Content uploaded by Marcos Román-González
Author content
All content in this area was uploaded by Marcos Román-González on Oct 05, 2015
Content may be subject to copyright.
RED-Revista de Educación a Distancia.Número 4615-Sep-2015 http://www.um.es/ead/red/46
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and
Foster Computational Thinking
Dr. Scratch: Análisis Automático de Proyectos Scratch para Evaluar y
Fomentar el Pensamiento Computacional
Jesús Moreno-León
Programamos.es. Spain.
jesus.moreno@programamos.es
Gregorio Robles
Universidad Rey Juan Carlos. Spain.
grex@gsyc.urjc.es
Marcos Román-González
Universidad Nacional de Educación a Distancia (UNED). Spain.
mroman@edu.uned.es
Abstract
One of the barriers to entry of computer programming in schools is the lack of tools that
support educators in the assessment of student projects. In order to amend this situation this
paper presents Dr. Scratch, a web application that allows teachers and students to
automatically analyze projects coded in Scratch, the most used programming language in
primary and secondary education worldwide, to check if they have been properly
programmed, learn from their mistakes and get feedback to improve their code and develop
their Computational Thinking (CT) skills. One of the goals of Dr. Scratch, besides
supporting teachers in the evaluation tasks, is to act as a stimulus to encourage students to
keep on improving their programming skills. Aiming to check its effectiveness regarding
this objective, workshops with students in the range from 10 to 14 years were run in 8
schools, in which over 100 learners analyzed one of their Scratch projects with Dr. Scratch,
read the information displayed as feedback by Dr. Scratch, and tried to improve their
projects using the guidelines and tips offered by the tool. Our results show that at the end of
the workshop, students increased their CT score and, consequently, improved their coding
skills.
Keywords
Computational thinking, learning, coding, Scratch, assessment
Resumen
Una de las barreras de entrada de la programación informática en las escuelas es la falta de
herramientas que ayuden al profesorado en la evaluación de los proyectos del alumnado.
Con el objetivo de resolver esta situación, este artículo presenta Dr. Scratch, una aplicación
web que permite a educadores y alumnos analizar automáticamente proyectos Scratch, el
lenguaje de programación más utilizado globalmente en educación primaria y secundaria,
para comprobar si se han programado correctamente, aprender de sus errores y recibir
retroalimentación para mejorar su código y desarrollar el Pensamiento Computacional (PC).
Uno de los objetivos de Dr. Scratch, además de ayudar al docente en las tareas de
evaluación, es ser un estímulo para animar a los aprendices a seguir mejorando sus
habilidades de programación. Para comprobar la efectividad de la herramienta en este
sentido, se organizaron talleres en 8 colegios con alumnos de entre 10 y 14 años en los que
los estudiantes analizaron uno de sus proyectos Scratch con Dr. Scratch, leyeron la
información del informe de resultados e intentaron mejorar sus proyectos usando los
consejos ofrecidos por la herramienta. Al finalizar el taller los alumnos mejoraron su
puntuación de PC así como sus habilidades como programadores.
Palabras clave
Pensamiento computacional, aprendizaje, programación, Scratch, evaluación
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 2 de 23
Introduction
In the last decade we have witnessed a resurgence of programming and Computational
Thinking (Wing, 2006) (CT) in schools (Lye & Koh, 2014). The educational use of
coding, which had been introduced in the 70s and 80s mainly with the Logo
programming language (Papert & Solomon, 1971), has come back strong due to new
visual programming languages, like Alice, Kodu and especially Scratch, which allow
young students to program applications without the need to learn the complex syntax of
traditional programming languages.
Scratch (Resnick et al., 2009) is a visual programming environment designed for
children over 6 years old, which also offers a website where users can share their
projects and exchange ideas or suggestions with other (young) programmers. Scratch is
massively used all over the world, with more than seven million registered users and
more than ten million shared projects in the repository
1
. One of the main goals of
Scratch is that programming becomes an educational tool to develop other skills and to
improve learning of other disciplines (Resnick, 2013). As a result, Scratch is being used
both in extracurricular activities (Kafai, Fields, & Burke, 2012) and in all levels of
formal educational environments, both in schools (Moreno-León & Robles, 2015), high
schools (Meerbaum-Salant, Armoni, & Ben-Ari, 2013) and even universities (Malan &
Leitner, 2007) worldwide.
Nevertheless, there is a lack of tools that support educators when evaluating student
programs and to assess the development of CT. This situation is partly caused by the
fact that there is a lack of agreement in a definition of the CT concept, and in the way it
should be included in the curriculum (Grover & Pea, 2013). This paper presents Dr.
