Designing a decision support system for self-managing teams in education

(1)

Designing a decision support system for

self-managing teams in education

Master’s Thesis Supply Chain Management Technology & Operations Management

22 June 2020 Author:

M.P. Bekink (S2958465) - m.p.bekink@student.rug.nl Supervised by:

prof. dr. J. Riezebos - j.riezebos@rug.nl Second supervisor:

prof. dr. J. de Vries - jan.de.vries@rug.nl

Abstract

This paper elaborates on the conceptualization of multiple team membership. This is an alternative way of educational organization that improves flexibility, by forming teacher teams. The current concept brings additional complexities to decision making processes, for which support is desired. Via a design science research approach an initial decision support system is designed based on classical optimization. The system supports decision making in the individual teacher to class allocation as well as the educational contents determination, while considering the multiskilled teachers in a team.

Keywords: decision support system, self-managing teams, educational logistics, special needs

(2)

1. Introduction

A rather new development in operations research is the introduction of teacher teams in an educational setting. Working in teams increases flexibility among teachers (SIG, 2014; Van Dartel & Koppens, 2019) and makes it easier to support students’ dynamic needs through the diverse teacher qualities within the team (Hew & Brush, 2007). Team-based decision making, however, comes with additional complexities, involving the consideration of a diverse set of opinions and options to choose from. Along with student’s dynamic needs, making certain decisions may become a challenge.

This paper aims at finding ways how teacher teams can make these challenging decisions effectively and efficiently. Therefore, firstly will be defined what these teacher teams are, secondly what kind of decisions they have to make, and thirdly which operational concepts can facilitate in making these decisions. Current research reports that ‘educational teams’ are either poorly defined or very broad (Vangrieken et al., 2015). In this paper, the term 'team' should be seen as a dedicated group of teachers that together educate one or more student-classes. This is in line with current trends that show a movement from one fixed teacher for one fixed subject and class towards a more flexible organization of education in which more than one teacher can serve a class or subject (Van Dartel & Koppens, 2019).

A recent contribution to operations research is the conceptualization of Multiple Team Membership (MTM) in education, incorporating the formation of teacher teams. In this MTM-concept, as conceptualized by Jans (2020), the idea is that a teacher is member of more than one team. Each team has its own specialty, but the overall aim for each team is to ensure qualitative and meaningful learning for all students. Each team is responsible for a set of student classes. It is up to the teacher team how they educate these classes. This comes with certain decisions that have to be made.

(5)

flexible in their teacher utilization by implementing MTM. They also hope this enhances meaningful learning for the students, as this is an important factor for them. Increasing the quality of education while efficiently using teachers’ talents, may attract new teachers and decrease teacher turnover. The organization is reflecting on the MTM-concept and reveals the need for support in team-based decision making. The teacher teams have to decide among its members which individual is allocated to which student class. Also, they have to decide which educational contents should be provided to each class. This should be based on student needs in order to manifest meaningful learning. Finally, the organization has experience with unexpected teacher absence, which has serious consequences for their abilities to provide qualitative education to its students. They do not have capacity to decide on re-arrangements in such a short timeframe.

This paper investigates how these challenging decisions could effectively and efficiently be taken. It will, therefore, explore the operational possibilities, aiming to improve educational logistics. Such educational logistics can formally be defined as the combination of all processes, systems, and information flow that enable the educational organization to operate sleekly (SIG, 2014).

The main concerns for teacher teams in MTM are the allocation of individual teachers to student classes, and determination of educational contents. Teachers have multiple skills and can provide several subjects. Current literature conceptualizes staff allocation of such a multiskilled workforce as an optimization problem. Such traditional staff allocation problems are not yet suitable for daily decisions in education’s dynamic environment. However, redesigning such an optimization problem for the educational context could bring support in teams’ decision making. This paper will investigate how optimization can support teacher’s decision making in an MTM-setting, followed by a first design of such support. Hence, the following research question is constructed:

(6)

make decisions? What should be considered to ensure qualitative and meaningful education for their students? Are there existing concepts that could inspire the design of decision support? The theoretical exploration will be complemented with a problem investigation. This problem investigation is carried out at the organization in The Netherlands that is considering the implementation of MTM. This should reveal any additional needs of educators in light of the MTM-concept. The main contribution of this paper is the design of decision support for teacher teams in an MTM-based organization.

(7)

2. Methodology

This research aims at improving team decision power by elaborating on the idea of Multiple Team Membership (MTM) in education. Further support may be desired for educational teams in their daily decisions. Such support can be formulated into the design of a supportive tool that incorporates the distinct needs and functions necessary in the educational context. Because there is not yet such a readily applicable tool for the educational context, this paper will contribute to its development, which will be created via a Design Science Research (DSR) approach. This type of approach can be formally described as the design and investigation of an artifact in a certain context. This artifact is designed to interact with a problem context in order to improve something in that context (Wieringa, 2014). The artifact in this paper will be a supportive solution to be designed. The context is an educational organization considering the implementation of MTM.

(8)

Figure 1: The design science cycle (adjusted from Wieringa, 2014)

Additionally to Wieringa’s framework, Hevner et al. (2004) state that the design should be both relevant and rigorous. It should, therefore, incorporate current foundations in the existing knowledge base, while simultaneously be of practical use for a particular domain. The literature and foundations discussed in the theoretical background, function as inspiration and input for the design. The problem investigation and validation will ensure its practical relevance. These provide opportunities to discover current needs in the applicable domain, as well as it could help answer potential ‘How?’ and ‘Why?’ questions that may arise in the development of the design (Benbasat et al., 1987; Yin, 2014).

2.1. Collecting data

(9)

interpretations of observations and to evaluate others' views and ideas that are developed throughout the research. Documentation within the organization is analyzed in order to understand the interviews better. Examples are schedules and individual student plans. The third source of data is direct observations in the form of presence inside classrooms and at teacher meetings throughout the day. Questions are asked every now and then while observing to understand specific observations better. To further improve validity, observations are compared with another researcher's observations. Also, the interviews are recorded when possible, and notes are taken. In order to guide interviews along with the interest of the research, a simple protocol is created that contains relevant questions for the research. The protocol ensures that interviews have sufficient flexibility while providing a structure to touch upon all necessary aspects of the research. These protocol-guided semi-structured interviews should enhance reliability. A summary of the interviews and observations is provided in Appendix A.

2.2. Data use and continuation

(10)

3. Problem investigation

In this section, the problem investigation is covered as the first phase of the Design Science Research (DSR) cycle. First, the problem context is described. Thereafter the problem and the stakeholders are discussed more concretely. The section ends with questions that follow from the problem investigation.

