Formation of Study Groups: Exploring Students’ Needs and Practical

Challenges

Cosima Schenk

1 a

and Sven Strickroth

2 b

Institute of Psychology, Goethe University Frankfurt, Frankfurt, Germany

LMU Munich, Munich, Germany

ﬁ

Keywords:

CSCL, Collaborative Learning, Student Learning, Group Composition, Self-Organised Collaborative Learning.

Abstract:

Learning in study groups offers students the opportunity to exchange ideas about lecture content, discuss

questions, and network with others. However, little is known about how self-organised study groups (i. e., study

groups that are organised and managed by students themselves) should ideally be composed to meet students’

needs. Following previous studies on group composition in collaborative learning, a requirement analysis was

carried out, consisting of a focus group and an online survey. Three factors were identiﬁed as being particularly

important to students: A similar level of conscientiousness, a similar attitude towards reliable attendance at

meetings, and a similar preference for online meetings. Based on these results, a tool was implemented that

uses a genetic algorithm for group formation. This prototype was tested and evaluated in a ﬁeld study in two

university courses. The ﬁeld study suggests that there is a general interest in using such a tool. However, it

seems to be a challenge for many students to establish contact and meet with the other members of the proposed

study group. Possible reasons and solutions to this problem are discussed.

1 MOTIVATION

Working and interacting together in groups is associ-

ated with many beneﬁts, ranging from academic ad-

vantages to positive social and psychological effects

(Laal and Ghodsi, 2012). For this reason, collaborative

learning methods, such as group work, are often used

in universities and colleges (Davidson et al., 2014).

In addition to the mandatory group work, which is

an integral part of many courses, students also have

the opportunity to join self-organised study groups.

Participation in such study groups is not obligatory

and the concrete design of the learning process is left

to the students themselves. Compared to mandatory

group work, there has been relatively little research on

these self-organised study groups. However, a better

understanding of them could help to support students

in their learning and promote networking among stu-

dents. This is particularly relevant in the context of

mass education, i. e., in courses that are attended by

several hundred students.

These types of courses require high self-regulatory

skills as the exchange between the lecturer and the

https://orcid.org/0009-0000-6137-8670

https://orcid.org/0000-0002-9647-300X

students as well as the exchange among the students is

often limited (Strickroth and Bry, 2022). Thus, these

courses in particular do not initially offer the opportu-

nity for in-depth exchanges with others and it is there-

fore up to the students themselves to ﬁnd other fellow

students for the purpose of learning together. One of

the main issues in this regard, is to ﬁnd the “right”

persons to study and possibly start a study group with.

It cannot be assumed that the students know all their

fellow students well enough to know which persons

they study well with or not. Hence, the use of technol-

ogy for the group formation seems essential to help

individuals ﬁnd the best study group for them.

This paper aims to contribute to a better under-

standing of study groups in order to provide initial

indications of how a tool for forming study groups

should be designed. The aspect of the group compo-

sition of study groups is examined in particular. The

appropriate composition of study groups is very impor-

tant for students in general (Rybczynski and Schussler,

2011). However, no systematic investigation has yet

been carried out with regard to the factors that are par-

ticularly relevant from the perspective of the students

themselves. Considering the students’ perspective is

important though, in order to gain valuable insights

into their experiences and to better identify how to

Schenk, C. and Strickroth, S.

Formation of Study Groups: Exploring Students’ Needs and Practical Challenges.

DOI: 10.5220/0012545000003693

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 16th International Conference on Computer Supported Education (CSEDU 2024) - Volume 2, pages 647-658

ISBN: 978-989-758-697-2; ISSN: 2184-5026

647

support their learning (Cook-Sather, 2002). Presum-

ably, the students’ perspective is particularly relevant

regarding the context of study groups: Firstly, because

it is a learning experience that is not externally inﬂu-

enced (e. g., by teachers or instructors), but only by the

students themselves. Therefore, students themselves

can probably best assess the factors that enhance or

hinder this experience. Secondly, because the partici-

pation in these self-organised study groups is voluntary.

Dissatisﬁed students, whose expectations are unmet,

can easily leave their study group. To prevent this,

it is crucial to ensure that students’ expectations are

fulﬁlled.

Therefore, the present work pursued two goals:

The ﬁrst step was to carry out a systematic require-

ment analysis with students in order to identify factors

which are according to students most important for the

formation of self-organised study groups. On this ba-

sis, a prototype was developed for a tool that students

can use to ﬁnd a study group. In a second step, the

prototype was tested in the ﬁeld in two introductory

Bachelor’s courses (computer science and math) to

determine how well the study groups formed by the

prototype worked and what difﬁculties and challenges

arose.

In summary, this paper addresses the following two

research questions:

•

What are relevant factors for the composition of

study groups from the students’ point of view?

•

What possible challenges need to be considered

when using a tool to form study groups?

This paper is structured as follows: Section 2 pro-

vides an overview of the previous research literature on

the composition of student groups. Section 3 describes

how the ﬁrst research question was addressed, present-

ing methods and results. Based on these ﬁndings, a

prototype for a tool for the formation of study groups

is introduced in Section 4. Section 5 presents the ﬁeld

evaluation of the tool. The results of the studies are

discussed in Section 6. Finally, the paper closes with

conclusions and an outlook.

2 RELATED RESEARCH

Students working together in groups is generally re-

ferred to as the concept of collaborative or cooperative

learning: Both of these terms describe a pedagogical

approach which is characterized by two or more peo-

ple interacting with each other to achieve a common

goal (Dillenbourg, 1999; Gillies, 2016). It should be

noted that there has been debate about the exact dis-

tinction between these two related concepts (Jacobs,

2015; Panitz, 1999). For the present work, we follow

the distinction made by Panitz who refers to collabora-

tive learning as a broader concept or philosophy that

requires group members to build a consensus by inter-

acting with each other and is more student-centered

than cooperative learning which is a form of cooper-

ation that is closely controlled by a teacher (Panitz,

1999). According to this view, study groups can be

seen as a form of collaborative learning. This is consis-

tent with previous research literature which also refers

to learning in study groups as self-organised collab-

orative learning, a form of learning that takes place

without external guidance but is instead mainly driven

by the initiative of the learners (Melzner et al., 2020).

