Conference PaperPDF Available

Learning programming through stepwise self-explanations


Abstract and Figures

In this article we present an approach where students self-explain small pieces of code while they are studying the process of building programs from videos used as instructional material. To evaluate our approach, we carried out an experiment in which we compared it with another approach where students self-studied the same videos. Although we could not confirm that the difference between the two groups was significant, we found a positive correlation within the instructional group between participants' answers to the self-explanation questions and participants' final results. Besides that, participants provided positive feedback regarding our approach. These findings suggest our approach should be investigated in further detail, especially with regard to which instructional conditions are more effective for it.
Content may be subject to copyright.
Learning programming through
stepwise self-explanations
Viviane C. O. Aureliano
Federal Institute of Pernambuco
Campus Belo Jardim
Belo Jardim, Pernambuco, Brazil
Patricia C. de A. R. Tedesco
Center for Informatics
Federal University of Pernambuco
Recife, Pernambuco, Brazil
Michael E. Caspersen
Centre for Science Education
Aarhus University
Aarhus, Denmark
Abstract In this article we present an approach where students
self-explain small pieces of code while they are studying the
process of building programs from videos used as instructional
material. To evaluate our approach, we carried out an
experiment in which we compared it with another approach
where students self-studied the same videos. Although we could
not confirm that the difference between the two groups was
significant, we found a positive correlation within the
instructional group between participants’ answers to the self-
explanation questions and participants’ final results. Besides
that, participants provided positive feedback regarding our
approach. These findings suggest our approach should be
investigated in further detail, especially with regard to which
instructional conditions are more effective for it.
Keywords - Learning programming; novices; videos; self-
The literature about programming education shows that
novice learners experience several difficulties in acquiring
programming skills. Novice programmers have fragile
knowledge and, because of that, struggle when they have to
apply the acquired knowledge in new situations of use [1].
Nevertheless, the major difficulty experienced by novices is
how to combine and use basic structures appropriately to build
a program [1][6][7].
As a way to overcome these difficulties, teachers should
guide novice learners while they are teaching the programming
process. One means of guiding novices is through the use of
Stepwise Improvement (SI) [6][7], a framework that describes
programming as an iterative and incremental process. Using
instructional material structured according to this framework,
novices can learn programming by developing small pieces of
code in a systematic and incremental way.
However, since students do not know how to study for the
needs of a programming course [25], the structure of the
material alone is not sufficient to make students learn. Thus,
besides the guidance regarding the structure of the instructional
material, novice learners should be guided while they are
learning the programming process. One evidence-based
practice for improving student learning by guiding them while
they are studying from some instructional material is through
using self-explanations (SEs) [2][3][4]. In the programming
education area, previous studies have shown SE practice is
beneficial for those learners who use it [4][13][16][17][18][19].
In this context, our approach combines the SI framework
with SEs in order to promote and guide novice learners’ SEs
while they are learning the programming process from
examples presented in videos used as instructional material.
Through this approach, we expect novices to study
programming appropriately from these examples, and, in
consequence become able to apply and use concepts that they
have learned in order to implement programs that match
defined requirements.
SI is a conceptual framework that describes programming
as a systematic and incremental process that encompasses three
types of activities: extension, refinement and restructuring,
organized in a three-dimensional space that is explored by
programmers while they are building programs [6][7].
Extension occurs when the specification is expanded in order
to cover more (use) cases. Refinement occurs when abstract
code is modified and becomes an executable code that
implements the current specification. Restructuring occurs
when an improvement of nonfunctional aspects of a solution is
made (i.e., this modification does not involve a change in the
apparent behavior of the solution). Figure 1 illustrates a
development sequence according to SI. It consists of five
ordered steps: extension, refinement, extension, refinement and
SI provides a general framework for the characterization of
the programming process. However, the primary motivation for
developing the SI framework was to use it for educating
novices in the skills of programming. Caspersen and Kölling
[7] advocate its application in a similar way to using guided
tours rather than leaving students to walk randomly on their
own. In order to conduct learners’ steps in the programming
space offered by the framework, the authors state that teachers
should be concerned with the right amount of guidance given
during instruction. Using SI guarantees that important aspects
of programming education are balanced: training learners’
programming skills and, at the same time, keeping the
cognitive load under control while students are learning.