3.1. Problem context description

(11)

(Dicker et al., 2019), which negatively influences employability for new teachers. Before a vicious circle starts to form in which the increased workload due to teacher leave becomes feedback for even more teacher leave and lower quality education, feasible and drastic solutions should be sought. RENN4 has been involved previously in the conceptualization of Multiple Team Membership (MTM) and is thinking of how they could use MTM in their organization. Currently, the MTM-concept is not complete enough to be implemented.

3.2. The problem analysis

Educating in teams rather than individually is proposed as a solution, aiming at the recent work of Jans (2020). The organization's management is open-minded but needs support as to how such a system could function in their organization once implemented. One aspect in their current organization is that some teachers stick to the belief of a high student-teacher dependency, making it contradictory for these teachers to start moving from single teacher to teacher team responsibility. The management itself does not necessarily stick to this dependency and would like to incorporate more flexibility in their organization. If this could be incorporated by using teams effectively, this would be of interest.

This would mean that a teacher is not providing education to one fixed class anymore. Rather, a team of teachers is responsible for a number of classes. To illustrate, a team of seven teachers has to serve four classes of students. Teaching still remains in the form of one teacher in front of one class. However, the teacher that teaches a class can frequently change, based on needs, insights, and availability. It could be that at the start of the day, one teacher is teaching a class, while a few hours later, another teacher from the team is teaching that class. In Jans (2020) concept, such a team, labeled as a level team, is responsible for four classes. These classes are bound to a certain level, and so is the teacher level team. The first level team would then serve educational level one, year one up to year four. In that way, the classes can get used to the level team throughout the full four years. The distinct teachers from that level team are providing several subjects to those four classes.

(12)

that this morning meeting remains in case that teams would be used. In that case, the teacher level team has to discuss the points of attention for their classes. Some data is already known, such as which classes they are responsible for, which timeslots for each class and which subjects each class is following. The decisions as to which individual teacher is providing which subject and in which relevant timeslot, remains for the teacher team to be made in the meeting. As it can be challenging in such little time to determine each class’s needs, checking the schedule and each other’s availability, support is desired that can help make the decisions.

Multiple stakeholders indicate that unexpected teacher absence is a problem. Since this absence cannot easily be processed by means of quick schedule alterations, student classes often are impacted by less qualitative education. The student class is then, for instance, distributed over other classes, causing an uneasy ambiance, or one or more school days are canceled in full for this student class. It would be very valuable if the support to be created can consider such unexpected absence. It may, therefore, be an idea to use the current teacher presence during the morning meeting as input for support in making decisions for individual teacher allocations. In that case, it could be avoided that classes are not served well due to unexpected absence of a solely responsible teacher. The overall idea is that the team of teachers take full responsibility for their classes and solve eventual problems among the team members. Together their goal is to ensure high-quality education for each of their classes, rather than only be responsible for a single class.

(13)

Some teachers emphasize the importance of the mentioned student-teacher dependency. From direct observations, a few teacher switches have been observed throughout the day for a class. One teacher switch was unexpected, as there was a teacher replacement due to absence. The class seemed to be able to cope with these switches quite well. This could mean that more flexibility, by moving away from full student-teacher dependency, may be possible. Imagined could be that this differs from time to time and between classes. A teacher that was observed mentioned that support, for example, a technical tool, in schedule alterations due to unexpected absence would be very useful and valuable. Another interviewee confirms this: "I wish we had such support. Currently, we do all by hand, which takes lots of time”. The problem can get very serious if there is a fever going around. If a teacher is sick for a longer period of time, it is possible that some students cannot be educated for a significant period. And this is not in line with the organization's ambition to provide high-quality education to its students.

Current student needs are also not known. At the time, subjects are scheduled according to a plan created in advance for a longer period of time. Sometimes a small alteration could be made based on a teacher's feeling, but this does not happen too often. Again, when teaching with teams, it can get even harder to determine student's current needs. Monitoring would be crucial here to generate data on which decisions should be based, responding to the current needs. Interviewees all indicate a desire to have such support. To be of value, however, support should come in easy and fast. The teachers only have ten minutes in the morning to make arrangements of the day, which is crucial to consider in designing a solution.

One interviewee, a teacher who is providing multiple subjects to a student class, indicates that as a teacher, one subject is better mastered than another, possibly influencing the quality of educating certain subjects. The same holds for the preference to teach a subject. Multiple interviewees are positive about the MTM-idea. Although they value the student-teacher relationship, the “students’ abilities should not be underestimated”. It is also indicated that a student class often has a different teacher each year, which means that there is a potential ability to get used to more than one teacher.

(14)

self-managing teams can allocate teachers to classes, (2) how self-self-managing teams can react to unexpected absenteeism, and (3) how monitoring can help make the right decisions. These questions have been discussed during the interviews, and a solution is required that provides such support in a short morning meeting of approximately ten minutes.

3.3. Stakeholders

The main stakeholders that are considered in the problem are summarized in Table 1. These are the stakeholders that are directly involved in the solution to be designed. The first stakeholder is the teacher team, as they need to make challenging decisions in a short amount of time. Their interest is in simple and effective support for making challenging decisions. Support should not add unnecessary complexity and should be easy to use. The second stakeholder in the problem is the student. The students have interest in structure and high-quality education. Cancellation of school days would be unfavorable. Getting education based on their needs and relation with a teacher would be valuable for them. If support for teachers could help in attaining this, their interest would be covered. Lastly, management is an overall stakeholder in the problem. Management wants to provide high-quality education to all students, while simultaneously creating a nice working atmosphere for their employees. By shifting the responsibility from individuals to teams, management hopes that teachers can attain more. Not only by feeling interconnected but also by sharing more knowledge and better being able to answer to the distinct needs that are present in a student class. Support for the teacher teams to answer these student's needs would, therefore, be in management's interest as well.

Table 1: Stakeholders

Stakeholder Description

Teacher Teachers want to provide the best education to the students possible Student Students want education based on their needs

(15)

3.4. Research questions

So far, the MTM-concept, the research design, and the problem investigation result in the research questions for this paper. The main research question is:

In which form brings classical optimization, value to the decision power of self-managing teacher teams for adequate determination of daily educational contents and individual teacher allocation? As interviewees indicated, it is important that teachers have autonomy; therefore 'self-managing' teams are key here. Also, support is mainly for teacher allocation and educational contents determination on a daily basis.