Scratch, a free/open source web tool that analyzes Scratch projects to (1) offer feedback
to educators and learners and (2) assign a CT score to the projects. Learners can use this
feedback to improve their programs, but also can realize how to improve their
programming abilities. To test the effectiveness of Dr. Scratch, we have set up a set of
workshops to measure the impact of its use on learning. Results show positive results
and hint areas of future development.
The paper is structured as follows: the Background section reviews different proposals
and tools that try to assist educators in the assessment of the CT of students; then the
features included in Dr. Scratch are explained; the approach followed in preparing the
workshops to test Dr. Scratch with programming learners is detailed in the Methodology
section; the results of the workshops, both quantitative and qualitative, are shown in the
Findings section; finally, in the Conclusions we summarize our study, discuss the
limitations of the tool and present some new features the development team is working
on.
Background
There is lack of tools that support educators in the assessment of the development of CT
1
See http://scratch.mit.edu/statistics/
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 3 de 23
and the evaluation of projects programmed by students. Regarding the Scratch
programming language, several authors have proposed different approaches to evaluate
the development of CT of learners by analyzing their projects, but most of these
approaches are based exclusively on a manual analysis.
Wilson, Hainey, and Connolly (2012) suggest a scheme to gauge the level of
programming competence demonstrated by a student by analyzing a project in terms of
programming concepts (such as threads, conditional statements or variables), code
organization (variable names, sprite names and extraneous blocks) and designing for
usability (like functionality, instructions or originality, among others).
In this line, Seiter and Foreman (2013) developed the school Progression of Early
Computational Thinking Model, a framework to assess CT in primary students
programming with Scratch by synthesizing “measurable evidence from student work
with broader, more abstract coding design patterns, which are then mapped onto
computational thinking concepts” (Seiter & Foreman, 2013, p. 59)
In the paper “New frameworks for studying and assessing the development of
computational thinking”, Brennan and Resnick (2012) introduced a strategy based on
project portfolio analysis using a visualization tool called Scrape (Wolz, Hallberg, &
Taylor, 2011), which seems to be no longer available, although their proposal is
completed with artifact-based interviews and design scenarios.
Aiming to assist educators with a tool that could be used to partly automate the
assessment of Scratch projects, Boe et al. (2013) developed Hairball, a static code
analyzer that detects potential issues in the programs, such as code that is never
executed, messages that no object receives or attributes not correctly initialized. After a
two-week Scratch-based summer camp, Hairball was used to assess Computer Science
learning in terms of event-driven programming, initialization of state and message
passing (Franklin et al., 2013).
The Hairball architecture, based on plug-ins, is ideal to add new features. In a previous
work, the authors developed two plug-ins to detect two bad programming habits we
frequently detect in our work as instructors with high school students (Moreno &
Robles, 2014):
convention.SpriteNaming
2
analyzes a Scratch project to check if the names of the
sprites begin with the string Sprite, which indicates that the programmer has not
modified the default name that Scratch assigns to an object. It should be noted
that while using the default name for sprites produces no error in the program if
its implementation is correct, it makes the readibility of the program more
difficult, especially when the number of sprites is high (i.e., more than ten).
duplicate.DuplicateScripts
3
analyzes a Scratch project to find duplicate scripts,
which are repeated programs within a project. For such type of structures,
Scratch custom blocks should be used.
In order to check if these bad programming habits are also common in the projects
2
https://github.com/jemole/hairball/blob/master/hairball/plugins/convention.py
3
https://github.com/jemole/hairball/blob/master/hairball/plugins/duplicate.py
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 4 de 23
shared in the Scratch website, we randomly downloaded and analyzed 100 projects,
detecting that 79% of the inspected projects presented not personalized object names,
while 62% included repeated code (Moreno & Robles, 2014). These figures encouraged
us to develop a tool to help both learners and educators to detect issues in the code to
improve their programming skills.
The fact that Hairball is executed from the command-line, as it is based on Python
scripts that the evaluator has to manually launch, makes it not suitable for many
educators that are not confident with such an environment, let alone for young students.
For this reason, we decided to create a web-based service, Dr. Scratch, that allows the
analysis of Scratch projects easily.
Introducing Dr. Scratch
Dr. Scratch
4
is a free/open-source web application that allows to easily analyze Scratch
projects using Hairball plug-ins, as well as to obtain feedback that can be used to
improve programming skills and develop CT. To analyze a project with Dr. Scratch an
.sb or an .sb2 file can be uploaded, as the tool supports both 1.4 and 2.0 Scratch
versions, or the users can directly copy the URL of the project. The ability to analyze
projects from the URL has been implemented using getsb2
5
.