When collaborative learning methods are combined

with learning technologies, e. g., to connect students

or facilitate resource sharing, it is also referred to as

computer-supported collaborative learning, abbrevi-

ated CSCL (Dillenbourg and Fischer, 2007).

Although study groups should be distinguished

from mandatory group work, it is still useful to exam-

ine studies regarding mandatory group work as well

when examining possible factors for group formation.

They provide a comprehensive overview of various

factors that have already been used in this context for

the composition of student groups and could also be

relevant for study groups. For this reason, results re-

lating to the context of mandatory group work will be

discussed ﬁrst.

Several studies have already addressed the issue

of appropriate group composition, partly in connec-

tion with possible algorithms that can be used for

group formation (Cruz and Isotani, 2014; Odo et al.,

2019; Konert et al., 2014). Konert et al. distin-

guish between person-related and group-related fac-

tors: While person-related factors describe the individ-

uals of the group (e. g., their personality traits, their

level of knowledge or sociodemographic variables),

group-related factors refer to the group as a whole

(Konert et al., 2014). A group-related factor that has

often been mentioned in the research literature is the

group size which is assumed to inﬂuence group pro-

ductivity (Shaw, 2013) and students’ motivation (Zhan

et al., 2022). Recommendations regarding group size

suggest small groups, as the productivity of the indi-

vidual person decreases in larger groups (Odo et al.,

2019). With regard to person-related factors, there

is the additional question of whether the group as

a whole should be homogeneous or heterogeneous

concerning these factors. Previous literature contains

arguments in favour of both homogeneous and hetero-

geneous groups: While it has been argued that group

heterogeneity may promote improved peer interactions

(Magnisalis et al., 2011) and overall creativity (Nijstad

CSEDU 2024 - 16th International Conference on Computer Supported Education

648

and De Dreu, 2002), it has also been indicated that

trust may be higher in homogeneous groups and that

this may positively inﬂuence performance outcomes

(Ennen et al., 2015). It should, however, be noted

that homogeneous group formation may also lead to

unfairness when groups are very different from each

other which may lead to some persons beneﬁting more

from group work than others. Considering this is par-

ticularly relevant if the grouping criterion is related to

the level of knowledge or performance as it has been

shown that low ability students beneﬁt more in hetero-

geneous groups than in homogeneous groups (Wang,

2013).

The use of algorithms is an effective way to carry

out automatic optimisation with regard to the chosen

factors for group formation. The most commonly used

algorithms for group formation in studies on CSCL are

probabilistic algorithms, such as genetic algorithms

(Cruz and Isotani, 2014).

It is not clear to what extent the factors for group

formation mentioned before also play a role in the

particular context of voluntary, self-organised study

groups, where students have a higher responsibility in

terms of organising the group or dealing with various

problems that may occur (Melzner et al., 2020). So

far, there is little evidence of how students perceive

the optimal group composition of study groups. A

study by Rybczynski and Schussler examined study

groups in an introductory biology course and investi-

gated students’ preconceptions via a survey (Rybczyn-

ski and Schussler, 2011). Using an open-ended ques-

tion, students were asked for any general comments on

study groups. One of the recurrent identiﬁed themes

is the high importance of group composition, which

according to the students inﬂuences factors such as use-

fulness, effectiveness, and productivity of the group.

Overall, the students in that study mentioned that the

level of knowledge of the other group members and the

willingness to participate were important factors for

them as well as the group size; interestingly, however,

the authors point out that students generally seem to

have different ideas about what makes a good group

member and whether they prefer to learn with friends

or strangers (Rybczynski and Schussler, 2011).

A more recent study also emphasizes the impor-

tance of a homogeneous problem perception within

the group which is characterised by a shared under-

standing about the presence and nature of problems

(Melzner et al., 2020). However, there are no clear rec-

ommendations mentioned on how the groups should

be composed so that students perceive them as effec-

tive and feel able to tackle problems in a meaningful

way.

These ﬁndings underline the importance of group

composition for students and already point to individ-

ual factors that could be important in group formation.

However, no systematic survey has yet been conducted

that covers a larger number of the person-related or

group-related factors that have already been used in the

context of mandatory group work, and analyses their

signiﬁcance in the context of voluntary study groups.

This paper aims to provide a more complete picture of

how students view the factors used in previous work

and what students’ needs are in terms of study groups.

REQUIREMENT ANALYSIS FOR

SELECTING GROUPING

CRITERIA

Probably the most important aspect of forming groups

is the grouping criteria. This section explores possi-

ble factors based on students’ needs and wishes and

describes the analyses that were conducted to deter-

mine which factors may be most important according

to students.

3.1 Method

To explore which factors students perceive as partic-

ularly relevant, a focus group was held as a ﬁrst step.

Focus groups are similar to group interviews in that

they are an organised discussion with a group of indi-

viduals to gain insights regarding a collective opinion;

however, focus groups are more interactive since the

group members do not just answer the interviewers

question but are also encouraged to discuss these ques-

tions and raise new points within the discussion (Gibbs,

2012).

Regarding the present study, the focus group was

held to get a more general overview on which factors

students consider important in the context of study

groups. The participants of the focus group should thus

be able to not only express their opinions on previously

used factors for group formation but also contribute

their own ideas based on their past experiences with

learning in study groups. Based on this general picture,

the factors mentioned in the focus group were then

evaluated with a larger sample in a follow-up online

survey as a second step.

The focus group was held at LMU Munich univer-

sity in Germany via the Zoom video conferencing tool

in January 2022. Based on the recommendations by

Benighaus and Benighaus, a guideline was developed

prior to this (Benighaus and Benighaus, 2012).

The complete guideline can be found in (Schenk, 2022).

Formation of Study Groups: Exploring Students’ Needs and Practical Challenges

649

particular, the guideline included a section containing

questions about different factors for group composi-

tion. Analogous to the categorisation made in previous

research literature (Konert et al., 2014), the guide-

line differentiated between person-related and group-

related factors. Students were presented with examples

for these factors that were obtained from the previous

reviews on the composition of groups in the context of

mandatory group work (Konert et al., 2014; Odo et al.,

2019; Cruz and Isotani, 2014) and they were asked for

their opinion on the importance of these factors for the

formation of voluntary study groups. Students could

also name other factors important to them and discuss

their signiﬁcance for group formation. In addition, the

guideline contained questions about general expecta-

tions that students have regarding a tool for forming

study groups, e. g., how much time the usage of the

tool should take.

