Bridging the Computer Science Teacher Shortage with a Digital

Learning Platform

L. M. van der Lubbe

, S. P van Borkulo

, P. B. J. Boon, W. P. G. van Velthoven and J. T. Jeuring

Freudenthal Institute, Utrecht University, P.O. Box 85170, 3508 AD, Utrecht, The Netherlands

Keywords:

Computer Science, Secundary Education, Co-Teach Informatica, Learning Platform, Student Model.

Abstract:

Although computer science is important in the many aspects of the world around us, computer science edu-

cation is insufﬁcient, partly due to a shortage of qualiﬁed computer science teachers. Co-Teach Informatica

offers a temporary solution to high schools in the Netherlands to overcome this shortage. With online learning

materials, local remote support desks, and guest teachers, it offers a computer science course following the

curriculum guidelines. This paper presents the design of an online learning platform speciﬁcally designed for

the Co-Teach Informatica program. The design has two important pillars: independent learning and progress

tracing. Both are important to ensure that students can follow their computer science course without a quali-

ﬁed computer science teacher in their classroom. Finally, we discuss the design of the student progress tracing

component using a student modelling approach and the requirements of the different user groups.

1 INTRODUCTION

Computer science is an important subject because of

the inﬂuence of the discipline on the world around us.

The number of jobs in software development keeps on

growing. For example, the demand for software de-

velopers in the manufacturing sector is now exceed-

ing the demand for production workers (Code.org,

2017a). Research from Code.org found that “Com-

puting occupations are the largest category of new

wages in the United States – ahead of management,

healthcare, ﬁnance, engineering, sales, or any other

category” (Code.org, 2016). Although the skills in-

volved in computer science are important to prepare

for future working and everyday life, computer sci-

ence education is insufﬁcient.

In many countries, there is a teacher shortage for

computer science, both in higher and secondary ed-

ucation (Shein, 2019). Moreover, in some cases, a

non-expert teacher (for example a teacher certiﬁed in

mathematics) is teaching computer science in a sec-

ondary school (Goode, 2007). There are different

causes of this shortage, among which is the difference

in salary for teaching compared to industry jobs for

computer scientists (Shein, 2019). Moreover, the ma-

jority of computer science students do not consider

https://orcid.org/0000-0003-1678-5159

https://orcid.org/0000-0001-6668-5282

https://orcid.org/0000-0001-5645-7681

teaching as their prior career choice when graduat-

ing (Yeni et al., 2020), as found in a research among

university students from the US. Computer science is

not the only subject with a teacher shortage, globally,

other science subjects, such as mathematics, suffer

from the same problem (McVey and Trinidad, 2019).

However, the number of qualiﬁed computer science

teachers who graduate is much lower compared to

mathematics and science teachers (Code.org, 2017b).

Different strategies can be applied to attract more

teachers, among which are ﬁnancial incentives such

as higher salaries and grants, and more opportunities

for teacher licensing for example with an emergency

license until the full license is achieved (McVey and

Trinidad, 2019).

Because schools often only offer a few com-

puter science courses, there is generally only a single

teacher. Thus, computer science teachers do not be-

long to a larger section within their school. This can

create practical challenges and makes collaborations

between teachers difﬁcult (Goode, 2007).

Computer science is an elective course in most

countries, and many schools are not offering it (Yeni

et al., 2020). This means that not all students

will be able to attend a computer science course at

their school or university (Shein, 2019). The con-

tent of computer science curricula in high schools

greatly varies per country and can also differ between

schools. It varies from a focus on basic digital skills

van der Lubbe, L., van Borkulo, S., Boon, P., van Velthoven, W. and Jeuring, J.

Bridging the Computer Science Teacher Shortage with a Digital Learning Platform.

DOI: 10.5220/0011971900003470

In Proceedings of the 15th International Conference on Computer Supported Education (CSEDU 2023) - Volume 1, pages 289-296

ISBN: 978-989-758-641-5; ISSN: 2184-5026

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

289

with a vocational emphasis to learning foundational

computational thinking skills and design principles in

a problem-solving context (Goode, 2007). Therefore,

students can have different skills at the end of their

high school education, preparing them for future jobs

or studies to different extents. In the Netherlands,

Computer Science is also an elective course and few

schools are offering it. There is a national curriculum

for the upper levels of high schools, however schools

still have freedom of how to translate this to their

lessons. Thus, although some general baseline knowl-

edge is required for all students, there are still differ-

ences in the knowledge and competencies of the stu-

dents completing a high school course on Computer

Science in the Netherlands.