In that sense, SI brings a significant contribution to the field
of programming education. It provides guidance in the
activities that it encompasses. Regarding the extension and
restructuring activities, it provides guidance in the way that the
instructional material is str uctured. Therefore, programming
textbooks, assignments, lectures and examples, among other
instructional materials, can be structured using the activities
defined by the framework. Regarding the refin ement activity,
Caspersen [7] defined an object-oriented programming process
for teaching n ovice learners.
Figure 1. An example of a development sequence in SI.
The SI framework provides guidance regarding the
structure of instructional material. However, it does not provide
any guidance regarding the way that students can study and
understand instructional material. In our approach, we propose
such guidance through the use of SEs. Clark, Nguyen and
Sweller [11] defined SE as “a mental dialog that learners have
when studying a worked example that helps them understand
the example and build a schema from it”. According to Chiu
and Chi [10], the activity of self-explaining promotes learning
through elaboration of information being studied, associating
this new information with learners’ prior knowledge, making
inferences, and connecting two or more pieces of the given
The benefits of SE (i.e., SE effect) were first shown by Chi,
Bassok, Lewis, Reimann and Glaser [2]. They found that good
students, who self-generated a greater number of explanations
while studying examples in the physics domain, scored better
at problem solving when compared to poor students. Good
students’ explanations provided justifications for steps in the
examples and related those steps to the concepts presented in
the instructional material. Those students also monitored their
understanding while studying the examples.
After that, the SE effect was demonstrated by many other
studies in different domains, instructional contexts and types of
instructional materials [3][10][11][12][14][15]. The studies by
Pirolli and Recker [13] and Bielaczyc, Pirolli and Brown [4]
were the first ones to demonstrate the SE effect in the
programming domain. Pirolli and Recker [13] suggested that
learners who could explain programs in an abstract manner or
could explain the operation of the program learned better.
Bielaczyc, Pirolli and Brown [4] demonstrated that students
who were trained to self-explain while studying from
instructional texts in cluding examples obtained a better
programming performance when compared to those students
who were not trained.
More recently, some other studies have investigated the
effects of SE in the programming domain. Kwon, Kulamasari
and Howland [16] demonstrated the benefits of using open SE
prompts while debugging web-program code in an online
learning environment. Yen-Chu [17] included additional
classes in a course where an instructional gr oup learned how to
self-explain computer architecture diagrams to learn assembly.
These classes fostered learning of assembly and computer
architecture in the instructional group. Lee, Ko and Kwan [18]
added multiple choice and SE assessment levels to a
programming educational game. Their findings suggested
greater student engagement and understanding. Vihavainen,
Miller and Settle [19] included SE questions in programming
assignments and demonstrated that students who received the
SE instructional material performed better than those who did
Our approach is built on two lines of research: (1) the SI
framework and (2) SEs. The SI framework was used to
structure a set of examples presented in videos. We have
chosen videos as instructional material in our approach because
of their advantages. First, in contrast to textbooks, which are
static, videos are an ideal instrument for showing the dynamic
process of programming to novices [5]. For instance, videos
play a fundamental role in showing how to use an IDE or how
to implement programs incrementally, as proposed by the SI
framework [5]. Second, students can manipulate those videos
(i.e., play, forward or rewind) at their own pace and as many
times as they need while studying programming [5]. Third,
learners can watch videos wherever they are, either inside or
outside the classroom. Fourth, watching videos to learn about a
subject can be a very familiar activity for today’s learners, also
known as digital natives [22].
Since our approach is intended to teach programming to
novice learners, we have used videos in which the teacher
(expert programmer) performed only a sequence of extension
and refinement activities. While performing an extension
activity, the teacher described briefly the goal he wanted to
achieve. Straight after that, while performing a refinement
activity, he described how he implemented the piece of code
that achieved that goal. Therefore, while learning from these
videos students should achieve the intended learning outcomes
(ILOs) defined from the SI framework and presented in Table
Activity Intended learning outcomes
Extension Describe a goal that should be achieved in the use case
being ext ended.
Refinement Apply concepts of programming language to build an
implementation that matches the goal previously
However, because many students do not know how to study
properly for the needs of a programming course [21], the
structure of these videos alone is not sufficient to make them
learn. For learning to happen, it is necessary that learners
understand the process of building programs presented by these
videos. Thus, to help learners in understanding this material,
after each extension and refinement performed by the teacher,
students are invited to engage in a SE learning activity aligned
with the extension and refinement activities they have just seen
in the video [9]. This combination of watching the
implementation of small pieces of code and then self-
explaining the steps taken to implement those pieces while the
example is being built, is illustrated in Figure 2. This approach
was named stepwise self-explanations (SSE), because students
are explaining to themselves small pieces of code while they
are being implemented incrementally in the video.
