Avg Res Pap Pub - RESORFACPeriod Res Pap Pub - RESORFACPeriod Res Pap Pub per Faculty - RESORFAC

Figure 21.2. Figures about Research Papers in the game played by the Research Oriented Faculty Member

GF Project Loads and Total FTE for Official Projects

88

80

44

40

000

0 2 4 6 8 10 12 14 16 18 20Time (semester)

Act Official Projects Load per GF - RESORFACHours Spent for UP per GF - RESORFACTotal FTE for Official Projects - RESORFAC

Figure 21.3. Figures about Project Motivations and Loads of Graduate Faculty Members in the game played by the Research Oriented Faculty Member

Player’s Decisions

707065

5

0000

0 2 4 6 8 10 12 14 16 18 20Time (semester)

GF Hiring Decision - BALANFACUGF Hiring Decision - BALANFACOfficial Projects Share for Faculty - BALANFACRelease Time for GF - BALANFAC

Figure 22.1. Decisions of the Balanced Faculty Member

Research Papers

1,0001,200

2

500600

1

000

0 2 4 6 8 10 12 14 16 18 20Time (semester)

Hist Avg Res Pap Pub - BALANFACPeriod Res Pap Pub - BALANFACPeriod Res Pap Pub per Faculty - BALANFAC

Figure 22.2. Figures about Research Papers in the game played by the Balanced Faculty Member

GF Project Loads and Total FTE for Official Projects

88

80

44

40

000

0 2 4 6 8 10 12 14 16 18 20Time (semester)

Act Official Projects Load per GF - BALANFACHours Spent for UP per GF - BALANFACTotal FTE for Official Projects - BALANFAC

Figure 22.3. Comparison of the behaviors of total FTE faculty for official projects in the games played by the Balanced Faculty Member (BALANFAC) and the Research Oriented Faculty Member (RESORFAC)