Six students from the ﬁelds of computer science,

physics, media informatics, psychology, and school

psychology were recruited via personal contact and

participated in the focus group. The age of the students

ranged from 21 to 31 years. All participants stated that

they already had experience with study groups. The

session of the focus group lasted for about two hours

and was recorded.

To analyse the discussion, an inductive categori-

sation analysis was conducted (Mayring, 2012). This

is a systematic method for analysing and structuring

qualitative data by incrementally developing a system

of categories from the data material. The categories

are formulated as terms or concepts that are then being

used to assign relevant text passages to the respec-

tive category. Thus, the resulting categories represent

umbrella terms that summarise the content of the dis-

cussion (Mayring, 2012).

Based on the categories developed using this

method, it was speciﬁed in more detail which factors

may play an important role in the formation of study

groups from the students’ point of view. While the cat-

egories refer to general topics of the focus group, the

factors represent concrete examples that the students

had mentioned as possibly important factors for group

formation in this context.

Based on the factors identiﬁed this way, an online

survey was conducted via the SoSci Survey web appli-

cation.

The aim of this survey was to prioritise the

factors to decide which factors should be taken into

account by an algorithm for group formation. Thus, it

should be determined which factors are most impor-

tant from the students’ point of view. The aim was

not to identify new factors. However, students had the

https://www.soscisurvey.de/, last accessed: 2023-11-22

opportunity to comment on the study (and thus, also

on possibly missing factors).

In order to be able to do a prioritisation and select

only the most important factors for group formation,

items were designed to indicate that a particular factor

is of high importance for study groups (e. g., for the

factor “Preference for Online Meetings”: “For study

groups in general, it is important that people in the

group have the same preference regarding the type of

meeting (online vs. face-to-face).”). These items could

be rated on a 5-point Likert scale (1=strongly disagree,

3=neither agree nor disagree, 5=strongly agree). To

further motivate respondents to decide which factors

were most important to them in relation to the other fac-

tors, at the end of the questionnaire the students were

asked to select a maximum of three factors that they

considered most important. This way, it was possible

to determine the importance of the factors through two

different measurements: On the one hand, the degree

of agreement with a statement. This value was then

averaged over all respondents. On the other hand, the

frequency with which a statement was mentioned as

one of the most important factors overall. In addition

to these prioritisation items, the questionnaire included

questions about demographic characteristics as well as

personality traits such as extraversion, conscientious-

ness, and perseverance. For the measurement of ex-

traversion and conscientiousness, the respective items

from the 30-item short version of the NEO Five-Factor

Inventory were used (K

orner et al., 2008), further-

more the six items of the scale “Perseverance” were

taken from the 12-item Grit Scale (Fleckenstein et al.,

2014).

The online survey was carried out in February and

March 2022 and lasted 18 days. It was advertised via

various email distribution lists at four German univer-

sities and was open to students from all disciplines.

The data was analysed from all respondents who indi-

cated that they are older than 18 years and accepted

the privacy agreement. The ﬁnal data set consists of

160 students (69 male, 87 female, 4 diverse) and was

analysed using the software R. The average age of the

respondents was 25 years (SD = 7.71, [18, 63]). In to-

tal, students from 37 different ﬁelds of study took part,

with computer science and media informatics being

the most common subjects.

3.2 Results

The results of the focus group are brieﬂy presented ﬁrst,

as they form the basis of the online survey. The topics

discussed in the focus group could be divided into the

The complete online questionnaire can be found in

(Schenk, 2022).

CSEDU 2024 - 16th International Conference on Computer Supported Education

650

following ten categories, using inductive categorisa-

tion analysis: “Interaction with Others”, “Common

Goals”, “Personality Traits”, “Organisation of Study

Groups”, “Attitude towards Reliability”, “Knowledge

and Skills”, “Sociodemographic Criteria”, “Consid-

eration of Health Limits” (e. g., acknowledging that

one also needs breaks and taking them consistently),

“Group Size”, and “General Requirements for a Tool

for the Formation of Study Groups”.

Major themes in the focus group concerned rea-

sons why students join study groups and the resulting

expectations regarding study groups. On the one hand,

the social component of study groups (i. e., the inter-

action with others) can be very important for some

students and may be one of the main reasons why they

decide to join a study group. On the other hand, the

participants of the focus group stated that interaction

with others can also be perceived as distracting and not

being conducive to their learning experience. It was

suggested by the participants of the focus group that

certain factors, such as personality traits, may be more

important for those students who join study groups

mainly for social reasons.

In general, the participants of the focus group dis-

cussed the possibility that students have different learn-

ing goals when it comes to learning in study groups:

For some students it might be a matter of (fully) un-

derstanding the material while for others it might be

just a matter of passing an exam. The participants of

the focus group therefore assumed that depending on

the learning goals, the level of commitment also varies

and that this may be why different learning goals could

probably have a negative impact on group dynamics.

The general organisation of study groups was also

discussed in the focus group. As mentioned before,

students have to take care of the organisation of study

groups themselves and can make different decisions

on this issue: For example, it is not speciﬁed that study

groups must meet in person. In fact, one focus group

participant said that she preferred learning in online

meetings because she perceived it as less distracting

and overwhelming. Another student from the focus

group emphasized that face-to-face meetings were very

important to him in order to get to know the other par-

ticipants in a study group better. It was also discussed

whether it is important to meet regularly or whether

students might prefer more ﬂexible meetings. The

“ideal” group size should be 3 or 4 students according

to the participants of the focus group. This statement

is consistent with the recommendation from the litera-

ture that smaller groups are more likely to promote the

productivity of individual group members than larger

groups (Odo et al., 2019).

Concerning the general requirements for a tool for

group formation, the students indicated that they are in

general not reluctant to use such a tool, but that aspects

such as transparency and trustworthiness are important

to them and that the tool should be restricted to fellow

students. The students also mentioned that using a tool

for group formation should not take more than 10 to

15 minutes of their time.