Figure 2. Stepwise self-explana tion appr oach.
To illustrate this approach, we have chosen to use a small
set of The Joy of Code videos [8] as instructional material.
These videos teach Java programming language using the
Greenfoot tool, and were produced according to the SI
framework ideas. We present a screenshot of a video section of
The Joy of Code in Figure 1. In this video section, the teacher
performed an extension followed by a refinement activity.
Because of that, he described the goal he wanted to achieve
with the new use case he was covering with extension. Also, he
described how he built an implementation that matched that
goal. In this case, the goal was to make the turtle move, and the
sequence of steps performed to build an implementation that
matched that goal was to write the line move(1); inside the
void act() method. After watching this video section, students
were able to answer corresponding SE prompts related to the
activities performed. These SE prompts are presented in Table
2. Similar to [4], we have defined these SE prompts using the
five Ws and one H (who/what/where/when/why/how)
principles of questioning.
Figure 3. Screenshot o f the Jo y of co de vide o recording.
The research by Pirolli and Recker [13] and Bielaczyc,
Pirolli and Brown [4] also used examples as part of their
instructional material, but their examples were presented to
students statically and entirely on paper. Instead, the examples
used in our approach are presented in videos and, because of
that, they are essentially dynamic. Besides that, our approach
has adopted the SI framework to structure the instructional
material and the SE questions. To our knowledge, no other
previous work in the programming domain has studied SE
from examples presented in videos.
Activity Questions
Extension What is the main goal of the piece of code that he has
just written?
Refinement What were the steps/actions that he performed to
achieve the goal he intended to?
How did he write code to make the turtle move?
What is the purpose of the parameter (number ‘1’) in
the move method?
Where did he write code to make the turtle move?
When did he use the move method?
After he had compil ed and run the code, how was th e
turtle behaving?
V. S
The main goal of the present study was to verify whether
there was any benefit in using our approach of stepwise self-
explanations of examples presented in videos. To reach this
goal, we have compared our approach again st a traditional one,
in which students studied programming from videos without
any SE prompts to guide their study. Besides that, we have also
verified whether a relationship exists between the learning and
programming phases within the group that studied according to
our approach. Thus, from the present study, we answered the
following research questions:
[RQ1] Was the score of the final programming practice
exercise of the instructional group greater than that of the
control group?
[RQ2] Was there any correlation between the SE questions
answered correctly and the score of the final
progr ammin g practice exercise of the instructional group?
A. Participants
This study was carried out in a technical high school
located in Aarhus, Denmark in September 2014. Fourteen
second-year students of this school volunteered to participate in
the study. They were divided into two groups randomly. The
instructional group (SE group) studied videos using SE
prompts while the control group (SS group) studied videos
using their own way of studying (i.e., self-study).
B. Instructional material
As mentioned before, we used Java as a programming
language and Greenfoot as a development tool. Therefore, the
instructional materials used during this study were:
(i) episodes 1, 3 and 4 of The Joy of Code videos [9]. These
videos were edited to make clear the boundaries between
different sequences of extension and refinement activities;
(ii) the Hedgehogs, Turtle and Crab scenarios available on
The Joy of Code website;
(iii) initial and final programming exercises based on the
exercises from Intr oduction to Programming with
Greenfoot [20]; and
(iv) a set of SE questions.
C. Procedure
The study consisted of five phases: (i) pre-information
phase; (ii) information phase; (iii) learning phase; (iv)
programming phase; and (v) post-programming phase,
performed in two sessions that took place on two different
days. The first session was performed online by using a
webpage and was planned to last approximately 1.5 hour. This
first session consisted of the first two phases of the study, pre-
information and information phases, as shown in Figure 4. The
pre-information phase consisted of three activities. First, we
explained the experiment to the student and collected the
student’s consent (or their parents’ consent if they were under
18 years of age). The consent form was given previously to
participants and they had to bring it back to the classroom
session. Second, we collected demographic information about
participants and data about their previous experience when
studying from videos through an online questionnaire. A
computer lab was not available for running our study;
consequently we had to ask students to use their own
computers. For this reason, at the end of this pre-information
phase, we asked students to install in their computers
Greenfoot and a software to record their computers’ screens.
Figure 4. Online session.