The following sections will further explore the design criteria for a solution. This starts with a theoretical background. This section elaborates on a few concepts that have been briefly discussed so far, as well as introducing new concepts that are valuable for a solution. The section functions as a conceptual overview that should be considered in the solution design. The theoretical background aims at findings answers to the sub-question:

(A) Which (support) concepts should be considered in designing a solution for automated decision support for teacher teams in SNE, and what are their implications?

After the theoretical background, the solution design section will present the design of the final solution. The design is based on input from the problem investigation and concepts discussed in the theoretical background. This section includes the support design and a first prototype that can be used for validation. The solution design should answer to the sub-question:

(B) Which functional, visual, and method design of automated support brings the most value to teacher teams' decision power for daily educational contents and individual teacher allocation?

(16)

educational practitioners from the problem investigation, who can reflect on a few critical aspects. This should result in the evaluation of the designed solution. The critical aspects are similar to the design evaluation factors of Hevner (2004). The critical aspects are of importance for the end-users and are derived from the problem investigation. Hence, the section aims at answering the sub-question:

(17)

4. Theoretical Background

4.1. Special needs education

The European Agency for Development in Special Needs Education (2003) defines Special Needs Education (SNE) as “education, based on a pedagogical project, that provides adapted schooling, care, and therapy for pupils whose general personal development cannot be

guaranteed, temporarily or permanently, in a mainstream school, or where such a guarantee is insufficient". The levels provided at secondary SNE are similar to that of a regular secondary school, but with increased attention for the distinct needs of students. The additional challenge in SNE compared to regular education is the larger variety of student needs and competencies. And since this study is focused on secondary SNE, this variety may be relatively wide, as the gap between students with special needs and peers without these needs increases with age (European Agency for Development in Special Needs Education, 2003). A teacher has to diversify the teaching skills among a variety of students, adapting to the individual student needs more than in regular education.

(18)

4.2. Agile and lean organization

The problem investigation reveals the need to be responsive to student needs. For such responsiveness, the concepts of lean and agility may be an inspiration, given the dynamic environment of education. Parsons & Maccallum (2019) confirm that it is increasingly important to be responsive to changes in requirements in any stage of development of students. From the management's perspective, it is essential to be able to answer to these students' needs with the current (scarce) capacity. Responsiveness (agility) and flexibility are key here. An interesting insight by Parsons & Maccallum (2019) is that much of the current literature on agile specifics in agile education comes from organizations and individuals with a practitioner focus, not from research. This means that contributions to this area of literature have practical relevance.

According to Peha (2011), agile thinking can be applied to education. It would be interesting to educate students based on their current individual performance, a lean concept labeled as ‘pull’ (Parsons & Maccallum, 2019). Here, the students pull towards themselves relevant materials based on current performance. With the right support, teachers could be facilitated to effectively do so. Another term within lean is, for instance, the term 'working software' as being the measure of progress. In education, such 'working software' should be altered in order to measure progress in education with the aim of 'meaningful learning' (Parsons & Maccallum, 2019). This makes individual and classical student monitoring of great importance. Students can then be educated on a subject at the right level and pace (Alp, 2001).

(19)

in which teachers can make notes, enter results and track performance, very useful for decision making. This paper assumes that such a system and its data is in place and could be used as input for a solution. The grading system in the Netherlands is on a scale from one (worst) to ten (best) performance. Such grades would be available as input from the student tracking system. The worse the grade, the more important to work on the related subject.

4.3. Teams

Educational organization in teams brings flexibility to the organization. This contributes to the trend of agile and lean education (Parsons & Maccallum, 2019). Teams are an essential part of this paper. The idea is to move away from the traditional way of teaching - with one individual teacher per class - towards the organization of a team that is responsible for, among others, that class. Not only can this provide more flexibility, but it also follows today's need to stimulate working in educational teams (SIG, 2014). Moreover, there are multiple reasons why one would do that. If the student population copes with a variety of relatively complex problems, which is especially the case in SNE, there is a need for intensified collaborations.

When there is no full distinction in concrete tasks, the integration of it is needed, and teams are unavoidable (Van Dartel & Koppens, 2019). It is not advisable to create full flexibility. A fixed team-to-student allocation is desired in order to minimize the transfer of students from one team to another. This enhances the creation of mutual knowledge and trust between the two (Van Dartel & Koppens, 2019). The incorporation of a possible student-teacher dependency is therewith embraced. The ideal scenario is that a team remains responsible for the same student throughout the full educational trajectory: first to last year.

(20)

the basis for students. Throughout their educational trajectory, students remain in this group (also called a mentor class). The elements being educated by the level team are centralized per class. For instance, the subject ‘History’ is educated in the mentor class by a teacher to all students. The teacher should, however, still differentiate among the different students. For more challenging and specific subjects, provided by the expert teams, students could be re-allocated in groups so that a student will be educated on a level that best fits the student’s current performance. Imagine, a student in a class educated by level team one (lower educational level) can be re-allocated to a group of students for the subject math (provided by an expert team) on the highest educational level (given that the student is good in math). For the subject physics (provided by an expert team), the student can again be re-allocated to another group of students on the most suitable level. Hence, level teams educate fixed classes based on class performance, whereas expert teams educate variable groups of students based on individual performance. The value of the level teams is that throughout the full educational trajectory students are guided by this team. This means that they get familiar with the team members, creating valuable relations. It is also the level team that stays in contacts with student’s parents or caretakers. Therefore, other than the provided subjects, social cohesion and cross-curricular student qualities are valuable level team contributions.

Teachers are to be allocated to both type of teams, which means the teacher is a member of two teams. This means that teachers find themselves in Multiple Team Membership (MTM) (Pluut et al., 2014). The first step to implement an MTM approach is allocating teachers to the teams. Another challenging step is creating a conflict-free schedule, as MTM has not been implemented in a timetabling problem before (e.g. Sabar et al., 2012; Sarin et al., 2010; Schimmelpfeng & Helber, 2007). Both steps have been covered in the conceptualization of Jans’ work (2020).

(21)

Imagine a level team of seven teachers, of whom four are present today in the morning meeting. They have to educate four classes in a number of specific timeslots today. Most of the present teachers were not working yesterday. In ten minutes, the four teachers have to decide who is providing education to which class(es), at what timeslot and which educational contents. What are the needs of each class? Which timeslots per class have to be filled by the team? How many hours per subject have been provided already this week? Whom of the team members can best provide which subject to which class? These questions take a considerable amount of time to answer and discuss. If a supportive solution could accelerate that process, decisions could be made in a timely and qualitative manner.