Reducing the teacher shortage takes time, and un-

til the situation has improved temporary solutions are

needed. To bring computer science to high schools

without a qualiﬁed computer science teacher, Co-

Teach Informatica

offers a remotely organized com-

puter science program following the curriculum for

Dutch schools. Partly, their program is delivered

via online materials. This paper presents the design

of the online learning platform designed speciﬁcally

for the Co-Teach Informatica program in the Nether-

lands. The design has two important pillars: inde-

pendent learning and progress tracing. A teacher at

a school participating in the Co-Teach Informatica

program has no or limited content knowledge, but

rather coaches students and manages the practicalities

of the course. Therefore, the platform should facili-

tate the learning of students in the best way possible.

Moreover, a teacher cannot support students in their

progress due to this lack of content knowledge. Trac-

ing the learning progress of the students with the use

of learning goals can help the teachers, but also the

students themselves to gain more insights into their

knowledge level and potential knowledge gaps.

To ensure that the design of the platform is suit-

able for the target group and the program, it is de-

signed in an iterative process and with close collab-

oration with different stakeholders. This paper de-

scribes the design of the platform. The remainder of

this paper will be as follows: The background section

describes the Dutch computer science curriculum, the

organization of the Co-Teach Informatica program,

and methods to trace students’ progress. In the sec-

tion ‘Platform Design’, the design of the platform is

described in more detail. In the ‘Future Work’ sec-

tion, we brieﬂy look forward to the studies that will

be performed to improve and evaluate the platform.

Finally, in the ‘Conclusions’ section, the importance

of the research is stressed with a prospect of results

https://www.co-teach.nl/

that can be expected from future studies.

2 RELATED WORKS

2.1 Dutch Computer Science

Curriculum

In the Netherlands, computer science is offered as an

elective course in the higher grades of general sec-

ondary education (in Dutch: havo) and pre-university

education (in Dutch: vwo) (Grgurina et al., 2018).

While regular subjects in the curriculum end with a

national exam in the ﬁnal grade, computer science

does not have such an exam. Instead, schools can cre-

ate their own exams within the given guidelines. The

content of the curriculum for secondary computer sci-

ence education was renewed in 2019 to be more up-

to-date. The curriculum consists of mandatory core

domains and elective domains. There are three pub-

lishers for learning material for computer science in

the Netherlands: Instruct, Informatica Actief and VO-

Content. Often teachers mix materials from different

publishers and other sources, including their own ma-

terials.

2.2 Co-Teach Informatica

Many vacancies for high school computer science

teachers in the Netherlands are unfulﬁlled and this

number will only increase in the coming years.

Schools are therefore forced to stop offering com-

puter science when their teacher leaves or are unable

to start teaching it (Lucassen, 2022). Co-Teach Infor-

matica aims to help those schools with a temporary

solution following the ofﬁcial curriculum. For the

mandatory domains, Co-Teach Informatica designed

online learning materials that students can use inde-

pendently while a school teacher is present to manage

and motivate the class and do the administration for

the course. This school teacher is not qualiﬁed for

computer science and does not need to have computer

science knowledge. To support students on the learn-

ing content, remote local support desks of teaching as-

sistants and subject didactics experts are organized by

Co-Teach Informatica. Multiple remote support desks

are each dedicated to schools in a speciﬁc region,

hence they are called local remote support desks. The

teaching assistants also correct and grade the work of

the students. For the elective domains, guest teachers

who are IT professionals are invited to give a project-

based module in a class. Those guest teachers receive

didactical training from Co-Teach Informatica. An

CSEDU 2023 - 15th International Conference on Computer Supported Education

290

overview of the organization of the Co-Teach Infor-

matica program can be found in Figure 1.

Figure 1: Overview of the organization of the Co-Teach In-

formatica program.