When a Scratch project is analyzed, Dr. Scratch informs the user of the degree of
development of CT demonstrated in that project, assigning a CT score. Being based on
Hairball, Dr. Scratch detects certain bad habits of programming or potential errors, such
as non-significant sprite names, repetition of code, code that is never executed and the
incorrect initialization of object attributes.
In order to assign the CT Score, Dr. Scratch infers the competence demonstrated by the
developer on the following seven concepts: abstraction and problem decomposition,
logical thinking, synchronization, parallelism, algorithmic notions of flow control, user
interactivity and data representation. The evaluation of the competence level of each of
these concepts follows the rules in Table 1, which was designed based on the proposals
presented in the Background section by remixing some of their ideas with the support of
educators from different educational levels who use Scratch in their classrooms.
Depending on the CT score, which may range from 0 to 21 points, distinct data is
displayed in the results page. Thus, if the CT level is low it is assumed that the user is a
novice programmer and, consequently, the tool will only show basic information of the
most important improvements to perform in the code. As the score increases, Dr.
Scratch will show more information of the analyzed projects. Thus, advanced users
receive a feedback report with all available information both in terms of CT skills and
bad programming habits. Figures 1 and 2 illustrate the differences in the quantity and
complexity of the information displayed on screen depending on the CT Score.
4
http://drscratch.org/
5
https://github.com/nathan/getsb2
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 5 de 23
Fig. 1. Dr. Scratch analysis results for a project with basic CT Score
Fig. 2. Dr. Scratch analysis results for a project with advanced CT Score
Figure 3 can be used to illustrate the operation of the CT assessment. Thus, following
the rules in Table 1, the first script of the picture would be cataloged as basic in terms of
data representation, as it modifies some of the object attributes (position and
orientation). The second script, however, would be considered to demonstrate a
developing level, because a variable is utilized. Finally, the third script would prove a
proficient level on this concept, as an operation on lists is performed.
In those aspects where there is room for improvement, the tool provides links to
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 6 de 23
information that can be used to improve. For example, if a project has been awarded
with one point in parallelism, Dr. Scratch provides a link to sample source code and an
explanation of how to perform several actions at the same time in a program (see Figure
4).
CT Concept
Competence Level
Null
(0)
Basic
(1 point)
Developing
(2 points)
Proficiency
(3 points)
Abstraction and
problem
decomposition
-
More than one
script and more
than one sprite
Definition of blocks
Use of clones
Parallelism
-
Two scripts on
green flag
Two scripts on key
pressed, two scripts on
sprite clicked on the
same sprite
Two scripts on when I receive
message, create clone, two
scripts when %s is > %s, two
scripts on when backdrop
change to
Logical thinking
-
If
If else
Logic operations
Synchronization
-
Wait
Broadcast, when I
receive message, stop
all, stop program, stop
programs sprite
Wait until, when backdrop
change to, broadcast and wait
Flow control
-
Sequence of
blocks
Repeat, forever
Repeat until
User Interactivity
-
Green flag
Key pressed, sprite
clicked, ask and wait,
mouse blocks
When %s is >%s, video,
audio
Data representation
-
Modifiers of
sprites
properties
Operations on
variables
Operations on lists
Table 1. Competence Level for each CT concept.
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 7 de 23
Fig. 3. Different competence levels of data representation: basic (top), developing
(center) and proficient (bottom).
Fig. 4. Ideas and tips provided by Dr. Scratch to improve projects by incorporating
parallelism.
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 8 de 23
Methodology
In order to assess the effectiveness of Dr. Scratch as a tool to assist programming
learners, we performed a number of workshops with students between 10 and 14 years
from eight Spanish primary and secondary schools (see Figure 5). These students had
previously learned to program with Scratch for several weeks in their schools.
During the one-hour workshop, students were given a questionnaire with some
questions they had to answer while performing different tasks. The tasks and questions
were as follows:
1) Visit the Dr. Scratch website:
a) What do you think about the website? Do you find it attractive?
b) After reading the information on the website, what do you think Dr. Scratch can
be used for?
2) Analyze one of your Scratch projects with Dr. Scratch.
a) Is it easy to analyze projects with Dr. Scratch?
b) What was your score?
c) According to Dr. Scratch, what is the CT level for that score?
d) How did you feel when you saw the results?
e) Why?