4.4. Team formation and allocation:

Jans (2020) model incorporates the teachers’ competencies and preferences in order to form and schedule the distinct teams. Such competencies and preferences could also be an essential factor in the allocation of individual teachers to a student class. Bahargam et al. (2017) state that the grouping and matching should facilitate peer interaction, meaning that a level class and a teacher should be matched on certain criteria. One usable criterion in Jans' paper (2020) is the teaching competence score per teacher. That score can be obtained using a self-assessment questionnaire for teachers if such score is desired in the teacher allocation. One can, for instance, allocate a teacher with a higher teaching competence score to a class that is performing badly, as the class has more need for it. Another criterion that Jans (2020) used is a teacher's knowledge level in a certain expertise. Such a criterion should also be used in individual teacher allocation. A score per subject as to how well the subject is mastered by a teacher could be used to allocate a teacher to teach a certain subject.

(22)

numerical notation used ranges from zero (not skilled or preferred at all) to one (highly skilled or preferred).

Table 2: Skills matrix

Teacher 1 Teacher n

Skill Preference Skill Preference

Task 1 0.80 1.00 0.6 0.75

Task n 0.55 0.6 0.95 1.00

4.5. Multiskilling and staff allocation

The main problem that teacher teams face is the actual allocation of a teacher that has the highest potential of increasing the students’ progress for a certain subject. Which teacher is most appropriate for a task is dependent on a variety of inputs, that can be derived from, for example, the skills matrix. In literature, there is abundant research that describes staff allocation problems (E.g. Bassett, 2000; Fernandez-Viagas & Framinan, 2014; Wu & Sun, 2006). In such problems, often in project (production) management, workers are allocated to certain tasks based on their skill or degree. The main inputs for this problem are a worker’s availability, the worker’s skill, and the tasks that have to be performed. Workers are then allocated by satisfying constraints or parameters that state that a worker is 1) available at timeslot X, and 2) that a worker is qualified for task Y. But what if the workforce consists of employees that each has a variety of skills at certain levels, like in the MTM-concept?

(23)

example of allocating the workers to certain production tasks, the schedules are for the longer term. Not to forget that the result of these optimization problems can be performed in a limited context. Combining the idea, however, with day-to-day management in which the autonomy remains at the workers, may be of value for teacher teams. Such an ILP model may be constructed to make it fit the educational context and to make it suitable for daily decision making. Also, where the existing problems optimize the schedule for fast and efficient processing times, the focus within education would lay in the highest (best matching) quality of a teacher to a class, while simultaneously determining the educational contents.

4.6. Decision support systems

(24)

educational context is very dynamic, it would be of much value that in the DSS classes and subjects can be selected that are taken care of by the team currently.

4.7. Design requirements

(25)

5. Solution design

The system to be designed is a Decision Support System (DSS) for teacher teams in the morning meeting. Day-to-day decisions have to be made in a short amount of time, and thus the system should not add any unnecessary complexities. This section starts by presenting the functional design of the solution, along with (IDEF0) interaction diagrams. This will then be used for the mathematical formulation of these functions. The mathematical formulation is based on staff allocation optimization, reformulated to suit the teacher teams’ decisions. The mathematical formulation is used in the design of the DSS. A prototype of the DSS is designed and presented. This section ends with the main optimistic scenario, illustrating how the DSS would function, what it does, and what it requires the users to do.

5.1. Functional design of the system

In this section, functional diagrams of the system to be designed are presented. The distinct functions of the system are mapped in hierarchical and interaction diagrams to show how they function. The system as a function, which is indicated as A0, as well as its sub-functions, use inputs and controls from its contexts in order to function properly. The outputs of the system are being fed back to its context. The direct context of the system is the educational organization. A detailed and functional representation of the system’s context can be found in Appendix B.

Figure 2: Functional architecture of system (A0)

A.0 P ide be den need ed ca i n F nc i n A.1 M ni ing F nc i n A.2 Ta k de e mina i n F nc i n A.3 Teache all ca i n F nc i n A.4 Ed ca ing den F nc i n hie P ide be den need ed ca i n

(26)

The aim of the system is to facilitate the provision of best student needs education. That means that based on student progress, teacher skills and competencies, teacher preferences, and teacher availability, the right decisions should be made that answer to the current needs of students. In Figure 2, the system is presented, which is decomposed of four sub-functions: 1) monitoring, 2) task determination, 3) teacher allocation, and 4) educating students. The order of the function is from A1 to A4. The core functions of the system to be developed are A2 and A3 since these functions take away most of the manual and time-consuming work.

Figure 3: Interaction diagram of system (A0)

The four functions are further mapped in the interaction diagram in Figure 3. The monitoring function ensures that there is data on current student (class) performance. Inputs for the monitoring function are the teacher teams and the student level classes. The function is controlled by the teacher team to student class allocation, as this determines which team is responsible for which

(27)

(28)

5.2. Functional design of system’s sub-functions

In this section, the system is further decomposed by its sub-functions. As task determination (A2) and teacher allocation (A3) are considered to be the system’s core function, adding most value, these will be further mapped hereafter. The sub-functions monitoring (A1) and educating students (A4) are provided in Appendix C. Starting with the second sub-function, task determination (A2), there are three sub-functions. These transform grades and scores into needs, alter these needs based on set out preferences, and prioritize the needs based on the curriculum. A2’s hierarchy is shown in Figure 4.

Figure 4: Hierarchical diagram A2

The interaction of A2’s sub-functions in Figure 5, illustrates in what steps the concrete tasks are determined. The student grades and points of attention are being fetched from the student tracking system that is in place. These can be considered outputs from the monitoring function (A1). These are inputs to A2’s first sub-function. Here the raw needs are determined; low scores indicate high needs. The raw needs are passed on to the second sub-function (A22) in which the preferences that the DSS’s user indicates on the dashboard are considered. These preference factors are for the user to steer the needs based on own insights as well as setting priorities. The raw needs are changed accordingly, and these are input for the final sub-function (A23) along with the taught hours per subject. This final sub-function guards the curriculum. The problem investigation revealed that it is important to follow a curriculum that sets out how many hours of each subject have to be provided each week. This function makes sure that needs are prioritized if the according subject

A.2 Ta k de e mina i n F nc i n A.21 T an f m g ade need F nc i n A.22 Al e need i h indica ed efe ence F nc i n A.23 P i i i e need ba ed n c ic l m F nc i n hie Ta k de e mina i n

(29)

has not been provided this week yet. All sub-functions are triggered by the subjects provided by the team: if a subject is not provided, then it is not considered as a need. The results of the final sub-function are the most important tasks per class.