The information phase con sisted of two activities. First,
participants had to watch sections of the episodes 1 and 3 of
The Joy of Code videos. The section of episode 1 presented
some examples of what students could build using Greenfoot
development tool and it was used to motivate students for
participating in the study. The section of the episode 3
presented some basic concepts of object oriented programming
in Java, such as classes, objects and methods, and it was used
to give students some prior knowledge in the subject. Students
were instructed not to watch these two videos more than twice
during this phase. After watching these videos, they had to do
an initial programming practice exercise. This exercise was
used as warm-up exercise and it aimed to assure that they could
(i) use Greenfoot, (ii) record th eir computers’ screens while
they were doing the exercise and, after finishing it, (iii) upload
the resulting video in a given Dropbox account.
The second session was performed at the participants
school and lasted 1.5 hour. It consisted of the last three phases
of the study; learning, programming and post-programming
phases, as showed in Figure 5. To this session, students were
randomly divided into two groups, SE group and SS group. To
do the activities in parallel, the groups were placed in different
classrooms and each student used their own computer and
headphone. To control the activities, each group had a different
mediator; the SE group was conducted by the first author of
this paper; the SS group was conducted by another researcher
from our research group at <omitted for submission> trained
for the mediation by the first author.
Figure 5. Classroom session.
During the learning phase, students had to watch a set of 4
sections of the episode 4 of The Joy of Code videos. Each
video section consisted of a sequence of extension and
refinement activities to teach the students how to make an
object act in the Greenfoot world. To make the boundaries
between two different sequences performed by the teacher,
these videos were cut after every refinement activity. After
watching each video section, students from both groups had the
same amount of time to study it. To collect data about how
students in the SE group self-explained each video section
using the corresponding SE prompts, students from the SE
group were instructed to study each video section by answering
the proposed questions on paper. Similarly, students from the
SS group had blank sheets of paper if they wanted to write
some notes while studying each video section. During this time
of study, students from both groups could watch the video
section one more time if they wanted to.
To compare the performance of both groups, during the
programming phase students were evaluated through a final
programming practice exercise. The time for doing this final
exercise and the practice exercise itself were identical for both
groups. To analyze the students’ dynamic process of
programming, they were instructed to record their computers
screens while they were answering the final exercise and, after
finishing it, to upload the resulting video in the Dropbox
account they had used previously.
Finally, at the end of this classroom session, students from
both groups answered a final questionnaire to evaluate their
experience of learning programming from videos in this study.
In addition, the SE group evaluated the proposed approach of
self-explaining these videos while learning programming.
The results we obtained from the data collected during the
online and classroom sessions are presented in the following
A. Data collected during online session
From the initial questionnaire, we have collected the
following data. The SE group had 7 participants, but one of
them had problems with his computer and his data was not
used. The remaining SE group students were between the ages
of 16 and 19 years old (mean = 17.33). They had no prior
programming experience. All of them, but one have used
videos from Youtube for learning some subject (e.g., they used
videos for learning how to use a software (n=3) and learning
how to make a game (n=3)). To study the mentioned subjects,
they have used videos in the ways depicted in Table III.
#students of
How have you used these videos? SE group SS group
I have paused, forwarded or r ewind ed in
order t o pay at tention in spe cific parts of
the video
5 3
I have watched more than one time the
same video
4 3
I have watched more than one time parts of
the same video
2 1
I have tried to reprodu ce what I was seein g
in the video
2 4
I have made notes while I was watching
the video
1 1
I have answered questions while I was
watchi ng the videos
1 1
I have simply watched the vi deo withou t
making any other activity
1 1
The SS group also had 7 students, but one of them had prior
programming experience and his data was not used. The
remaining SS group students were between the ages of 16 and
18 years old (mean = 17.17) and they had no prior
programming experience. Similar to SE group, all 6 students in
SS group, but one have used videos from Youtube for learning
some subject (e.g., they used videos for learning how to make
games (n=4)). To study the mentioned subjects, they have used
videos in the way depicted in the Table III.
The videos of students’ computer screens, recorded during
the online session showed that all of them could (i) use
Greenfoot and the software to record their computers’ screens
after installing them; (ii) use the Dropbox folder to upload the
resultant videos; and (iii) finish the initial programming
practi ce exercise. Then, all of them could use the tools needed
for the classroom session of the study.