This paper presents the design of a learning plat-

form developed for the Co-Teach Informatica pro-

gram. The focus of this platform is to facilitate stu-

dents, teachers and local remote support desks in their

learning process and work. The platform has two

important pillars: independent learning and progress

tracing. These two pillars are particularly important

because the teacher at the participating school is not

a computer science content expert and therefore has

limited content knowledge. Therefore, it is important

that the independent learning of the students is sup-

ported in the best way possible. That means for ex-

ample that communicating with the local remote sup-

port desk should be easy and accessible. Although

the teacher present in the classroom can motivate and

guide students, the teacher might not be able to over-

see their learning progress as the teacher is not famil-

iar with the content. Tracing the students’ progress

by modelling their knowledge based on learning goals

can help to give students, teachers, and teaching assis-

tants insights into the progress of students and classes.

2.3 Tracing Student Progress

To follow the knowledge state of a student and to pre-

dict future performance, knowledge or skill modelling

can be used (Pel

anek, 2017). This is the most used

form of what is called student modelling. Besides

a student’s knowledge and skills, other aspects can

be incorporated into a student model, such as student

motivation, preference or affect. While those aspects

are dynamic, static aspects like gender or age can also

be used (Chrysaﬁadi and Virvou, 2013).

Student models are useful for teachers or educa-

tional content developers to gain insights in for exam-

ple item difﬁculty or learner clusters (Pel

anek, 2017).

Learner clustering puts students into a cluster of sim-

ilar students. The system that uses this clustering can

adapt to those clusters, instead of treating all students

as if they were coming from a homogenous group.

For example, if student clusters are created around

learning difﬁculties, each cluster can receive different

forms of information.

Making a student model available to students of-

fers them a tool for reﬂection and self-monitoring,

supports discussions between students and teachers

and supports self-regulated learning (Pel

anek, 2017).

In such cases, the model is often visualized, for ex-

ample in a (directed) graph. In adaptive educational

systems, student models are a way to provide person-

alization for individual students (Chrysaﬁadi and Vir-

vou, 2013). Based on the information in the student

model about the needs, preferences, and knowledge

state, the system can adapt the learning path.

A student model can cover different aspects: the

process of learning and forgetting, using observa-

tional data, domain modelling, and learner clustering

and individualization (Pel

anek, 2017). There are dif-

ferent approaches to modelling learning and forget-

ting. For this type of modelling, there is a distinc-

tion between models that are based on assumptions

about learning, such as the Bayesian knowledge trac-

ing that includes forgetting, item difﬁculty and indi-

vidualization, and models without these assumptions

about learning that instead use approaches such as

(exponential) moving averages. For both types of

models simple and more complex approaches exist.

Observational data is often the correctness of an item,

but can also be the response time, use of hints, wrong

answers or the history of attempts. To model the do-

main, there are different ways in which the knowl-

edge components (items such as learning material or

exercises) can be organized. Knowledge components

can be organized in disjoint sets, meaning that each

item belongs to one learning component. However, it

is also possible that multiple knowledge components

apply to one item. Knowledge components can also

be arranged in a hierarchy or a structure with prereq-

uisites. Basic student models assume that all learners

come from one homogeneous population. However,

Bridging the Computer Science Teacher Shortage with a Digital Learning Platform

291

in reality, learners are often diverse which can be ex-

plored with learner clustering.

Intelligent tutoring systems are digital tools that

can be used as a (partial) alternative to one-to-one hu-

man tutoring and can be used for computer science

education. For example, intelligent tutoring systems

give feedback on the syntax or semantics of code writ-

ten in special environments (Crow et al., 2018). In

general, intelligent tutoring systems have two impor-

tant roles for which student modelling is used: a diag-

nostic role and a strategic role (Chrysaﬁadi and Vir-

vou, 2013). The diagnostic role means that the sys-

tem understands the knowledge of the student. The

strategic role means that the system can plan a re-

sponse to that knowledge state. In their review of

intelligent tutoring systems for programming educa-

tion, Crow et al. (2018) found that such systems often

lack reference materials for the students and mainly

focus on programming tasks and the required instruc-

tions. Reference material is often closely related to

the domain model, which is an important component

of a student model. User interactions with reference

material can normally be used as input for the stu-

dent model. Although intelligent tutoring systems for

programming often lack reference material, it is still

possible to use student modelling, for example for

generating hints and feedback. However, the underly-

ing knowledge domain model might not be visible to

the student because it cannot be linked to explaining

resources and therefore it might be more difﬁcult to

understand the hints and feedback. When reference

material is available, students can seek clariﬁcation

based on the feedback of the student model (Crow

et al., 2018).