As described above, the ten categories of the focus

group were used as a basis to identify speciﬁc fac-

tors that may be important for group formation and

to develop respective questionnaire items for them as

described in the previous section. A total of 20 factors

were found this way. The items that each describe the

importance of one speciﬁc factor were then used in

an online survey. Table 1 shows the factors and sum-

marises the descriptive statistical analyses regarding

the item prioritisation: Shown are the averaged agree-

ment with an item and the frequency of mention in the

evaluation of the factors. In addition, the correspond-

ing category from the focus group is given for each

item. Note that while group size was mentioned in

the focus group, the question of an appropriate group

size was not listed in the online questionnaire. As the

statements from the focus group correspond with rec-

ommendations from the literature, a group size of 3–4

students was decided in advance instead.

The following ﬁve factors received the highest

average agreement (cf. Table 1): “Sense of Duty”,

“Attitude towards Reliable Attendance of Meetings”,

“Attitude towards Regular Meetings”, “Preference for

Online Meetings”, and “Accuracy in Working”. The

ﬁve factors that were mentioned most frequently in-

clude “Common Goals”, “Sense of Duty”, “Attitude

towards Reliable Attendance of Meetings”, “Prefer-

ence for Online Meetings”, and “Goal-Orientation and

Perseverance”.

In the questionnaire, students were also asked for

their attitudes regarding study groups (“I like learning

in study groups”), and how much experience they had

with learning in study groups (“I have often learned

in study groups”). Both items were again rated on the

same 5-point Likert scale as described in Section 3.1.

The average agreement regarding the ﬁrst item was

3.24 (SD = 1.25), the average agreement regarding

the second item was 3.56 (SD = 1.33). In addition,

students were also asked whether they could imagine

using an application or tool for ﬁnding a study group

(“I could imagine using an application to ﬁnd a new

study group.”). The average agreement regarding this

item was 3.6 (SD = 1.2).

Formation of Study Groups: Exploring Students’ Needs and Practical Challenges

651

Table 1: Factors for Group Formation and Items used in the Online Study, along with Descriptive Data for Prioritisation.

Factor Category in Focus Group Mean SD Frequency

Common Goals Common Goals 3.88 1.04 74

Homogeneity with regard to Gender Sociodemographic Criteria 1.24 0.69 1

Homogeneity with regard to Age Sociodemographic Criteria 2.25 1.23 10

Preference for Online Meetings Organisation of Study Groups 4.04 0.89 50

Attitude towards Regular Meetings Organisation of Study Groups 4.19 0.86 31

Setting one Appointment per Week Organisation of Study Groups 3.26 1.25 9

Attitude regarding the Distribution of Tasks Organisation of Study Groups 3.64 1.08 20

Attitude towards Reliable Attendance of Meet-

ings

Attitude towards Reliability 4.25 0.89 51

Similarity in Personality Interaction with Others 2.94 1.00 10

Difference in Personality Interaction with Others 2.74 0.94 7

Similarity in Extraversion Personality Traits 2.73 1.07 8

Difference in Extraversion Personality Traits 3.16 1.03 7

Sense of Duty Personality Traits 4.26 0.76 73

Accuracy in Working Personality Traits 3.91 0.95 26

Common Interests Interaction with Others 2.73 1.20 9

Common Hobbies Interaction with Others 2.16 1.08 1

Goal-Orientation and Perseverance Personality Traits 3.78 0.93 35

Pursuing of Goals despite Setbacks Personality Traits 3.54 1.06 6

Consideration of Health Limits Consideration of Health Limits 3.85 1.00 18

Similar level of Knowledge Knowledge and Skills 3.12 1.15 18

Different level of Knowledge Knowledge and Skills 2.81 1.07 11

3.3 Selection of Relevant Factors for

Group Formation

The analysis of students’ needs and perceptions regard-

ing study groups presented in the previous sections

formed the basis for the development of a prototype.

In particular, based on the analysis, the selection of

factors for group formation was made. This selection

was made in two steps: Firstly, the averaged agreement

was used, i. e., the extent to which the students agreed

with the presented items. Secondly, it was examined

how often the students had counted a factor as one of

the three most important factors in total.

The following three factors were selected since

they showed both, a high averaged agreement and a

high frequency of mention: “Sense of Duty”, “Atti-

tude towards Reliable Attendance of Meetings”, and

“Preference for Online Meetings”. It should be noted

that the factor “Common Goals” was also considered

as a criterion for group formation, as it was mentioned

by most students (

n = 74

) as one of the three most im-

portant criteria for study groups. Interestingly though,

the mean agreement was lower and the factor also

had a higher standard deviation than the three factors

that were ultimately chosen for group formation. It

can therefore be assumed that this factor was less im-

portant for the entire sample which is why it was not

considered for the algorithm for group formation here.

In addition, difﬁculties with regard to the measurement

of “Common Goals” also complicated the selection of

this factor. The online study gave students the oppor-

tunity to mention in free text possible goals that they

personally have when learning in study groups. A wide

range of different goals were mentioned, such as a gen-

eral exchange of information about university courses

or the encouragement of interdisciplinarity. These ﬁnd-

ings indicate that students are pursuing study groups

with different objectives. However, it is unclear how

to summarise or operationalise these goals. To collect

data or preferences on this factor from students, further

analysis is needed.

In summary, it seems to be particularly important

for students that the other people are similarly duti-

ful or conscientious, that they have a similar attitude

regarding the reliable attendance of meetings, and a

similar preference for online meetings. Thus, an algo-

rithm for group formation should form groups that are

as homogeneous as possible regarding these factors.

4 DEVELOPMENT OF A

PROTOTYPE FOR GROUP

FORMATION

The prototype consists of two components: the stu-

dent front-end for allowing students to register and

CSEDU 2024 - 16th International Conference on Computer Supported Education

652

for asking for their preferences, and a back-end for

the group formation and sending notiﬁcation emails

to the students containing contact information of their

recommended fellow students for a study group.

The student front-end was developed as a simple

web-application using Spring Boot. It allowed stu-

dents to register before a deadline, and stored all data

in a database. Shibboleth Single-Sign On was used to

restrict usage to students enrolled at the same univer-

sity and to receive their (validated) names and email

addresses. The registration required students to ﬁll out

a form that on the one hand included administrative

data such as their name, email address, university and

semester (all pre-ﬁlled). Further questions were de-

veloped to operationalise the three factors for group

formation found in the previous section: Two questions

were developed to measure preference for online meet-

ings, and three items to measure the attitudes towards

the reliable attendance of meetings in a study group.