B. Data collected during classroom session
To answer [RQ1], we examined written answers and videos
of students’ final programming practice exercise from both
groups. These data complement each other, so we looked at
them together to grade students’ practice exercises. The final
practice exercise had 5 questions and students’ grades could be
a maximum of 40 points. The score in the SE group ranged
from 25 to 37 points, with the median at 34.5 and a mean of
33.33 (SD = 4.32). The score in the SS group ranged from 29
to 35 points, with the median at 32 and a mean of 32 (SD =
2.76). However, we cannot confirm that the difference between
scores of SE group and scores of th e SS group is significant
(Mann Whitney’s U=11.5, z=-1.052; p=0.29) regarding the
final pr ogramming practice exercise. Consequently, our data
does not enable us to distinguish between the approaches used
in the SE gr oup and in the SS group respectively.
To answer [RQ2], we examined students’ SE written
answer s together with the scor es of SE group students’ final
programming practice exercise. Each answer to the SE
questions was graded in a maximum of 1 point (wrong answer
– 0; half-right answer – 0.5; correct answer – 1). We proposed
a set of 38 SE questions, so in total, students could have a
maximum score of 38 points. The scores for the SE answers
ranged from 16 to 26.5 points, with the median at 22 and a
mean of 21.36 (SD = 3.76). We found a significant correlation
(Spearman’s rho=0.899, p=0.007) between the scores obtained
from students’ SE answers during learning phase and the
scores of students’ final programming practice exercises
answer s during programming phase.
At the end of the study, we asked for students’ feedback
through a final questionnaire. When asked about their
experience studying from videos, SE group presented positive
feedback. Students found it easy (n=4), because they could re-
watch the videos in case of doubts (n=2) or reproduce the
content they learned from the videos using Greenfoot (n=1).
Other adjectives used were: interesting, great and fun. When
asked about using videos combined with SE questions, SE
group also presented positive feedback. They found it helpful
(n=4), because the questions helped them to remember things
(n=2), the questions were important to summarize the
information learned (n=1), and because they could monitor
what they had learned (n=1). Despite this positive feedback,
our approach of videos combined with SE questions was
considered very time-consuming (n=4).
Similarly, SS group also presented positive feedback when
asked about their experience. Students found to study from
videos was easy to understand the subject (n=4). In particular,
they said The Joy of Code videos wer e very good, simple and
more understandable than others they had watched previously
because the teacher presented the code as a simple step-by-step
guide and at an appropriate speed. When asked about the
activities they performed while studying from videos, they said
they wrote down on paper what they found important and hard
to remember (n=2).
C. Discussion
Examining the scores from students’ final practice
exercises in SE group and SS group, we could not confirm the
difference between the scores from the two groups was
significant. Therefore, the SE effect could not be replicated in
this study because of a combination of two reasons. First, the
content learned might have been too simple for the participants.
In general, studies that support the SE effect in programming
only looked at more complex content than those we have
investigated. However, due to time constraints, students had to
learn very simple tasks during its learning phase.
Consequently, students in both groups did not experience a
high level of difficulty in studying and solving associated
problems in the present study. This issue was confirmed with
students’ feedback about learning from videos.
Second, the instructional explanations presented in the
videos also might have contributed for this result. These
instructional explanations contained the answers to the SE
questions that the instructional group had to answer. Besides
that, we observed from SS group’s notes that some of them
were equivalent to answers to SE questions. For example, notes
from five students in SS group were related to the code
produced. These notes correspond to answers to the questions
related to a refinement activity in SE group. Thus, it seems that
students in SS group benefited from instructional explanations
presented in the videos and, in consequence, engaged in a
learning activity that was similar to that of SE group. As a
result, students in SS group had similar learning outcomes
when compared to the students in SE group.
Besides that, we found a significant positive correlation
between scores for the SE answers and scores of students’ final
programming practice exercise in the SE group. Our analysis
showed that almost 90% of the variation in the scores of SE
group final practice exercise can be explained by the variation
in the scores for their answers to SE questions. Thus, students
from SE group who answered more corr ectly SE questions in
learning phase had the best performance in programming
Besides that, students had a positive feedback regarding our
approach. Regarding the use of videos as instructional
materials in this study, students’ opinions from both groups
suggested they enjoyed it. Regarding the use of SE questions,
most of the students in the SE group found it helpful. Although
students of SE group had a good experience using our
approach, most of them considered the approach very time-
In this article, we presented our approach to learning
programming through stepwise self-explanations of examples
presented in videos. To evaluate our approach, we carried out
an experimental study with second-year students in a technical
high school in Denmark. We assumed that by learning
programming through our approach students in SE group
would outperform students in SS group. However, we could
not confirm that the difference between the two groups was
significant, and, as a consequence, we could not replicate the
SE effect in the context of this study. In our opinion, the lack of
effect happened because of these reasons: (i) the low level of
difficulty of the content chosen and (ii) the instructional
explanations presented in the videos.