3) From the results page, after analyzing a project, clickon some of the links to
receive information that could help you improve your code.
a) Write the title of the page you clicked on.
b) Do you understand the information in theresults page?
c) After reading the information, do you feel like trying something new?
4) Using the information that appeared in the help page you selected, try to improve
your project by adding something new.
a) Are the ideas and tips in the results page enough to improve your program?
b) After performing some modifications, analyze again your project with Dr.
Scratch. What is the new score?
5) Do you have any other comments?
Fig. 5. Dr. Scratch workshop at Lope de Vega Primary School, Madrid.
Characteristics of the study sample
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 9 de 23
Table 2 shows the characteristics of the study sample regarding the age of the
participants, formed by a group of 109 students between 10 and 14 years. The mean age
of the sample was 11.50, while the median was 11 and the mode was 10.
N
Valid
109
Missing
0
Mean
11.50
Median
11.00
Mode
10
Standard Deviation
1.392
Variance
1.937
Minimum
10
Maximum
14
Table 2. Age of students participating in the investigation
Figure 6 shows the percentage of participating students for each age group. As can be
seen, a majority of participants were 10 or 11 years old.
Figure 6. Percentage of students by age group (in years old)
In terms of gender, Table 3 shows the percentage of boys and girls participating in the
investigation; 57.8% of participants were boys while 42.2% were girls. This difference
can be explained by the fact that in some of the schools the experiment was carried out
in non-compulsory subjects, such as ICT or technology, where there usually is a
majority of boys. On the other hand, the gender of the students was not recorded in one
of the participating schools, which explains the 19 missing records on Table 3.
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 10 de 23
Although there has been significant research into gender issues in computer science
education (Beyer, Rynes, Perrault, Hay & Haller, 2003) and programming (Carter &
Jenkins, 1999), that topic is out of the scope of this study. Nonetheless, future research
on if boys and girls react to and/or learn different with Dr. Scratch is planned.
Frequency
Percentage
Valid Percentage
Cumulative Percentage
Valid
Boy
52
47.7
57.8
57.8
Girl
38
34.9
42.2
100.0
Total
90
82.6
100.0
Missing
System
19
17.4
Total
109
100.0
Table 3. Percentage of students by gender
Finally, regarding the educational stage of the students, as shown in Table 4, a majority
of participants was enrolled in Primary Education, although we also had a significant
group of students from Secondary Education.
Frequency
Percentage
Valid Percentage
Cumulative Percentage
Valid
Primary Education
75
68.8
68.8
68.8
Secondary Education
34
31.2
31.2
100.0
Total
109
100.0
100.0
Table 4. Percentage of students by educational stage
Findings
Figures 7, 8, 9, 10 and 11 show the answers of the students to some of the questions of
the questionnaire. According to the responses, a majority of the students found the Dr.
Scratch website attractive, as can be seen in Figure 7, and most of students believed that
analyzing projects with Dr. Scratch was easy (Figure 8). Regarding their feelings after
analyzing their projects, shown in Figure 9, a majority of learners felt good when they
saw the CT score, although 3% of the respondents indicated that they felt bad. In regard
to the information displayed by Dr. Scratch, Figure 10 shows that most of the students
were able to understand it; however, 5.6% of the learners answered that they did not
understand the information obtained. Finally, being one of the goals of Dr. Scratch to
stimulate self-learning by offering a gamifyied environment, we were interested in the
response of students after obtaining feedback. In this sense, Figure 11 shows that Dr.
Scratch boosts the willingness to improve programming skills.
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 11 de 23
Figure 7. What do you think about the website? Do you find it attractive?
Figure 8. Analyze one of your Scratch projects with Dr. Scratch. Is it easy to analyze
projects with Dr. Scratch?
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 12 de 23
Figure 9. How did you feel when you saw the results?
Figure 10. Do you understand the information in the results page?
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 13 de 23
Figure 11. After reading the information, do you feel like trying something new?
Table 5, Figures 12 and 13 show the results of the analysis of the Scratch projects
developed by the students before and after reading the feedback reports by Dr. Scratch.
As it can be seen, there was an increase in the results, as the mean of the pre-test
analysis is 11.82, while the mean of the post analysis is 13.52.
Before Dr. Scratch Feedback (Pre-test)
After Dr. Scratch Feedback (Post-test)
Mean
11,82
13.52
Median
12
14
Mode
[11, 12, 15]
16
Std. Deviation
3.093
3.257
Skewness
.028
-.171
Minimum
5
5
Maximum
20
21
Percentiles
10
8
9
20
9
10.40
30
10
11
40
11
13
50
12
14
60
13
15
70
14
16
80
15
16
90
15.60
17
Table 5. Results of the projects’ analysis before and after reading feedback by Dr.