In addition, the six-item conscientiousness scale of the

30 item short version of the German NEO Five-Factor

Inventory (K

orner et al., 2008), based on the English

NEO Five-Factor Inventory (Costa and McCrae, 1989),

was used to assess the sense of duty. Two further items

asked for the importance of the grade and the envi-

sioned grade for that course. The form thus comprised

a total of 17 items which are listed in Table 2.

The back-end was developed as a stand-alone Java

application that needs to be invoked (manually) after

the registration deadline to compose study group rec-

ommendations using a genetic algorithm approach. A

genetic algorithm was chosen, because this type of

algorithm performs well (also for large data sets) and

has been used successfully by many previous studies

on group formation (Krouska et al., 2019). For each

course, the grouping algorithm uses the data of the

registered students on online preference, their attitude

towards reliability, and conscientiousness for the ﬁt-

ness function to form homogeneous groups, i. e., the

average differences on the answers for the three factors

should be minimal (Schenk, 2022). The implementa-

tion of the genetic algorithm is inspired by a previous

study using a genetic algorithm for the formation of

student groups of three (Moreno et al., 2012), the prob-

ability parameters for crossover (1.0) and mutation

(0.5) were determined systematically by an evaluation

with test data (Schenk, 2022). The genetic algorithm

composes groups of three students as requested in the

focus group. If the number of students is not a multiple

of three, in the ﬁnal step one or two “fourth students”

are added to the best ﬁtting groups. After all students

have been assigned to groups, the tool sends emails to

all students containing the names and email addresses

of their fellow students of the respective recommended

group and asks to contact them.

To ensure that the genetic algorithm actually

achieves a higher ﬁtness (i. e., forms more homoge-

neous groups regarding the three factors) than, for

example, a random algorithm, the algorithm was eval-

uated using randomly generated data. A comparison

was made between the genetic algorithm, a backtrack-

ing (brute force) algorithm (which considers all the

options for grouping and then selects the one with the

best ﬁtness), and a random selection of students. The

backtracking method is capable of ﬁnding an optimal

solution for small groups and was used to estimate

how close the genetic algorithm could come to that

solution; however, it cannot be used effectively with

more than 30 students.

For comparing the three algorithms, the number

of students was systematically varied and the ﬁtness

values (calculated via the ﬁtness function which in-

dicates how homogeneous groups are with regard to

the three factors) were compared with each other. The

evaluation showed that the genetic algorithm does not

achieve a better ﬁtness for a smaller number of stu-

dents (7–11 students) than a random selection. From a

number of 12 students onwards, the genetic algorithm

achieved a better ﬁtness and it was able to achieve bet-

ter ﬁtness values in less than two minutes for a number

of 3,000 students (Schenk, 2022).

5 EVALUATION IN A CASE

STUDY

The prototype and the proposed factors were evaluated

in two different ﬁrst-semester Bachelor’s courses in

a ﬁeld study at Technical University of Munich (ﬁrst-

semester computer science course, CS in the following,

approx. 1,500 students) and Aalen University of Ap-

plied Sciences (math course, approx. 80 students) in

winter term 2022/2023 in Germany. Both courses were

attended by students of different courses of study.

5.1 Method

For both courses the prototype was made available

to students in the ﬁrst week of the semester and the

registration was open for one week. In the math course

the study was announced by the teacher and in the CS

course by one of the authors. The composed groups

were announced using email on November, 1st (begin-

ning of the third week of the semester).

These two courses were selected, to investigate two

hypotheses: 1) study groups formed by the proposed

algorithm are perceived as better and persist longer

Formation of Study Groups: Exploring Students’ Needs and Practical Challenges

653

Table 2: Items used in the Web-Application (translated from German into English).

Item Formulation Response Format

Your Name Free text ﬁeld (preﬁlled)

Your Email Address Free text ﬁeld (preﬁlled)

For which university are you looking for a study group? Free text ﬁeld (preﬁlled)

Which semester are you in? Drop-down menu (1,2,3,4,5,6,>6)

I prefer to learn with a study group in presence. 5-point Likert scale

I prefer to learn with a study group online. 5-point Likert scale

When I am in a study group, I try to attend every meeting of the study

group.

5-point Likert scale

It is important to me that other people in a study group attend every

meeting.

5-point Likert scale

It is important to me that other people in a study group are willing to

cancel other appointments for attending meetings of the study group, if

necessary.

5-point Likert scale

I keep my belongings clean and neat. 5-point Likert scale

I’m pretty good about pacing myself so as to get things done on time. 5-point Likert scale

I try to perform all the tasks assigned to me conscientiously. 5-point Likert scale

When I make a commitment, I can always be counted on to follow

through.

5-point Likert scale

I am a productive person who always gets the job done. 5-point Likert scale

I never seem to be able to get organised. 5-point Likert scale

A good grade is very important to me. 5-point Likert scale

My target grade for this module:

Drop-down menu (Grade A, Grade

B, Grade C, Grade D, no target

grade)

than randomly composed groups, and 2) there is no

difference in these terms between a course with a very

high number of students and a course with a lower

number of students. To investigate the ﬁrst hypothesis,

the students of the CS course were either assigned to

a study group formed by the genetic algorithm or to a

study group formed by a random process. Due to the

size of the math course (and “only” 26 registered stu-

dents for the study), all groups were composed using

the genetic algorithm for that course.

To evaluate both hypotheses, a web-based ques-

tionnaire was developed and a personalized link was

sent to all registered students four weeks before the

end of the term and a reminder one week later. The

questionnaire was accessible online for three weeks. It

included items asking whether there had been contact

to other members of the study group and, if so, how

often and regularly the students had met. Addition-

ally, items were included asking how satisﬁed students

were with their study groups. To also assess how pleas-

ant the atmosphere in the group and the cohesion was

perceived, the four items of the Participative Safety

scale of the German version of the Team Climate In-

ventory were used (Brodbeck et al., 2000). In addition,

the online questionnaire included the eight items of

the “Cohesion” scale of the German questionnaire on

teamwork (Kauffeld and Frieling, 2004, “Fragebogen

zur Arbeit im Team”).

5.2 Results

The prototype was used by a total of 287 students (CS

course: 261, math course: 26), 151 of whom were

in their ﬁrst semester (CS course: 133, math course:

18). For the CS course, 87 groups were composed (44

randomly and 43 by the genetic algorithm) and for the

math course 9 groups were composed.