Despite this result, the significant correlation between
learning and programming phases within SE group suggests the
proposed SE questions served as a useful starting point in our
approach. This result also suggests our approach should be
investigated in further detail, especially with regard to which
instructional conditions are more effective for it. Because of
that, as future work, we should (i) choose a more difficult
content in the programming domain to carry out a next
experiment and (ii) be concerned not only about how to phrase
SE questions but also about the most appropriate time to ask
those SE questions while students are watching videos.
[1] Robins, A.; Rountree, J.; Rountree, N. Learning and Teaching
Programming: A Review and Discussion, Computer Science Education,
v. 13, n. 2, p. 137-172, 2003.
[2] Chi, M. T. H.; Bassok, M.; Lewis, M.; Reimann, M., & Glaser, R. Self-
explanations: How stu dents study and use examples in learning to solve
problems. Cognitive Science, 13, 145-182, 1989.
[3] Chi, M. T. H; Deleeuw, N.; Chiu, M.; Lavancher, C. Eliciting self-
explanations improves understanding. Cognitive Science, 18, 439-477,
[4] Bielaczyc, K.; Pirolli, P.; Brown, A. L. Training i n self-explanation and
self-regulation strategies: Investigating the effects of knowledge
acquisition activitie s on problem solving. Cogniti on and Instruction, 13,
221-253, 1995.
[5] Bennedsen, J.; Carpersen, M. E. Exposing the Programming Process. In
Reflection on the Theory of Programming: Methods and
Implementation, J. Bennedsen, M. E. Carpersen, M. Kölling (Eds.)
Springer-Verlag, 6-16, 2008.
[6] Caspersen, M. E. Educating Novices in the Skills of Programming. PhD
dissertation PD-07-4, Department of Computer Science, University of
Aarhus, 2007.
[7] Casper sen, M. E.; Kölling, M. STREAM: A First Program ming Process.
Trans. Comput. Educ., v. 9, n. 1, p. 1-29, 2009.
[8] Kölling, M. The Joy of code [Video recordings]. 2012. Available:
[9] Biggs, J.; Tang, C. T eaching for Qua lity Lea rning a t University. New
York, N.Y.: McGraw-Hill/Society for Research into Higher
Education/Open University Press, 2011.
[10] Chiu, J. L.; Chi, M. T. H. Supporting self-explanation in the classroom.
In V.A. Benassi, C.E. Overson, & C.M. Hakala (Eds.). Applying science
of learning in education: Infusing psychological science into the
curriculum, 2014. Available:
[11] Clark, R. C.; Nguyen, F.; Sweller, J. Efficiency in Learning: Evidence-
Based Guidelines to Manage Cognitive Load. Wiley, 2005. ISBN 978-
[12] Berthold, K.; Renkl, A. Fostering the understanding of multi-
representational examples by self-expla nation prompts. In: BARA, B.
G.;BARSALOU, L., et al, Proceedings of the CogSci 2005, 2005.
Erlbaum. p.250-255.
[13] Pirolli, P.; Recker, M. Learning Strategies and Transfer in the Domain of
Programming. Cognition and Instruction, v. 12, n. 3, p. 235-275,
1994/09/01, 1994.
[14] Renkl, A. Learning from Worked-Out Examples: A Study on Individual
Differences. Cognitive Science, v.21, n.1, p.1-29, 1997.
[15] Crippen, K. J.; Earl, B. L. The impact of web-based worked examples
and self-explanation on performance, problem solving, and self-efficacy.
Comput. Educ., v. 49, n. 3, p. 809-821, 2007.
[16] Kwon, K.; Kumalasari, C. D.; Howland, J. L. Self-Explanation Prompts
on Problem-Solving Performance in an Interactive Learning
Environment. Journal of Interactive Online Learning, v. 10, n. 2, p. 96-
112, 2011.
[17] Yen-Chu, H. Combining Self-Explaining With Computer Architecture
Diagrams to Enhance the Learning of Assembly Language
Programming. Education, IEEE Transactions on, v. 55, n. 4, p. 546-551,
[18] Lee, M. J.; Ko, A. J.; Kwan, I. In-game assessments increase novice
programmers' engagement and level completion speed. Proceedings of
the ninth annual international ACM conference on International
computing education research. San Diego, San California, USA: ACM:
p. 153-160, 2013.