Figure 5: Interaction diagram A2

The third sub-function of the system (A3) is the actual allocation of level team members to the determined tasks and classes. It consists of matching teachers to the tasks and creating an optimal schedule for the day (Figure 6).

Figure 6: Hierarchical diagram A3

C ec ed eed ba ed e i da a M i a a e c a P i f a e i Ra eed ba ed c e S de g ade S bjec ided b ea Ta gh h e bjec A.21 T a f g ade eed A.22 A e eed i h i dica ed efe e ce A.23 P i i i e eed ba ed c ic idef0 Ta de e i a i U i e i Edi i - F Acade ic U e O Da e: Ma 26, 2020 A.3

Teacher alloca ion F nc ion A.31 Ma ch eacher o a k F nc ion A.32 Crea e op imal da ched le for eam F nc ion hier Teacher alloca ion

(30)

Figure 7: Interaction diagram A3

The interaction diagram in Figure 7 illustrates that the matching uses teacher competencies and the most important tasks per class as input. The matching is dependent upon the preferences of the teachers, as these can influence the matching. Once the best-teacher-to-tasks-match options are generated, these options are fed to the second sub-function (A32). Here the best options that are possible, based on the overall schedule and the individual teacher availability, are used to generate an optimal schedule for the day, along with the distinct subjects for today and the teachers, per class. The DSS should then present the selected optimal schedule for the day, sorted per teacher and class, clearly showing what is provided to which class, by whom, when, and why. If the users are satisfied with the solution, the schedule can be followed. If not, some parameters can be changed, and a new solution can be generated.

(31)

5.3. Mathematical formulation of the system

5.3.1. Input parameters, variables and objective function

For the Decision Support System (DSS) to function properly, it is important that the practical problem, as investigated previously, is translated accurately into a mathematical formulation. This mathematical formulation follows the process and concept as being illustrated in the previous section with the functional diagrams. The mathematical formulation functions as the algorithm in the background of the decision support system and will cover mainly sub-function A2, task determination, and A3, teacher allocation, from the main system A0 as conceptualized. As indicated in the functional, a teacher team can use the system by the use of a dashboard. Within the dashboard, the user can change certain preference factors and update some parameters if necessary. By changing some of the preference values, the user can influence the outcome of the results, which is important to keep autonomy at the user level. The result of the system is a day-schedule in which all timeslots necessary are filled, including the allocation of teachers and subjects for each class.

(32)

Jans (2020). The final set is the schooldays (D), which contains the days the school is providing education to the students. This is Monday up to Friday.

Table 3: Sets

Set Description Values No.

𝐶 Classes See Appendix D (1)

𝑇 Timeslots per day See Appendix D (2)

𝐵 Subjects See Appendix D (3)

𝐼 Individual teachers See Appendix D (4)

𝐷 Schooldays See Appendix D (5)

Using these sets, the parameters that make up the actual mathematical formulation are presented in Table 4. The parameter D contains the value of a teacher’s expertise in each subject. The better the teacher is in this subject, the higher the value. How these values are assigned is up to the school, but from the different discussions with some of the teachers and management, it seems that teachers (and/or the team of teachers) are able to do a self-assessment as to how competent they are for a certain subject (relatively to their colleagues). The same principle holds for the second parameter P, which is the preference of a teacher to teach a certain subject. The teachers should be able to change these two parameters from time to time.

Table 4: Input parameters for the objective function

Notation Description Value No.

𝐷_!" Knowledge level individual i in subject b 𝐷_!" ∈ {0.00; … ; 1.00} (1) 𝑃_!" Preference of individual i to teach subject b 𝑃_!" ∈ {0.00; … ; 1.00} (2) 𝐹_! Teaching competence of individual i 𝐹_! ∈ {0.00; … ; 1.00} (3) 𝑅_# Preference factor teacher dependency class c 𝑅_# ∈ {0.00; … ; 1.00} (4) 𝑅!# Familiarity of teacher i with class c 𝑅!# ∈ {0.00; … ; 1.00} (5) 𝑆# Preference factor social aspect priority class c 𝑆# ∈ {0.00; … ; 1.00} (6)

𝐽_" Subject b is a social subject 𝐽_" ∈ [0,1] (7)

𝐻𝑇_"# Scheduled hours per week subject b for class c 𝐻𝑇_"# ∈ {0,1,2, … , 𝑛} (8) 𝐻"# Hours subject b already provided to class c this week 𝐻"# ∈ {0,1,2, … , 𝑛} (9) 𝐺_#" Grade of class c for subject b 𝐻_"# ∈ {1.00; … ; 10.00} (10)

(33)

(34)

The grades could be based only on student exam performance, but ideally would be a combination of exam performance and teacher assessment each time the subject is provided. This would result in a more accurate and updated representation, as exams are only organized a limited amount of times.

Table 5: Decision variables

Notation Description Value

𝑋_#$"!% _{Class c on time t has subject b provided by individual i on day d} _𝑋_#$"!% _{∈ [0,1]}

In Table 5, the decision variables for the objective function are provided, which is limited to one. The binary decision variable, X, represents the teacher and subject allocation for each class. If the value holds one, it means that class ‘c’ on time ‘t’ will be provided with subject ‘b’ by teacher ‘i’ on day ‘d’. The variable should be positive for a combination of allocation if this combination would yield the highest benefit for all students combined. In order to find this combination, the following objective function is constructed:

𝑀𝑎𝑥: 2 32 2 2 2 𝑋!"#$% %∈' $∈( #∈) "∈* ∙ (𝐷$#∙ 𝑃$#∙ 𝐹$#) ∙ (1 + 𝑅$!∙ 𝑅!) ∙ =1.1 −_𝐻𝑇𝐻#! #!A ∙ (1 + 𝑆!∙ 𝐽#) ∙ (11 − 𝐺!#)E !∈+

The objective function consists of the decision variable and five following terms and can be considered as function A32, ‘create optimal day-schedule for teams’, from the diagrams. The idea is that learning and future student performance is maximized. Two indices in the objective function are not influencing the allocation result directly: 't' and 'd'. In the objective function, they act as the ability to sum all possible allocation scores, while in the constraints these indices are used to limit the 't' and 'd' values allowed according to schedule and user input. This means, for instance, that if a 't' or a 'd' value is not allowed, the associated allocation score equals zero. The function starts with the first term, which is the combination of the teacher’s knowledge level in a subject, the teacher’s preference to provide that subject, and the teacher’s overall teaching competence. This term has a minimum value of zero and a maximum value of one. The higher this value, the higher the potential that this teacher can improve the students’ performance. The term can be considered

(35)

(36)

5.3.2. Constraints

The model is subject to certain boundaries, constraints. For these constraints, there are some additional parameters, summarized in Table 6.