There were 39 persons from both universities par-

ticipating in the evaluation study (CS course: 31, math

course: 8). The students’ ages ranged from 18 to 29

(M = 20.74, SD = 2.62). The majority of the students

(

n = 31

) was in their ﬁrst semester. 17 students were in

a randomly composed group (only CS course), and 22

in a group that was composed by the genetic algorithm

(CS course: 14, math course: 8).

Out of the 39 students, 15 stated that contact had

been established with all members of their study group

(CS course: 11, math course: 4), 7 stated that they

had been in contact with some members of their study

group (CS course: 4, math course: 3), and 17 students

stated that there had been no contact at all (CS course:

16, math course: 1). Out of the 22 students who replied

that a contact had – to some extent – been established,

CSEDU 2024 - 16th International Conference on Computer Supported Education

654

14 students stated that they had not yet had a meeting

with other members of their group, four students stated

that they had met less frequently than once a week,

three students stated that they had met once a week,

and one person stated that they had met multiple times

a week.

Due to the small number of people who took part

in the ﬁnal questionnaire and had actually met in their

groups, no further in-depth analyses regarding the two

formulated hypotheses were conducted.

A free text ﬁeld gave people the opportunity to in-

dicate difﬁculties they had encountered in connection

with their study group. Two students stated that other

members did not reply, and thus no contact could be

established. Furthermore, it was mentioned that people

did not have time to schedule a meeting for studying:

One student remarked that an actual meeting for study-

ing did not happen because the others could not ﬁnd

the time. Another student stated that no meeting took

place because one person from the study group never

got in touch and the other was too busy. To facilitate

an exchange, it was suggested to pay more attention to

a common study program and interests.

6 DISCUSSION

The requirement analysis described in this paper identi-

ﬁed three factors that are important to students. These

factors can be described as person-related factors that

relate to personality traits of the group members and

to their preferences regarding the organisation of the

meetings.

In general, the current work suggests that students

themselves prefer a homogeneous group composition

regarding the mentioned factors. This may be due to

an enhanced trust with group members that are simi-

lar to oneself, as previous research literature suggests

(Ennen et al., 2015) and is in line with previous ﬁnd-

ings that have shown that when group formation is

left to students themselves, they tend to form more ho-

mogeneous groups, e. g., regarding person-related fac-

tors such as level of knowledge or sociodemographic

variables (Razmerita and Brun, 2011; Freeman et al.,

2017). It is conceivable that a similar level of consci-

entiousness may also result in a similar perception of

problems, which has already been shown to be impor-

tant for self-organised collaborative learning (Melzner

et al., 2020).

In contrast to previous studies, the current study

also highlights the importance of students’ preference

for either face-to-face or online meetings: The group

formation algorithm proposed in this study also takes

the factor “Preference for Online Meetings” into ac-

count which seems to be very important for many stu-

dents. This is a factor that has not been investigated in

particular by previous research literature. The fact that

students attribute high importance to this factor may

be due to the COVID pandemic during which meetings

in person were hardly possible or not possible at all; as

a result, online meetings and in particular, online learn-

ing became increasingly important and commonplace

for students (Shaid et al., 2021).

Two limitations should be noted regarding these

ﬁndings: Firstly that it cannot be clearly stated whether

the three factors identiﬁed in the requirement analysis

are indeed the factors that are most important to the

population of students since this was a purely descrip-

tive statistical analysis. Efforts were made to obtain a

representative sample by addressing various students

at multiple (German) universities. However, the ﬁ-

nal participating sample of students was inﬂuenced by

self-selection and it is not clear how representative the

data is for all students with an interest in study groups

since a majority of participants seemed to study STEM

subjects such as computer science or media informat-

ics and some participants stated that they themselves

were not actually interested in learning in study groups.

Furthermore, only one focus group of six people was

used to select potentially relevant factors. Although

care was taken to select students from different dis-

ciplines who already had experience of learning in

self-organised study groups, it cannot be ruled out that

other (possibly relevant) factors would have been men-

tioned in a further survey with different people. Due to

restricted resources, the inductive categorisation anal-

ysis for analysing the focus group was carried out by

one person which limits the reliability of this analysis.

Lastly, the results of the requirement analysis may

be biased because of the already mentioned potential

impact of the COVID pandemic: the focus group and

the online study took place in the beginning of 2022

and, therefore, might be inﬂuenced by the COVID

pandemic. It cannot be deﬁnitively stated whether

these ﬁndings are still valid in current times when

everyday life is no longer restricted by the pandemic.

Secondly, it should be noted that the factors iden-

tiﬁed in the requirement analysis are only based on

students’ own perceptions and therefore do not neces-

sarily have to have a (signiﬁcant) positive impact on

the learning experience. As pointed out before, the

aim of the present study was in particular to take stu-

dents opinions into account since they are probably

best placed to assess which factors they perceive as

beneﬁcial or detrimental to learning in study groups.

However, students may not only pursue the goal of

achieving learning success with study groups – the re-

sults of the requirement analysis indicate that students

Formation of Study Groups: Exploring Students’ Needs and Practical Challenges

655

may pursue various different goals. It is therefore

possible that students consider certain factors to be

important for study groups which may not necessarily

be related to an improved learning experience or aca-

demic outcomes but to other goals that students had in

mind.

The evaluation study attempted to address this ﬁnd-

ing by using different items in the online questionnaire

to assess not only learning achievements, but also as-

pects such as the sense of cohesion within the study

group. However, no clear statements can be made

as to whether the group formation carried out by the

genetic algorithm (which forms homogeneous groups

according to the three factors) is also better in terms

of learning success or group cohesion than the group

formation carried out by a random algorithm due to

the small sample size in the evaluation study. The

questions of whether groups formed by the proposed

algorithm are perceived as better and last longer and

whether there is a difference between the two courses

of different size could thus not be answered. Neverthe-

less, the evaluation study provided certain insights into

the challenges that may hinder students from learn-

ing with their study groups and that should be taken

into account when designing a prototype for group

formation of study groups.