Scratch.
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 14 de 23
Figure 12. What was your score? (Pre-test scores)
Figure 13. After performing some modifications, analyze again your project with Dr.
Scratch. What is the new score? (Post-test scores)
To be able to prove that the improvement experienced by the students was statistically
significant, we performed a t-test for paired samples, establishing a 95% confidence
level (α = 0.05) for our statistical decisions.
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 15 de 23
Table 6 shows the statistics of the 88 students who correctly indicated the pre and post
analysis results of their projects. There were 21 not completed records out of the 109
participating students, because pre-test or post-test scores were not correctly specified
by students. Several circumstances apply here: from students who specify just their CT
level (low, medium, high) instead of their CT score, to Internet connection problems or
an error in Dr. Scratch. For the 88 complete records, the mean value of CT score
increased from 12.00 in the pre-test to 13.45 in the post-test.
Paired Samples Statistics
Mean
N
Std. Deviation
Std. Error Mean
Pair 1
Pre-test Score
12.00
88
2.983
.318
Post-test Score
13.45
88
3.216
.343
Table 6. Statistics of paired data
Paired Samples Test
Paired Differences
t
df
Sig.
(2-tailed)
Mean
Std.
Deviation
Std. Error
Mean
95% Confidence
Interval of the
Difference
Lower
Upper
Pair
1
Pretest Score -
Postest Score
-1.455
1.523
.162
-1.777
-1.132
-8.959
87
.000
Table 7. t-test for paired data in the pre-test and post-test
The results of the t-test for paired data are shown in Table 7. As p(t) = 0.000 << 0.05,
the null hypothesis of equality of means is rejected and we can therefore state that there
are significant differences between the pre-test and post-test, which indicates that the
use of Dr. Scratch helped the participating students develop their CT.
Aiming to assess the real impact of using Dr. Scratch on the development of CT, the
effect size of the experiment is considered to be a good indicator as it is independent of
sample size, and can be calculated according to the following formula (Cohen, 1990):
For our investigation the effect size was 0.47, which is considered as a moderate effect
(Ellis, 2010). Nevertheless it must be noted that this effect was experienced after just 1
hour of treatment, consisting in the workshop, which highlights the impact that Dr.
Scratch had on the learners and draws attention to its potential as a tool to foster CT by
using it as a supporting tool in a programming course.
As expected, there is a positive and very significant correlation between pre-test and
post-test scores, which is shown in Table 8.
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 16 de 23
Paired Samples Correlations
N
Correlation
Sig.
Pair 1
Pretest Score * Postest Score
88
.882
.000
Table 8. Correlations of paired data
Figure 14 shows the scatter plot for the sample when comparing the pre-test and post-
test scores. As can be seen, all cases fall into the improvement area, which means that
the scores in the post-test were equal or bigger than the pre-test scores for all of the 88
students; therefore, none of the learners decreased his/her score during the experiment.
Figure 14. Scatter plot comparing post-test and pre-tests scores. Cases when the post-
test score is equal or bigger than the pre-test score fall into the improvement area. Cases
when the post-test score is less than pre-test score fall into the no improvement area.
We performed several significance analysis of the differences of pretest-postest
depending on several factors. In the first place, the significance analysis was performed
based on the initial score. Thus, we divided the sample in three sub-samples depending
on the pre-test score:
Sub-sample “Low initial score”: pre-test scores ≤ 10 (blue color in Figure 14)
Sub-sample “Medium initial score”: pre-test scores ≤ 15 (red color in Figure 14)
Sub-sample “High initial score”: pre-test scores ≤ 21 (green color in Figure 14)
Table 9 shows the significance and correlation values both for the total sample and for
each sub-sample. The results indicate that the use of Dr. Scratch generated significant
improvements in the development of CT in the “medium initial score” sub-sample (11-
15) and in the “low initial score” sub-sample (0-10), although the improvement is
slightly lower in the latter case. However, Dr. Scratch did not generate a significant
Improvement area
No improvement area
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 17 de 23
improvement in the “high initial score” sub-sample (16-21). These results are, to some
degree, in line with what was expected, as they indicate that students with basic and
medium initial levels, where there is more room for improvement, are able to make use
of the information provided by Dr. Scratch to enhance their score in just one hour, while
the tool seem less useful for those students with a high initial score where improvements
are harder to achieve. The differences between the initial and medium levels were also
relatively expected, as novice learners have to struggle with the difficulties of the early
steps in Scratch and the feedback provided by the tool. Nonetheless, Dr. Scratch offers
CT-dependent feedback as in first tests with learners the authors noted that too much
information in early phases were counterproductive. Future research will help us modify
and adapt the feedback reports in this direction.