Table 6: Input parameters for the constraints

Notation Description Value No.

𝐴_!$ Availability individual i in timeslot t today 𝐴_!$ ∈ [0,1] (11) 𝐿𝑇_# Level team is responsible for class c 𝐿𝑇_# ∈ [0,1] (12) 𝐿𝑇_"# Level team provides subject b to class c 𝐿𝑇_"# ∈ [0,1] (13) 𝐿𝑇_#$% Class c is allocated to level team in time t on day d 𝐿𝑇_#$% ∈ [0,1] (14) 𝑀_"# Maximum timeslots subject b for class c per day 𝑀_"# ∈ {0,1,2, … , 𝑛} (15)

𝑇_% Today is day d (on the schedule) 𝑇_% ∈ [0,1] (16)

(37)

general schedule for the week. Hereafter follow the constraints the model is subject to, using the parameters as described above. First, the constraint is explained, where after the mathematical formulation follows.

Constraint 1 – Only selected classes and subjects

To start with the first constraint: it is important for a level team that only the classes and subjects that it teaches are allocated in the daily allocation. The model has incorporated all subjects and classes that are present in the school. The level team selects the classes they are currently responsible for and select the subjects that are being provided by the team, per class. The model then only considers the selected classes and subjects and will look to the data that is related to these. Logically follows that for each class, for each timeslot, for each subject and for each day, the allocation value 𝑋_#$"!% should be equal or smaller than the level class value 𝐿𝑇_# multiplied by the subject to class value 𝐿𝑇_"#. If the class and/or the subject for this class is not selected, then the allocation can never be positive.

1 𝑋_#$"!% ≤ 𝐿𝑇_#∙ 𝐿𝑇_"# ! ∈ (

∀ 𝑐 ∈ 𝐶, 𝑡 ∈ 𝑇, 𝑏 ∈ 𝐵, 𝑑 ∈ 𝐷 (1)

Constraint 2 – Only adjacent timeslots

(38)

constraint regards a single subject and holds for every class apart. The sum of the allocation of all timeslots should not exceed the value two. Alternatively, a non-quadratic constraint could be used that ensures that a maximum of one or two succeeding timeslots can be used for one subject. This, however, limits the ability to be able to provide more than two timeslots for a subject on a day. The added flexibility to be able to use as many succeeding timeslots as desired (e.g. four timeslots) for a subject, is chosen to be worth the additional complexity as long as the model returns a solution in an acceptable amount of time. In the case of upcoming exams, for instance, it is then possible to spend a full day on one subject.

1 1 1 <𝑋_#$"!% − 𝑋_#($*+)"!%> -% ∈ . ! ∈ ( $ ∈ / ≤ 2 _{∀ 𝑐 ∈ 𝐶, 𝑏 ∈ 𝐵} ₍₂₎

Constraint 3 – Only available teachers and only allocated once

A very simple, yet important constraint, is that only the teachers that are available in a certain timeslot are allocated to provide a subject to a class. For every day, for every timeslot, and for every individual teacher, the sum of the allocation values 𝑋_#$"!% of every class and subject may not exceed the availability value of the teacher 𝐴_!$. The constraint makes sure that a teacher can only be allocated if he or she is present and can only be allocated once to a timeslot.

1 1 𝑋#$"!% # ∈ 0

" ∈ 1

≤ 𝐴!$ ∀ 𝑡 ∈ 𝑇, 𝑖 ∈ 𝐼, 𝑑 ∈ 𝐷 ₍₃₎

Constraint 4 – All scheduled hours should have an allocation

It is important that the classes have a teacher allocated to them in every scheduled hour. This makes that this constraint is a hard constrained and must be obtained. The number of allocations for a day 𝑋_#$"!% should be equal to the total number of timeslots that are scheduled for the level team for this class. This constraint is for each distinct class and day.

1 1 1 𝑋_#$"!% ! ∈ ( " ∈ 1 $ ∈ / = 1 𝐿𝑇_#$% ∙ 𝐿𝑇_# ∙ 𝑇_% $ ∈ / ∀ 𝑐 ∈ 𝐶, 𝑑 ∈ 𝐷 (4)

Constraint 5 – Only allocations in scheduled level hours

(39)

beyond the level team’s responsibility. The fifth constraint, therefore, states that the allocation value 𝑋#$"!% should be smaller or equal to today’s value 𝑇% multiplied by the timeslot-is-level-team-for-class value 𝐿𝑇_#$%. If the day is not today, there should be no allocations positive, as only the timeslots of today should be considered. Additionally, only if a timeslot for a class is scheduled to be a level team timeslot, an allocation could be made. This constraint holds for every day, every timeslot and every class.

1 1 𝑋_#$"!% ≤ 𝐿𝑇_#$%∙ 𝑇_% ! ∈ (

" ∈ 1

∀ 𝑐 ∈ 𝐶, 𝑡 ∈ 𝑇, 𝑑 ∈ 𝐷 (5)

Constraint 6 – Maximum number of timeslots per subject

As indicated in the second constraint and as obtained from the case (RENN4, personal interview, March 20, 2020), teachers may choose to provide a subject for more than one timeslot on a day. As this preference may change from day to day and may be different for every class, a preference parameter is included in the model. The parameter could be changed any time and is a positive integer. It indicates the maximum number of timeslots that one single subject may be provided to a distinct class on one day. For instance, if this is set to two, a maximum of two timeslots can be used for a subject. Whereas the second constraint only ensures that multiple timeslots of a subject are adjacent, this sixth constraint would make it such the number of timeslots on a day do not exceed the entered number. If for instance ‘two’ is entered, there may be provided two timeslots on a day to a class of a certain subject, which have to be adjacent. If not, only one timeslot may be used, ensured by the second constraint. The constraint takes the allocation values 𝑋_#$"!% for the full day per day, per class and per subject.

1 1 𝑋#$"!% ≤ 𝑀"# ! ∈ (

$ ∈ /

∀ 𝑐 ∈ 𝐶, 𝑏 ∈ 𝐵, 𝑑 ∈ 𝐷 (6)

Constraint 7 – Only one subject and teacher per timeslot

(40)

1 1 1 𝑋_#$"!% % ∈ . " ∈ 1 ! ∈ ( ≤ 1 _{∀ 𝑐 ∈ 𝐶, 𝑡 ∈ 𝑇} ₍₇₎

5.3.3. Prototype of the Decision Support System

Now that the mathematical formulation of the model is finished, a prototype of the model into a Decision Support System (DSS) should be performed. In order to build a prototype, a prescriptive analytics application has been used, named AIMMS Prescriptive Analytics Platform. It is a well-known technological tool in operations research and could be used to build and deliver solutions to improve business performance (AIMMS, n.d.). There are two versions of the tool: AIMMS Developer and AIMMS PRO. This paper has access to AIMMS Developer and therefore uses this version to create a prototype of the model in question. The PRO version could be used to deliver the solution to the end-user, which could be deployed in the cloud, ensuring accessibility and sufficient computational power.