3 PLATFORM DESIGN

This section describes the design and the rationale

behind the design of the learning platform. Impor-

tant prerequisites for the newly designed platform are

allowing independent learning and tracing the stu-

dents’ progress, while supporting the speciﬁc work-

ﬂow of Co-Teach Informatica. The emphasis on

student progress tracing, using a student model ap-

proach, makes this platform unique compared to other

existing platforms. In the next paragraph, the ratio-

nale behind this is explained in more detail. In the fol-

lowing paragraphs, the main features to facilitate in-

dependent learning within the Co-Teach Informatica

program are explained per user type, i.e., the learners,

the support desk, and the teachers.

3.1 Student Progress Tracing Using

Student Modelling

To trace the progress of the student, a student model

approach is used. This student model represents the

knowledge state of the student and is based on a do-

main model consisting of all learning goals of the

learning materials. This domain model is organized

on different levels. Learning goals are connected with

other learning goals to represent prior knowledge. For

each mandatory domain of the curriculum, the learn-

ing goals are organized by chapter. This helps to

reduce the complexity of the graph and allows easy

and comprehensible ﬁltering. Connections with prior

knowledge can go beyond the learning goals of the

chapter or even the domain.

All learning activities in the online learning mate-

rials are linked to learning goals. For each learning

goal, a so-called probability score can be calculated

based on the student’s achievement on activities re-

lated to the given learning goal. The probability score

expresses the probability that a learner will answer

the next question linked to the learning goal correctly.

Every time a student completes an activity, the prob-

ability score for the linked learning goals (and their

prior knowledge) changes according to the evaluation

of the activity. The exact mathematics behind this will

evolve during initial testing. Figure 2 shows an exam-

ple of what a learning goal graph for a programming

chapter could look like.

As mentioned in the background section, student

models have two roles in intelligent tutoring systems:

a diagnostic role and a strategic role. In our platform,

the student model also has these two roles.

When our student model is used as a diagnosti-

cian it is mainly a tool for reﬂection, both for the

learner and the teacher or the local remote support

desk. Aside from being a tool that can be used to

give them insight into their progress, it can also stim-

ulate self-regulated learning. Within the learning ma-

terial, there are different types of activities: exercises

for practising and so-called milestone assignments.

Exercises for practising are not mandatory and are

often automatically corrected or manually assessed

by the learner. Milestone assignments are obligatory

and larger activities, often corrected by teaching as-

sistants. Learners are free to decide whether they

want to follow the learning material linearly and com-

plete (all) the exercises for practising before starting

with the milestone assignment or cherry-picking the-

ory and exercises to prepare for the milestone assign-

ment. The information on the learning goals can help

them to make informed decisions about this.

CSEDU 2023 - 15th International Conference on Computer Supported Education

292

Figure 2: Example learning goal graph for a programming chapter.

The use of our student model can also be extended

towards having a strategic role. Instead of only giving

the learners the freedom to customize their learning

path by for example skipping exercises for practis-

ing, the system could also actively make recommen-

dations to the learners. With these recommendations,

the system can help learners to increase their prob-

ability score (and thus their knowledge) on weaker

learner goals. For example, if a learner has a probabil-

ity score lower than 50% for multiple learning goals

for the next milestone assignment, the system can rec-

ommend exercises that cover those learning goals or

prior knowledge if that is still lacking. This can help

learners to prepare for milestone assignments or ex-

ams. This role has the potential to be extended to

other types of recommendations, such as recommend-

ing supporting peers that can complement each other.

The prerequisite for the ﬁrst application of the stu-

dent model is that all the learning material can be rep-

resented in the domain model. This means that all

learning activities are linked to nodes in the domain

model graph. Tweaking the mathematics behind this

model, such as weight factors representing the impact

of an item evaluation on the probability score, can be

done based on experiences in the classroom. For the

second way of applying the student model, the pre-

requisites are more complicated. To be able to rec-

ommend suitable exercises for learners, there needs

to be a signiﬁcant number of exercises. There have

to be multiple ways in which a learning goal can be

practised, and there have to be exercises using differ-

ent combinations of learning goals to be able to ﬁll

knowledge gaps. The current Co-Teach Informatica

program material does not include such a variety of

exercises yet.