[19] Vihavainen, A.; Miller, C. S.; Settle, A. Benefits of Self-explanati on in
Introductory Programming. Proceedings of the 46th ACM Technical
Symposium on Computer Science Education. Kansas City, Missouri,
USA: ACM: p. 284-289, 2015.
[20] Kölling, M. Introduction to Programming with Greenfoot: Object-
orient ed Progra mming in Java with Games and Si mulati ons. Pre ntice
Hall, 2010. ISBN 9780136037538.
[21] Gomes, A., Mendes, A.J. Learning to program - difficulties and
solutions. International Conference on Engineering Education – ICEE
2007, Coimbra, Portugal.
[22] Prensky, M. R. Teaching Digital Natives: Partnering for Real Learning.
SAGE Publications, 2010. ISBN 9781452271408.
... Worked examples have been identified as an effective strategy that can help learners manage cognitive load. This instruc- tional strategy has been evaluated in several disciplines, including math (Carroll 1994), physics ( Chi et al. 1989), and computer programming ( Aureliano et al. 2016;Morrison et al. 2015;Vieira et al. 2015). Worked examples are especially useful for novice learners, who still may not have the required schemata to solve problems on their own. ...
... This study explores the use of in-code comments as a self-explanation strategy in the context of glass-box and black-box approaches for CSE. Students' self-explanations of programming code have been explored as written self-explanations from instructional videos ( Aureliano et al. 2016), as subgoal labeling within the examples ( Morrison et al. 2015), and as in-code comments ( Vieira et al. 2015) in an introductory programming course. A preliminary analysis of the effective charac- teristics students' explanations has also been previously presented by Vieira and colleagues ). ...
This article presents two case studies aimed at exploring the use of self-explanations in the context of computational science and engineering (CSE) education. The self-explanations were elicited as students’ in-code comments of a set of worked-examples, and the cases involved two different approaches to CSE education: glass box and black box. The glass-box approach corresponds to a programming course for materials science and engineering students that focuses on introducing programming concepts while solving disciplinary problems. The black-box approach involves the introduction of Python-based computational tools within a thermodynamics course to represent disciplinary phenomena. Two semesters of data collection for each case study allowed us to identify the effect of using in-code comments as a self-explanation strategy on students’ engagement with the worked-examples and students’ perceptions of these activities within each context. The results suggest that the use of in-code comments as a self-explanation strategy increased students’ awareness of the worked-examples while engaging with them. The students’ perceived uses of the in-code commenting activities include: understanding the example, making a connection between the programming code and the disciplinary problem, and becoming familiar with the programming language syntax, among others.
In this article, we propose the development of a context-sensitive tool for providing personalized 3I (informative, interactive and iterative) feedback to novice programmers during the programming problem solving process. To achieve this aim, we have carried out different research stages, where the first is to understand the provision of feedback for novice programmers, investing in carrying out a study in two perspectives, theoretical and experimental. Thus, this study was divided into three stages: systematic literature mapping, systematic literature review and an experiment. As one result of this study we organize the acquired knowledge and elaborate a Context-Aware Taxonomy for Feedback (TaFe). In addition, we designed a Conceptual Architecture Based on Multi-Agents and Computational Context, considering functional requirements identified in the knowledge represented in TaFe. For the next steps, we plan to validate TaFe and the architecture using appropriate methodological instruments. Finally, we intend to develop and validate a feedback solution based on a real problem identified in an experiment.
Conference Paper
Studies in educational psychology suggest that people learn better when visual learning materials are accompanied by audio explanations rather than textual ones. Research on how this modality effect applies to computing education is scarce and inconclusive. We explore whether modality of instruction affects learning from videos that use a series of example programs to explain how variables work in Python. Learners (n=186) were crowdsourced from the internet and randomized in three groups, who received explanations as audio, text, or both, respectively. We did not find significant differences between the groups in near transfer to code-tracing tasks or perceived cognitive load. The result affirms the need to further investigate instructional modalities in programming education. There are a number of theoretical, methodological, and instructional-design factors that may explain these and earlier findings; we trace out future research that could explore those factors.