The sets, parameters, variables, objective function, and constraints have been entered into AIMMS Developer. The resulting declarations are presented in Figure 8, which is the model tree that contains a declaration section. A full representation of the model can be found in Appendix F, which contains a textual representation of the AIMMS project file.

(41)

Running the model does not result in any errors, indicating that the model is functional. Using a simple page designer in AIMMS, a prototype dashboard is created. On this dashboard, parameters are presented, of which some should be changed daily. The dashboard is presented in Figure 9.

Figure 9: Prototype dashboard

(42)

can be entered in the according tables and cells for each class. The fifth column presents all subjects again and, in this table, there can be indicated which subjects are considered to be a social subject. If the social aspect has a high priority today, the selected subjects are prioritized in the allocation solution. The sixth and final column represents the preferential factor as to how many timeslots can be assigned to one subject on a day per class. In the table, a value between one and nine (as positive integers) can be entered for each subject and each class. In the lower-left bottom of Figure 9, there is a button labeled ‘Generate Today’s Allocation!’. After entering all relevant data on the dashboard, the button can be clicked in order to start solving a solution. In order to generate a solution, a use case study has been performed. The used parameter data for this use case can be found in Appendix E. This includes the optimal concept schedule that resulted from Jans’ (2020) work. A solution with the entered data is presented in Figure 10.

Figure 10: Optimized day allocation solution for teachers

(43)

average grade of the class for the subjects to be provided and the familiarity of the teacher with that class. The lower the class’ grade, the higher the chance the subject is listed in the solution. And when dependency for a class is important, a teacher with higher familiarity will be prioritized. For instance, in the solution, Teacher 1 is allocated to Class 1 in Timeslot 1. The familiarity of Teacher 1 with Class 1 is only ‘0.10’. If the preference factor for dependency for Class 1 was set to ‘1.00’, there is a low chance that Teacher 1 was allocated to Class 1. Additionally, there is also an overview of the schedule per class that the level team has to carry out. This is presented in Figure 11. This functions as an additional overview. The full dashboard, in which each part is presented as a whole, is shown in Figure 12. The prototype system consists in total of 32,292 constraints and 136,081 variables. The total amount of time between clicking the button and having a solution equaled ‘10.30’ seconds. A time short enough to be able to make decisions in a short morning meeting of approximately ten minutes. Changing parameters and running the model multiple times consistently resulted in valid solutions. This indicates that the prototype is fully functional and can be used in the educational organization's context. In order to reflect critically on the model, a validation phase follows in the next section.

(44)

(45)

5.4. Main optimistic scenario

The main optimistic scenario contains the ideal scenario of a teacher team using the Decision Support System (DSS). The main optimistic scenario is visually presented in the BPMN diagram in Figure 13. The scenario starts at the morning meeting in which the team has to determine today’s educational contents and teachers per class. The team discusses the priorities per class for today and sets the parameters on the dashboard accordingly.

Usually, a teacher team has four fixed classes all the time. In case this is changed by management for any reason someday, the team can select the new classes that are applicable to them on the dashboard. If not, the classes that are selected (e.g. one up to four) should not be changed. The same holds for the subjects provided by the team per class. This is set by management and should not change regularly. Both the selected classes and subjects per class should only be changed if management has changed these. This means that other than the first time the DSS is used, these parameters are not touched often by the team.

What should be changed daily by the team is the presence of each team member per timeslot for the day. The schedule, as well as unexpected absence, result in a different teacher presence every day. Other parameters that can be changed daily are the preference factors per class: dependency and social. The team can set the dependency preference for a class to high if they think that the class has a high need for the most familiar teacher. Likewise, the team can set the social preference factor to high if they think that the class has a need for subjects with a social character. The subjects that are perceived to be of social character can be selected by the team.

The set parameters and preferences are stored in the DSS’s database. In the database are also the skills matrix and teacher-class familiarity stored, but these are not shown on the dashboard. The teacher team should be able to update these from time to time, but this is not likely to be necessary on a daily basis. After filling out the parameters, the user can request a solution.

(46)

that is assumed to be in place. The DSS requires teachers to update the student tracking system for each class, each timeslot and each subject.

(47)

This should result in the most adequate solution per day. From the tracking system data, the DSS can determine the current needs for each class. The DSS computes all possible scenarios of educational contents and teacher allocations. Each scenario is assigned a score, which is the sum of how well the needs of each class are answered in the scenario. The DSS selects the optimal scenario, which is the one with the highest score.

The optimal solution, containing the educational contents per class per timeslot and the teacher who provides each subject, is then presented. In the solution, also the average class grade for a subject and the familiarity of the allocated teacher with the class is shown. These function as clarification why a subject and a teacher are selected. The solution is sorted on the teacher, so that each teacher can see his/her schedule for the day. Additionally, all subjects and allocated teachers are presented per class in order to provide additional overview.

The teacher team can now assess the solution to see if it is to their liking. If it is not to their liking, they can make some minor alterations manually and follow the altered solution. If it is not to their liking at all, or complex changes are desired, the user should head back to changing the parameters and priorities to steer a new solution. Once an acceptable solution has been found, the team can follow the solution and be assured that they provide qualitative and meaningful education to their students for the day.

Throughout the day, it is important that the teachers report progress each timeslot in the student tracking system. This includes notation which subject has been provided in each timeslot, how this subject performed throughout the hour (score), additional grades of, e.g. exams (score), and additional notes that are found to be important for colleagues. The entered data is used the next day in the morning meeting by the DSS.

(48)

6. Solution validation

In this section, the designed Decision Support System (DSS) will be evaluated. Referring to Wieringa’s (2014) Design Science Research (DSR) cycles, validation is the third (and final) phase of this paper. Validation can be considered as a form of evaluation. A design is effective and complete as soon as it satisfies the initial requirements of the problem it was meant to solve (Hevner et al., 2004).