3.2 Learners’ Requirements

Learners use the system to access their learning mate-

rials, including theory, exercises for practising, mile-

stone assignments and exams. Figure 3 shows the stu-

dent view of learning material on the platform. Ex-

ercises and milestone assignments can be completed

within the platform. Short answers can be given

through text boxes, different closed answer types are

facilitated and Jupyter Notebook is integrated to al-

low students to work on Python code. For some ex-

ercises, like writing a report, or when speciﬁc soft-

ware is needed, exercises are solved outside of the

platform and can be submitted to the platform man-

ually. Exercises are assessed in three different ways:

automatically (for example for multiple choice ques-

tions), manually by the learners (with the help of ex-

ample answers), or by the local remote support desk

(in the case of milestone assignments). In the last

case, learners can read the feedback on their exercises

in the platform and hand in improvements if neces-

sary. Exams can be taken and graded via the platform

as well.

Aside from accessing the learning materials,

learners can follow their progress in different ways.

All exercises for practising are marked with check-

boxes that indicate whether the exercise is completed

(see Figure 3). This gives learners a clear overview of

their progress within a chapter but does not say any-

thing about whether the exercise was correct or not.

Because the exercises are meant for practicing and are

not graded, the score is less important, and thus the

focus is on completing them. For milestone assign-

ments, which are graded, the checkbox shows more

information. A score (0-100%) is shown together

with a green or red checkmark indicating whether the

Bridging the Computer Science Teacher Shortage with a Digital Learning Platform

293

Figure 3: Screenshot of the platform from the student’s view.

score is sufﬁcient to pass or whether another attempt

is needed.

In addition to this, learners can follow their

progress via a learning goal graph, as shown in Figure

2. For each chapter, the part of the learning goal graph

with the relevant learning goals, coloured according

to their probability scores for a student, is shown in-

teractively so that students can explore it. For every

learning goal, they can read a description, ﬁnd an ex-

ample and link to the relevant reference material.

Lastly, the platform allows learners to communi-

cate with the local remote support desk, the teacher,

the class group, or other individual learners. An inte-

grated chat-like message service is offered to facilitate

these different ways of communication.

3.3 Local Remote Support Desks’

Requirements

The local remote support desk uses the platform to

prepare learning materials and make them available

for different classes.

As explained earlier, some of the learners’ work

is corrected by the teaching assistants from the local

remote support desk. The work of students can be

downloaded or viewed on the platform and a rubric

can be ﬁlled in to evaluate the work. This rubric is

also linked to learning goals, to strengthen or weaken

the probability scores of those learning goals (see Fig-

ure 4).

The local remote support desk can communicate

with individual learners, class groups, teachers and

each other via the integrated chat service.

3.4 Teachers’ Requirements

Teachers have the same user type (technically) as the

local remote support desk. However, teachers will

not use the possibility to change learning material and

grade assignments. Teachers need to edit the planning

as this is class speciﬁc and can follow the progress of

their students. Moreover, they can communicate with

their class, individual learners or the local remote sup-

port desk.

4 CONCLUSIONS AND FUTURE

WORK

The platform that is introduced in this paper will ulti-

mately be used within the Co-Teach Informatica pro-

gram. To develop the platform, intermediate eval-

uations will ensure an informed design. Therefore,

the stakeholders (students, teachers, and local remote

support desks) are involved in the design process. Af-

ter initial brainstorming to get insights into the user

requirements, the prototype of the system was built.

This prototype will be evaluated in a pilot study, in

which multiple classes will be using the platform for

a short course on programming. The research with

CSEDU 2023 - 15th International Conference on Computer Supported Education

294

Figure 4: Screenshot of a rubric to evaluate the work of students.

this platform will give insights into how independent

learning within a classroom setting can be facilitated.

Requirements for the platform, and prerequisites that

need to be met to fulﬁl these requirements, will be-

come clear.