N
Post-Pre
difference
t
p (t)
Significant difference
95% confidence?
r
p (r)
Full sample
88
1.455
8.959
0.000 < 0.05
Yes
0.88
0.000
Sub-sample
Low IL
25
1.440
6.896
0.000 < 0.05
Yes
0.89
0.000
Sub-sample
Medium IL
55
1.618
6.951
0.000 < 0.05
Yes
0.61
0.000
Sub-sample
High IL
8
0.375
1.426
0.197 > 0.05
No
0.90
0.002
Table 9. Significance and correlation values depending on pre-test score
Aiming to check if the use of Dr. Scratch had a different effect on primary and
secondary students, we performed the significance analysis on the differences of pretest-
posttest based on the educational stage. Table 10 shows the significance, effect size and
correlation values, both for the total sample and for each sub-sample (Primary and
Secondary education). As can be seen, both in primary and secondary education the use
of Dr. Scratch generated a significant improvement, although the improvement was
bigger in the latter stage.
N
Post-pre
difference
t
p (t)
Significant
difference
95%
confidence?
r
p (r)
d
Effect
Size
Full
sample
88
1.455
8.959
0.000 <
0.05
Yes
0.88
0.000
0.47
Moderate
Sub-
sample
Primary
64
1.234
7.294
0.000 <
0.05
Yes
0.90
0.000
0.40
Moderate
(-)
Sub-
sample
Secondary
24
2.042
5.540
0.000 <
0.05
Yes
0.87
0.000
0.60
Moderate
(+)
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 18 de 23
Table 10. Significance and correlation values depending on education stage
The analysis of covariance, shown in Table 11, confirms the significant effect of the
educational stage on the post-test controlling the baseline differences in the pre-test, as p
(F Educational Stage) = 0.03 < 0.05
Tests of Between-Subjects Effects
Dependent Variable: Post-test Score
Source
Type III Sum of Squares
df
Mean Square
F
Sig.
Corrected Model
710.667a
2
355.333
159.678
.000
Interception
22.806
1
22.806
10.248
.002
Pre-test Score
70.541
1
708.541
318.401
.000
Educational Stage
10.801
1
10.801
4.854
.030
Error
189.151
85
2.225
Total
16830.000
88
Corrected Total
899.818
87
a. R Squared = .790 (Adjusted R Squared = .785)
Table 11. Analysis of covariance. Dependent variable: post-test score.
Figure 15 can be used to illustrate the bigger improvement of secondary students. As
can be seen, there is a bigger density of cases with the same pretest-posttest score
(students do not improve, although they do not decline either after using Dr. Scratch) in
Primary Education, while in Secondary Education the bigger density is translated to the
improvement area. These results encourage us to plan future research in order to
determine whether the differences between primary and secondary students are due to
our tool, because of the language used and the examples included in the feedback report,
or are related to the maturation meta-cognitive development that, in consequence,
improves with age.
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 19 de 23
Figure 15. Scatter plot. Pretest Score * Postest Score depending on the Educational
Stage
Finally, a Spearman’s Rho was calculated to detect correlation between the data-
responses measured in an ordinal level, which are the ones corresponding to the
following questions of the questionnaire:
Attractiveness of the website. What do you think about the website? Do you find
it attractive? (1=Not attractive; 2=Just fair; 3=Yes, attractive)
Easiness of analysis. Is it easy to analyze projects with Dr. Scratch? (1=No;
2=Just fair; 3=Yes)
Feelings after results. How did you feel when you saw the results? (1=Bad;
2=Something in between; 3=Good)
Readability of results. Do you understand the information in the results page?
(1=No; 2=Not completely; 3=Yes)
Willingness of improvement. After reading the information, do you feel like
trying something new? (1=No; 2=Maybe; 3=Yes)
Feedback sufficiency. Using the information that appeared in the help page you
selected, try to improve your project by adding something new. Are the ideas and
tips in the results pages enough to improve your program? (1=No; 2=Just fair;
3=Yes)
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 20 de 23
Easiness
Feelings
Readability
Willingness
Feedback
Rho of
Spearman
Attractiveness
.241*
.063
.208*
.195*
.172
Easiness
.324**
.005
-.015
-.016
Feelings
.071
-.086
.120
Readability
.005
.294**
Willingness
.211*
Table 12. Spearman’s Rho correlations for ordinal responses
As shown in Table 12, some significant correlations were found:
The attractiveness of the web is correlated positively and significantly, albeit
with low intensity, with the perception of easiness of analysis, the readability of
the results and subsequent motivation to improve.