First of all, it should be pointed out that a large

number of students stated in the evaluation study that

contact with other group members did not take place

at all. This may be due to the design of the prototype

where the students were only given the names and the

email addresses of the other group members as well as

a prompt to establish contact. The actual establishment

of contact was ultimately left to the students them-

selves. It is possible that the students felt reluctance

to write to other students they did not know or that

there was a general lack of clarity about who should

write a message to the others ﬁrst. Another possibility

is that contacting students via email (as conﬁgured in

the university’s directory; default was the university

email account) is generally not convenient for them.

They may ﬁnd this way of communication more com-

plicated and they may not check their emails often

enough to respond in a short period of time. Another

reason could be that the study group assignment was

made too late in the semester (beginning of the third

week) and groups might already have formed on other

ways. Nothing was noted about this in the evaluation

questionnaire.

Even if the contact was established, scheduling

difﬁculties also posed a challenge for the students

which sometimes prevented them from studying to-

gether. There are several such comments in the eval-

uation study. The problem that students like to learn

with others but cannot ﬁnd a suitable time slot was

also discussed in the focus group. It was suggested by

the participants of the focus group that the tool could

ask students about the possible free time slots or pre-

ferred days. Especially for students who want to study

regularly, it could be important to ﬁnd a common time

slot per week. Though, the results of the online study

showed that one shared meeting per week was less

important to students than the other factors mentioned.

Hence, features for ﬁnding an appointment, were not

further considered in the design of the prototype. But

since the courses in which the prototype was presented

turned out to be quite diverse regarding the attend-

ing students’ semesters and courses of study, students

might have been overwhelmed with the task of ﬁnding

an appointment on their own. Thus, taking information

about the students’ semester or course of study into

account could further reduce the difﬁculties of schedul-

ing an appointment. This assumption is supported by

one student commenting in the evaluation that all three

people in their study group studied a different major

which caused problems in scheduling.

In general, it seems as if there is an interest by a

notable proportion of students who used the prototype

in the beginning of the semester to ﬁnd a study group

(CS course: 261 of approx. 1,500 students, 17 % vs.

math course: 26 of approx. 80 students, 32 %). This

however, can only be seen as a rough tendency, as the

interest depends on various factors such as the way

of advertising, the university size or types, and the

concrete courses.

7 SUMMARY AND OUTLOOK

The basic aim of this study was to gain a better un-

derstanding about the formation of study groups. In

particular, two questions were addressed: The ques-

tion of students’ needs in relation to the factors used

for the formation of study groups and the question of

possible challenges that need to be considered when

forming study groups. Through a requirement analysis

it was examined how students imagine their ideal study

groups. Study groups were perceived as most useful

by students when being as homogeneous as possible

regarding the three factors “Sense of Duty”, “Attitude

towards Reliable Attendance of Meetings” and “Pref-

erence for Online Meetings”. Based on this ﬁnding, a

prototype was implemented using a genetic algorithm

for group formation. To evaluate the prototype, a ﬁeld

study was carried out. Due to a small sample size in

the ﬁnal questionnaire, the main questions of how well

the groups formed by the prototype actually learn to-

gether and whether there are differences between large

CSEDU 2024 - 16th International Conference on Computer Supported Education

656

or smaller university courses could not be answered,

but difﬁculties were identiﬁed that should be consid-

ered in the future: On the one hand, that contact is

often not established when the task of making contact

is left to the students themselves. On the other hand,

that even if contact is made, it does not also mean that

meetings of the study group will take place. This is

probably due to scheduling difﬁculties.

The paper thus provides initial indications of what

may be important to students when learning in study

groups and what needs to be considered when im-

plementing a tool for the formation of self-organised

study groups. Future studies should, for example, ex-

amine which goals students pursue with learning in

study groups. The present work indicates that there

are apparently different goals that motivate students

to learn in study groups, and that common goals may

be important for group formation. In addition, future

research would be needed to examine how the students

perceive learning in the groups formed by the genetic

algorithm presented in this paper. With regard to the

implementation of a tool for the formation of study

groups, the question of how to facilitate contact re-

mains to be answered. Future studies could explore

this question further and, for example, try to get in-

formation through surveys about what could motivate

students to contact others and what tends to prevent

them from doing so. In addition, it could be help-

ful to consider characteristics such as the number of

semesters or the subject of study when forming study

groups in order to make it easier for the students to

schedule meetings.

All in all, an application for ﬁnding study groups

seems to be considered useful by the students – the fact

that the prototype presented in this paper was used by

287 students overall indicates that there is a need for

such a tool in higher education. Nevertheless, further

research is needed to fully ensure the expectations of

students – especially regarding their different goals –

are met and to facilitate the contact establishment and

scheduling of meetings.

The source code of the registration form (incl. the

questionnaire) as well as the grouping tool (Schenk and

Strickroth, 2024a), and the evaluation questionnaire as

well as the raw data (Schenk and Strickroth, 2024b)

are available on Zenodo.

ACKNOWLEDGEMENTS

The authors thank the two educators Ruediger Wester-

mann and Armin Egetenmeier at Technical University

of Munich respective Aalen University of Applied Sci-

ences and also all participants.

REFERENCES

Benighaus, C. and Benighaus, L. (2012). Moderation,

Gespr

achsaufbau und Dynamik in Fokusgruppen. In

Fokusgruppen in der empirischen Sozialwissenschaft:

Von der Konzeption bis zur Auswertung, pages 111–

132. Springer.

Brodbeck, F. C., Anderson, N., and West, M. (2000). Das

Teamklima-Inventar. Hogrefe.

Cook-Sather, A. (2002). Authorizing students’ perspectives:

Toward trust, dialogue, and change in education. Edu-

cational researcher, 31(4):3–14.

Costa, P. T. and McCrae, R. R. (1989). The NEO PI/FFI

Manual Supplement. Psychological Assessment Re-

sources.

Cruz, W. M. and Isotani, S. (2014). Group formation algo-

rithms in collaborative learning contexts: A systematic

mapping of the literature. In Proc. CRIWG, pages

199–214. Springer.

Davidson, N., Major, C. H., and Michaelsen, L. K.

(2014). Small-group learning in higher educa-

tion—cooperative, collaborative, problem-based, and

team-based learning: an introduction by the guest edi-

tors. JECT, 25(3-4):1–6.

Dillenbourg, P. (1999). What do you mean by collabora-

tive learning? Collaborative-learning: Cognitive and

Computational Approaches, 1:1–19.