Since this is still a limited prototype, the stu-

dent model will be limited to having a diagnostic

role. Future work will focus on incorporating more

learning material to extend the roles of the student

model. Moreover, the exact embedding and visualiza-

tion of the student model is an important focus point

of the ﬁrst pilot study. Aside from this, certain design

choices, such as the way that completion of exercises

for practising and milestone assignments are shown

to the learners are important questions for this pilot

study. Finally, the pilot study aims to improve the

user experience for all involved stakeholders to best

facilitate their workﬂow.

Aside from the user experience with the platform,

the goal is also to integrate a student model. Our re-

search will give insights into how a student model

can be applied to the Dutch Computer Science cur-

riculum. Moreover, a suitable way of visualizing this

student model and the way that it is embedded in the

material is also explored in this research. Those ﬁnd-

ings apply to other courses as well.

For the prototype design, the scope of the plat-

form with respect to the learning content is still lim-

ited. The learning materials of one mandatory domain

(programming) are included, which also means that

the student model is only developed for that domain.

In the future, the learning materials will include all

mandatory domains. Each domain within the Dutch

computer science curriculum will have its demands

for the platform. While programming is a rather prac-

tical domain, other domains can be more theoretical,

which may pose other challenges. For example, do-

main model graphs representing the student models

might become less connected. We also expect that

extending the learning material in the platform cre-

ates the need to change the initial design or extend it

with new requirements. Thus the platform needs to

be versatile, which makes it also broader applicable,

possibly for other subjects as well.

The platform is designed to be used within the Co-

Teach Informatica program, but our platform could

also be used by qualiﬁed computer science teachers

that want to follow their students’ progress using a

student model based on learning goals. And in addi-

tion, we see potential for it to be used in other contexts

as well. More secondary school subjects are suffering

from teacher shortages. It might be interesting to ap-

ply a similar approach using the infrastructure of our

platform for those subjects to help overcome the prob-

lem of teacher shortage more broadly.

Bridging the Computer Science Teacher Shortage with a Digital Learning Platform

295

ACKNOWLEDGEMENTS

The research reported here was part of a larger

project, (partly) ﬁnanced by EFRO and REACT EU,

project number KVW-368.

REFERENCES

Chrysaﬁadi, K. and Virvou, M. (2013). Student modeling

approaches: A literature review for the last decade.

Expert Systems with Applications, 40(11):4715–4729.

Code.org (2016). Computing occupations are now the #1

source of new wages in america.

Code.org (2017a). In manufacturing, software jobs outpace

production jobs for ﬁrst time.

Code.org (2017b). Universities aren’t preparing enough

computer science teachers.

Crow, T., Luxton-Reilly, A., and Wuensche, B. (2018). In-

telligent tutoring systems for programming education:

A systematic review. In Proceedings of the 20th Aus-

tralasian Computing Education Conference, ACE ’18,

page 53–62, New York, NY, USA. Association for

Computing Machinery.

Goode, J. (2007). If you build teachers, will students come?

the role of teachers in broadening computer science

learning for urban youth. Journal of Educational

Computing Research, 36(1):65–88.

Grgurina, N., Tolboom, J., and Barendsen, E. (2018). The

second decade of informatics in dutch secondary ed-

ucation. In Pozdniakov, S. N. and Dagien

e, V., ed-

itors, Informatics in Schools. Fundamentals of Com-

puter Science and Software Engineering, pages 271–

282, Cham. Springer International Publishing.

Lucassen, M. (2022). Scholen schrappen vak informatica

— de algemene onderwijsbond.

McVey, K. P. and Trinidad, J. (2019). Nuance in the noise:

The complex reality of teacher shortages. Bellwether

Education Partners.

Pel

anek, R. (2017). Bayesian knowledge tracing, logistic

models, and beyond: an overview of learner modeling

techniques. User Modeling and User-Adapted Inter-

action, 27(3):313–350.

Shein, E. (2019). The cs teacher shortage. Communications

of the ACM, 62(10):17–18.

Yeni, S., Aivaloglou, E., and Hermans, F. (2020). To be

or not to be a teacher? exploring cs students’ percep-

tions of a teaching career. In Proceedings of the 20th

Koli Calling International Conference on Computing

Education Research, Koli Calling ’20, New York, NY,

USA. Association for Computing Machinery.

CSEDU 2023 - 15th International Conference on Computer Supported Education

296