The perception of easiness of analysis correlates positively and significantly,
although with low to moderate intensity, with good feelings after receiving the
results.
The readability of the results correlates positively and significantly, albeit with
low to moderate intensity, with the perception of sufficiency of feedback.
Finally, the motivation to improve is significantly and positively correlated,
although with low intensity, with the perception of sufficiency of feedback.
Conclusions and future work
This paper presents Dr. Scratch, a free/open-source web tool that allows analyzing
Scratch projects to automatically assign a CT score as well as to detect potential errors
or bad programming habits, aiming to help learners to develop their coding and CT
skills as well as to support educators in the evaluation tasks.
In order to assess the effectiveness of Dr. Scratch as a tool to assist programming
learners, we run a series of workshops with 109 students between 10 and 14 years from
8 different schools that had prior coding experience with Scratch. The students analyzed
one of their Scratch projects with Dr. Scratch, read the information in the feedback
report provided by the tool, tried to improve their code following the instructions and
finally analyzed again their projects. The results show that, in average, students
enhanced their CT Score in 1.45 points, from 12.00/21 to 13.45/21, which represents a
statistically significant improvement. In this line, the overall effect size, d = 0.47,
indicates a moderate effect that, taking into account that was generated during a one-
hour workshop, highlights the real impact that the use of Dr. Scratch had in the coding
skills and the development of CT of participants.
The results indicate that the feedback report provided by Dr. Scratch was especially
useful for secondary students with an initial medium (developing) CT score. However,
the tool does not seem to be as helpful for students with an initial high (proficient)
score, at least in the tested, one-hour workshop environment. We will devote future
work to ascertain how to enhance the feedback provided by the tool. In addition, new
research could help us discover if differences of performance between secondary and
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 21 de 23
primary students are due to the tool itself or are related to the maturation meta-cognitive
development of learners.
In the near future we also plan to perform new investigations to try to find correlations
with other tools that assess CT of students, such as the Computational Thinking Test
(Román-González, 2015), as well as to test the effectiveness of the Dr. Scratch
assessment comparing its results with the ones from Primary and Secondary Education
expert evaluators. These results would help us adjust and improve the CT analysis
operation.
Regarding the features of Dr. Scratch, at the time of writing this paper we are working
on several enhancements:
Plug-ins for browsers: with the plug-ins for Firefox and Chrome, learners will be
able to analyze their projects while programming in the Scratch website.
User accounts: students will be able to keep the log of their analysis to study
their evolution on time.
Teacher accounts: educators will be able to group and follow their students, and
to keep track of their progress.
Translation into new languages: at this moment the tool is available in Spanish
and English, but we plan to increase the number of languages with the support of
the community.
Gamification and Social network functionalities: we plan to incorporate new
features that allow users to communicate with each other to exchange ideas and
challenges.
‘App’ for mobile phones: we are developing an HTML5 ‘app’ to expand the
social network and gamification features.
Acknowledgements
The work of all authors has been funded in part by the Region of Madrid under project
“eMadrid - Investigación y Desarrollo de tecnologías para el e-learning en la
Comunidad de Madrid” (S2013/ICE-2715). The authors are very thankful to the
teachers and pupils of the schools and middle schools participating in the investigation:
Lope de Vega, Carles Salvador, Miguel de Cervantes, Villa Devoto, Camp de Túria,
Mestre Ramón Esteve, Vicente Aleixandre and Fuente San Luis. We would like to
thank as well Eva Hu Garres and Mari Luz Aguado for their technical support with Dr.
Scratch.
Sent: 31 July 2015
Accepted: 2 September 2015
Published: 15 September 2015
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 22 de 23
Moreno-León, J., Robles G., & Román-González, M. (2015). Dr. Scratch: Automatic
Analysis of Scratch Projects to Assess and Foster Computational Thinking.
RED-Revista de Educación a Distancia. Número 46. 15 de Septiembre de 2015.
Consultado el (dd/mm/aaaa) en http://www.um.es/ead/red/46
References
Beyer, S., Rynes, K., Perrault, J., Hay, K., & Haller, S. (2003). Gender differences in
computer science students. SIGCSE Bull., 35(1), 49. doi:10.1145/792548.611930
Boe, B., Hill, C., Len, M., Dreschler, G., Conrad, P., & Franklin, D. (2013). Hairball:
Lint-inspired static analysis of scratch projects. Proceedings of the 44th ACM
Technical Symposium on Computer Science Education (SIGCSE 2013), 215-220.