Dillenbourg, P. and Fischer, F. (2007). Computer-supported

collaborative learning: The basics. Zeitschrift f

Berufs- und Wirtschaftsp

adagogik, 21:111–130.

Ennen, N. L., Stark, E., and Lassiter, A. (2015). The impor-

tance of trust for satisfaction, motivation, and academic

performance in student learning groups. Social Psy-

chology of education, 18:615–633.

Fleckenstein, J., Schmidt, F. T., and M

oller, J. (2014). Wer

hat biss? Beharrlichkeit und best

andiges interesse von

lehramtsstudierenden; Eine deutsche adaptation der

12-ltem Grit Scale. Psychologie in Erziehung und

Unterricht, 61(4):281–286.

Freeman, S., Theobald, R., Crowe, A. J., and Wenderoth,

M. P. (2017). Likes attract: Students self-sort in a class-

room by gender, demography, and academic character-

istics. Active Learning in Higher Education, 18(2):115–

126.

Gibbs, A. (2012). Focus groups and group interviews.

Research methods and methodologies in education,

186:192.

Gillies, R. M. (2016). Cooperative learning: Review of

research and practice. Australian Journal of Teacher

Education (Online), 41(3):39–54.

Jacobs, G. M. (2015). Collaborative learning or coopera-

tive learning? the name is not important; ﬂexibility is.

Online Submission, 3(1):32–52.

Kauffeld, S. and Frieling, E. (2004). Fragebogen zur Arbeit

im Team: FAT. Hogrefe.

Konert, J., Burlak, D., and Steinmetz, R. (2014).

GroupAL: ein Algorithmus zur Formation und

Qualit

atsbewertung von Lerngruppen in E-Learning-

Szenarien. i-com, 13(1):70–81.

Formation of Study Groups: Exploring Students’ Needs and Practical Challenges

657

orner, A., Geyer, M., Roth, M., Drapeau, M., Schmutzer,

G., Albani, C., Schumann, S., and Br

ahler, E.

(2008). Pers

onlichkeitsdiagnostik mit dem NEO-F

unf-

Faktoren-Inventar: Die 30-Item-Kurzversion (NEO-

FFI-30). PPmP, 58(06):238–245.

Krouska, A., Troussas, C., and Virvou, M. (2019). Applying

genetic algorithms for student grouping in collabora-

tive learning: A synthetic literature review. Intelligent

Decision Technologies, 13(4):395–406.

Laal, M. and Ghodsi, S. M. (2012). Beneﬁts of collaborative

learning. Procedia-social and behavioral sciences,

31:486–490.

Magnisalis, I., Demetriadis, S., and Karakostas, A. (2011).

Adaptive and intelligent systems for collaborative learn-

ing support: A review of the ﬁeld. IEEE transactions

on Learning Technologies, 4(1):5–20.

Mayring, P. (2012). Qualitative Inhaltsanalyse–ein Beispiel

ur Mixed Methods. In Mixed Methods in der em-

pirischen Bildungsforschung, pages 27–36. Waxmann.

Melzner, N., Greisel, M., Dresel, M., and Kollar, I. (2020).

Regulating self-organized collaborative learning: The

importance of homogeneous problem perception, im-

mediacy and intensity of strategy use. International

Journal of Computer-Supported Collaborative Learn-

ing, 15:149–177.

Moreno, J., Ovalle, D. A., and Vicari, R. M. (2012). A

genetic algorithm approach for group formation in col-

laborative learning considering multiple student char-

acteristics. Comput. Educ. J., 58(1):560–569.

Nijstad, B. A. and De Dreu, C. K. (2002). Creativity and

group innovation. Applied Psychology, 51(3):400–406.

Odo, C., Masthoff, J., and Beacham, N. (2019). Group forma-

tion for collaborative learning: A systematic literature

review. In Proc. AIED, pages 206–212. Springer.

Panitz, T. (1999). Collaborative versus cooperative learning:

A comparison of the two concepts which will help us

understand the underlying nature of interactive learn-

ing. U.S. Department of Education. Retrieved from

https://ﬁles.eric.ed.gov/fulltext/ED448443.pdf.

Razmerita, L. and Brun, A. (2011). Collaborative learning

in heterogeneous classes: towards a group formation

methodology. In Proceedings of the 3rd International

Conference on Computer Supported Education, pages

189–194.

Rybczynski, S. M. and Schussler, E. E. (2011). Student

use of out-of-class study groups in an introductory

undergraduate biology course. CBE—LSE, 10(1):74–

82.

Schenk, C. (2022). Anforderungsanalyse und Design

eines Tools zur Bildung von Lerngruppen. On-

line, https://www.tel.iﬁ.lmu.de/pubs/2022/lmu/

anforderungsanalyse-und-design-eines-tools-zur-

bildung-von-lerngruppen.pdf. Bachelor’s Thesis,

LMU Munich.

Schenk, C. and Strickroth, S. (2024a). Algorithm. Online,

http://doi.org/10.5281/zenodo.10678081.

Schenk, C. and Strickroth, S. (2024b). Evaluation question-

naire and data. Online, http://doi.org/10.5281/zenodo.

10678094.

Shaid, N. A. N., Kamruzaman, F. M., and Sulaiman, N.

(2021). Online learning during ongoing covid-19 pan-

demic: A survey of students’ satisfaction. International

Journal of Academic Research in Business and Social

Sciences, 11(7):924–937.

Shaw, R.-S. (2013). The relationships among group size, par-

ticipation, and performance of programming language

learning supported with online forums. Computers &

Education, 62:196–207.

Strickroth, S. and Bry, F. (2022). The future of higher ed-

ucation is social and personalized! experience report

and perspectives. In Proceedings of the 14th Interna-

tional Conference on Computer Supported Education,

volume 1, pages 389–396. INSTICC, SciTePress.

Wang, Z. (2013). Effects of heterogeneous and homoge-

neous grouping on student learning. Master’s Thesis,

University of North Carolina at Chapel Hill university.

https://doi.org/10.17615/j3k8-fz95.

Zhan, Z., He, G., Li, T., He, L., and Xiang, S. (2022). Effect

of groups size on students’ learning achievement, mo-

tivation, cognitive load, collaborative problem-solving

quality, and in-class interaction in an introductory

ai course. Journal of Computer Assisted Learning,

38(6):1807–1818.

CSEDU 2024 - 16th International Conference on Computer Supported Education

658