Full-text available
Multiple representations in learning materials are usually em- ployed in order to foster understanding. However, they also impose high demands on the learners (e.g., need for integration). By embedding multi-representations in worked- out examples, cognitive capacity is released that can be used for self-explanations on the integration and understanding of multiple representations. The effects of two types of self- explanation prompts were investigated by conducting an experiment comprising three conditions (domain: mathematics). The learners ( N = 62) received either (1) self- explanation prompts, (2) self-explanation prompts in a scaffolding-fading procedure, or (3) no prompts. Both types of self-explanation prompts fostered procedural and conceptual knowledge. With respect to procedural knowledge, the different self-explanations prompts did not differ in their effectiveness. However, conceptual knowledge and especially knowledge indicating the integration of multiple representations was particularly fostered by scaffolded self- explanation prompts. Thus, for enhancing conceptual understanding, such self-explanation prompts should be provided because they scaffold the learners to reach their zone of proximal development.
One approach for helping students learn to program is the use of self-explanation assignments. In these assignments, students explain instructional materials using domain knowledge covered in the course. In this work, we describe a randomized experiment where students in an introductory programming course were given two kinds of self-explanation assignments. One randomly selected group worked on self-explanation assignments with supporting questions while the alternate group had the same self-explanation questions but no additional supporting exercises. The combined groups performed better on comparable test questions than students from the previous year, who did not use self-explanation questions. The group with supporting questions performed better than the group with no additional support. Based on our results and previous research on self explanation, we argue that embedding self-explanation questions into programming material is beneficial for students. Moreover, further gains are achieved from supporting questions that help focus their explanations.
This study examined the effects of self-explanation prompts on problem-solving performance. In total, 47 students were recruited and trained to debug web-program code in an online learning environment. Students in an open self-explanation group were asked to explain the problem cases to themselves, whereas a complete other-explanation group was provided with partial explanations and asked to complete them by choosing correct key-words. The results indicate that students in the open self-explanation condition (a) outperformed in a debugging task, (b) perceived higher confidence for their explanations, and (c) showed a strong positive relationship between the quality of their explanation and their performance. These results demonstrate the benefits of the open self-explanation prompts. Cognitive load of self-explanation and quality of explanation are discussed.
Conference Paper
Assessments have been shown to have positive effects on learning in compulsory educational settings. However, much less is known about their effects in discretionary learning settings, especially in computing education and educational games. We hypothesized that adding assessments to an educational computing game would provide extra opportunities for players to practice and correct misconceptions, thereby affecting their performance on subsequent levels and their motivation to continue playing. To test this, we designed a game called Gidget, in which players help a robot find and fix defects in programs that follow a mastery learning paradigm. Across two studies, we manipulated the inclusion of multiple choice and self-explanation assessment levels in the game, measuring their impact on engagement and level completion speed. In our first study, we found that including assessments caused learners to voluntarily play longer and complete more levels, suggesting increased engagement; in our second study, we found that including assessments caused learners to complete levels faster, suggesting increased understanding. These findings suggest that including assessments in a discretionary computing education game may be a key design strategy for improving informal learning of computing concepts.
This paper investigates the impact of combining self explaining (SE) with computer architecture diagrams to help novice students learn assembly language programming. Pre- and post-test scores for the experimental and control groups were compared and subjected to covariance (ANCOVA) statistical analysis. Results indicate that the SE-plus-diagram approach had a significant effect on student achievement and improved novice students' learning experience by increasing their interest and participation in the course and helping them to develop assembly programming skills.
Previous research has found positive correlations between particular strategies students use while studying to explain instructional materials to themselves and student performance on associated problem-solving tasks (Chi, Bassok, Lewis, Reimann, & Glaser, 1989; Pirolli & Bielaczyc, 1989; Pirolli & Recker, 1994). In the study reported here, we investigate the causal nature of this relation. This was accomplished by identifying a set of self-explanation and self-regulation strategies used by high-performance students in our earlier studies. We used strategy training to manipulate students' application of these strategies and examined the impact of their use on student explanations and performance. Twenty-four university students with no prior programming experience worked through a sequence of programming lessons. Following introductory lessons, participants received interventions involving explicit training in the strategies (instructional group) or received a similar set of interventions but no explicit training (control group). The instructional group showed significantly greater gains than the control group in the use of self-explanation and self-regulation strategies from the pre- to postinterventions lessons. Increased strategy application was accompanied by significantly greater performance gains. The results indicate that the particular self-explanation and self-regulation strategies used in training contribute to learning and problem-solving performance.