Brennan, K., & Resnick, M. (2012). New frameworks for studying and assessing the
development of computational thinking. Proceedings of the 2012 Annual Meeting of
the American Educational Research Association (AERA 2012)
Carter, J., & Jenkins, T. (1999). Gender and programming. SIGCSE Bull., 31(3), 1-4.
doi:10.1145/384267.305824
Cohen, J. (1990). Things I have learned (so far). American Psychologist, 45(12), 1304-
1312. doi:10.1037/0003-066x.45.12.1304
Ellis, P. (2010). The essential guide to effect sizes. Cambridge: Cambridge University
Press.
Franklin, D., Conrad, P., Boe, B., Nilsen, K., Hill, C., Len, M., ... Kiefer, B. (2013).
Assessment of computer science learning in a scratch-based outreach
program. Proceedings of the 44th ACM Technical Symposium on Computer Science
Education (SIGCSE 2013), 371-376.
Grover, S., & Pea, R. (2013). Computational thinking in K–12. A review of the state of
the field. Educational Researcher, 42(1), 38-43.
Kafai, Y. B., Fields, D. A., & Burke, W. Q. (2012). Entering the Clubhouse: Case
studies of young programmers joining the online Scratch communities. In A.
Dwivedi & S. Clarke (Eds.), End-User Computing, Development, and Software
Engineering: New Challenges (pp. 279-294). Pennsylvania, PA: IGI Global.
Lye, S. Y., & Koh, J. H. L. (2014). Review on teaching and learning of computational
thinking through programming: What is next for K-12? Computers in Human
Behavior, 41, 51-61.
Malan, D. J., & Leitner, H. H. (2007). Scratch for budding computer scientists. ACM
SIGCSE Bulletin, 39(1) 223-227.
Meerbaum-Salant, O., Armoni, M., & Ben-Ari, M. (2013). Learning computer science
concepts with scratch. Computer Science Education, 23(3), 239-264.
RED. Revista de Educación a Distancia. Número 46 15-Sep-2015 http://www.um.es/ead/red/46
_____________________________________________________________________________________
Dr. Scratch: Automatic Analysis of Scratch Projects to Assess and Foster Computational Thinking. Jesús
Moreno-León, Gregorio Robles and Marcos Román-González. Página 23 de 23
Moreno, J., & Robles, G. (2014). Automatic detection of bad programming habits in
scratch: A preliminary study. Frontiers in Education Conference (FIE), 2014
IEEE, 1-4. doi:http://dx.doi.org/10.1109/FIE.2014.7044055
Moreno-León, J., & Robles, G. (2015). Computer programming as an educational tool
in the English classroom: A preliminary study. Proceedings of the 2015 IEEE
Global Engineering Education Conference (EDUCON 2015), 961-966.
Papert, S., & Solomon, C. (1971). Twenty things to do with a computer. Retrieved from
http://18.7.29.232/bitstream/handle/1721.1/5836/AIM-248.pdf
Resnick, M. (2013). Learn to code, code to learn. EdSurge, May.
Resnick, M., Maloney, J., Monroy-Hernández, A., Rusk, N., Eastmond, E., Brennan, K.,
... Silverman, B. (2009). Scratch: Programming for all. Communications of the
ACM, 52(11), 60-67.
Román-González, M. (2015). Computational Thinking Test: Design Guidelines and
Content Validation. Proceedings of the 7th Annual International Conference on
Education and New Learning Technologies (EDULEARN 2015), 2436-2444.
Seiter, L., & Foreman, B. (2013). Modeling the learning progressions of computational
thinking of primary grade students. Proceedings of the 9th Annual International
ACM Conference on International Computing Education Research (ICER 2013), 59-
66.
Wilson, A., Hainey, T., & Connolly, T. (2012). Evaluation of computer games
developed by primary school children to gauge understanding of programming
concepts. Proceedings of the 6th European Conference on Games-Based Learning
(ECGBL), 4-5.
Wing, J. M. (2006). Computational thinking. Communications of the ACM, 49(3), 33-
35.
Wolz, U., Hallberg, C., & Taylor, B. (2011). Scrape: A tool for visualizing the code of
scratch programs. Poster presented at the 42nd ACM Technical Symposium on
Computer Science Education (SIGCSE 2011)