There is a variety of evaluation methods available in literature. In this paper, the DSS should be evaluated, which can be categorized as an algorithm. In a literature review by Peffers et al. (2012) for DSR evaluation methods, regarding algorithms, 60 out of the 65 papers used a technical experiment as its evaluation method. Peffers et al. (2012) define such technical experiment as “a performance evaluation of an algorithm implementation using real-world data, synthetic data, or no data, designed to evaluate the technical performance, rather than its performance in relation to the real world”. In this paper, a basic technical experiment was performed by randomized data and experiments to verify that the DSS returned valid solutions. As this was the case, the DSS has a first positive evaluation.

(49)

Table 6: Evaluation criteria

Critical aspect to be evaluated Hevner et al.’s related evaluation criteria (2004)

Usability Usability

Speed of decisions Performance, fit with organization

Solution adequacy Accuracy

Flexibility & Autonomy Completeness, fit with organization

User-friendliness Functionality

Confidence & acceptance Reliability, consistency

Table 7: Validation questions

Critical aspect Questions

Usability How useful is the DSS? Would it enhance effectivity and efficiency?

Speed of decisions Would use of the DSS result in a satisfactory speed of decision making?

Solution adequacy Are the solutions returned by the DSS valid enough and adequate? Can the solution be used/followed?

Flexibility & Autonomy Does the DSS incorporate a sufficient level of flexibility? Can the parameters and solution be changed to an acceptable extent? Can the solution be influenced to an acceptable extent?

User-friendliness Is the DSS’s workflow easy to follow? Is the visual dashboard organized in a logical way?

(50)

6.1. Validation results

Usability

Overall, the DSS could be very useable for RENN4. Most valuable so far is the fast generation of a day-schedule based on current capacity. In the prototype, as a concept, there are no names included yet. RENN4 would find it of high value if the DSS could be changed to reflect team-member names, improving clarity in allocation. One concern is their current capacity. In the design's concept, a level team may have seven teachers while serving four classes. In reality, it could be that there are only three teachers present while serving four classes. This may cause problems. RENN4 can imagine that the DSS, with regards to generalizability, may be easier to implement in regular education. Overall, they see potential in the DSS.

Speed of decisions

In the demonstration, the solution was generated within ten seconds. If only minor changes have to be made each day, one is familiar with the DSS, and the solution is generated within ten seconds, the speed is satisfactory. Decisions could then be made in a short amount of time. Especially in case of unexpected absence, the speed is beyond expectations. One thing of improvement could be the simplicity of the DSS; this should be less complex and overwhelming. This could further enhance speedy decisions.

Solution adequacy

The solution is useful and easy to follow. As in prior interviews, RENN4 emphasizes that some classes may have a significant student-teacher dependency. If this can be reflected in the solution, by filling out per class how important the dependency is, the solution’s adequacy would increase. This would as well enhance the acceptance level for the solution.

Flexibility & autonomy

(51)

user-friendliness. It may be better to have them in the background, accessing them only when necessary. With regards to autonomy, RENN4 does not encounter problems so far. Input from the student tracking system is used, the DSS’s solutions seem valid, and a solution can be influenced in multiple ways. Currently, teachers are used to getting a schedule from management, which tells them what to do in a certain period. The level of autonomy does not change if such schedule would instead be provided by the DSS, especially if there are ways in which the solution can be influenced.

User-friendliness

The dashboard is overwhelming. One generic and a simpler design is desired in a later stadium. It is indicated that it would be too much of a hassle to use the DSS as a team. As a workflow, they would prefer it if one or two people are appointed to use the DSS and make the decisions. This may even be someone from higher up, such as a scheduler. This avoids discussion within the team. Also, they would prefer to set preferences and priorities per class only periodically, e.g. once every month. This saves abundant switching between priorities and takes away the need to fill out redundant parameters every morning. Visualization could be added by adding colors to indicate certain points of attention. RENN4 indicates that it would be very useful if schedules could be created for the longer term as well, while allowing to generate a new day-schedule if there are unexpected changes such as absence. Currently, the solution is presented only in the morning, telling the teachers what to do. The teachers then have no time to prepare for the subjects they are providing today. This makes the solution, as it is designed right now, harder to follow. This has a negative impact on the over-all user-friendliness.

Confidence & acceptance

(52)

6.2. Validation conclusions

(53)

7. Critical success factors

This section discusses the Critical Success Factors (CSFs) that should be kept in mind when considering the use of the designed solution. “CSFs are those characteristics, conditions, or variables that when properly sustained, maintained, or managed can have a significant impact on the success of a firm competing in a particular industry” (Leidecker & Bruno, 1984). CSFs can be specific to either the firm, a certain department, regulations, or environment. This paper makes a distinction between design specific CSFs and context-specific CSFs. In order for an organization to gain optimally from utilization of the design, the CSFs should be satisfied. Throughout the research, several CSFs have been identified, which are presented below.

7.1. Design specific success factors

Human – machine interaction

As indicated by Fransoo et al. (2011), there a several automation levels that a Decision Support System (DSS) can have. In the current design, the DSS has level (4) automation, which “suggests one alternative”, based on user input. That means mutual control between the human and the algorithm. From validation, it was found that this was indeed the desired level of human-machine interaction. The user can influence the outcome, and the complex mathematical work is being carried out quickly by the DSS. It is important to integrate this balance in order for users to generate valuable and custom solutions. A design specific success factor is, therefore, that there is mutual control between the user and the DSS to generate a solution.

Availability of data

(54)

Accuracy of data

As the DSS bases its solutions on the available data, the quality of the solution is dependent on the quality of this data. This means that the more accurate the data, the higher the solution quality. Accuracy of data means that the data is valid and objective, frequently reported, and that the available data to the DSS is up to date. If this is not the case, the quality of the solution may not fit the current needs or may be less accurate. A design specific success factor is, therefore, that data is generated objectively and frequently while being made directly available to the DSS.

7.2. Context-specific success factors

Dynamic

The DSS is designed for the educational context, which is a dynamic environment. This means that the DSS would bring value to organizations that mostly face constant change of needs or capacity. If needs and capacity would be constant, the DSS would rather add complexity than reducing it. A context-specific success factor is, therefore, that the context is complex and faces a constant change of needs and capacity.

Multiskilled workforce

The DSS assumes workers (teachers) to have multiple skills, which are represented in the skills matrix of the DSS. This means that workers can be interchanged if capacity changes and that workers can satisfy a variety of needs. If the workforce is not multiskilled and task or skill specialization is in place, flexibility would decrease to a minimum, and much value of the MTM-concept that the DSS incorporates would disappear. A context-specific success factor is, therefore, that the context consists of workers that contain a variety of skills.

Relational nature

Designing a decision support system for self-managing teams in education