Professional Documents
Culture Documents
1
tation. These dimensions are statically evaluated by the grouping of some dimensions, thus simplifying the
inspecting the source code of the analyzed project and feedback provided to users.
given a score from 0 to 3, resulting in a total mastery
score that ranges from 0 to 21 when all seven dimen- 3.1 Ecological Validity
sions are aggregated. With this information the tool
One of the main objectives of Dr. Scratch is to en-
generates a feedback report that is displayed to learn-
courage students to improve their programming skills.
ers, as shown in Figure 1.
Aiming to check its effectiveness regarding this goal,
we organized a series of workshops with over 100 stu-
dents in the range from 10 to 14 years in 8 different
schools [MLRRG15]. During the workshops, partici-
pants analyzed one of their projects with Dr. Scratch,
read the feedback report provided by the tool, and
tried to improve their projects using the tips included
in the report. After 1 hour of working with Dr.
Scratch, students increased their CT score and im-
proved their programming skills, which proves that the
tool is useful for learners of these ages.
Nevertheless, even though Dr. Scratch offers a CT-
dependent feedback report, the workshops showed that
the tool was especially useful for older students (12-
14 years old) with an intermediate initial CT score.
Figure 1: Dr. Scratch Analysis Result for Star Wars Future interventions will help us modify the feedback
StickMan DressUp, a Scratch project available at report so that younger students as well as learners with
https://scratch.mit.edu/projects/101463736/. basic or advanced CT levels can also make the most
out of the tool.
2
Table 1: Level of Development for Each CT Dimension Assessed by Dr. Scratch [MLRRG15]
Cabe’s Cyclomatic Complexity and Halstead’s met- projects are subjected to the Dr. Scratch mastery
rics [MLRRG16a]. By comparing the results after an- score. For instance, while if and if else instructions
alyzing 95 Scratch projects, we found positive, signif- are commonly present at games, this is not the case
icant, moderate to strong correlations between mea- for stories, which usually have a linear structure with-
surements, which could also be considered as a valida- out branches. Therefore, games tend to score high in
tion of the complexity assessment process of the tool. the Dr. Scratch logical thinking dimension, while the
Finally, in an intervention involving an 8-weeks pro- opposite is true for stories.
gramming course with Scratch in three Spanish mid-
dle schools, with a total sample of n=71 students, 3.4 Face Validity
we compared the Dr. Scratch scores with the re-
Another main goal of the tool is to support teach-
sults of the CT-test. The findings show a posi-
ers in the evaluation tasks. It is thus important that
tive, moderate, and statistically significant correla-
educators feel that Dr. Scratch does what it claims
tion between tools, both in predictive and concurrent
to do (i.e. assessing CT). Pursuing this goal we pre-
terms [RGMLR17]. As we expected, the convergence
pared a survey for teachers who participate in a 40-
is not total since, although both tools are assessing
hours Scratch coding training course. At the mo-
the same psychological construct, they do it from dif-
ment of writing this paper more than 320 educators
ferent perspectives: summative-aptitudinal (CT-test)
have submitted the survey and another 120 are taking
and formative-iterative (Dr. Scratch).
the course. The partial results indicate that 84% of
participating teachers feel that the CT analysis of the
3.3 Discriminant Validity tool is measured in a correct way. Having educators
who teach at different grades and who have reached
Projects shared in the Scratch repository are catego- distinct levels of development of CT skills during the
rized under one or more project types, being the most course will allow the study of potential differences in
common games, animations, music, art and stories, al- their opinions about the usefulness of the tool based
though there are other categories such as tutorials or on these variables.
simulations. Aiming to check whether Dr. Scratch
is able to detect differences in the CT dimensions de-
3.5 Factorial Validity
veloped when programming different types of Scratch
projects, we randomly downloaded 500 projects from A key pending validation action is to study the re-
the main categories in the repository. Although this lationships between the different CT dimensions as-
work has not been published (at the moment of writ- sessed by Dr. Scratch aiming to identify clusters of
ing this paper it is still under revision), the results dimensions that share sufficient variation. This fac-
of a K-means cluster analysis confirm that different torial analysis, if performed on a big enough num-
types of projects can be used to develop distinct CT ber of projects, could lead to the grouping of some
dimensions, and consequently, the existing typology of the dimensions, which would therefore simplify the
of Scratch projects is empirically replicated when the feedback report that the tool displays to learners.
3
For this action we will use the dataset made avail- the 2016 ACM Conference on Interna-
able from [AH16], which consists of 250,166 projects tional Computing Education Research,
scraped from the Scratch repository. pages 53–61. ACM, 2016.
4
[MLRGHR17] Jesús Moreno-León, Marcos Román- [RMLAH17] G. Robles, J. Moreno-León, E. Aival-
González, Casper Harteveld, and Gre- oglou, and F. Hermans. Software
gorio Robles. On the automatic assess- clones in scratch projects: on the pres-
ment of computational thinking skills: ence of copy-and-paste in computa-
A comparison with human experts. tional thinking learning. In 2017 IEEE
In Proceedings of the 2017 CHI Con- 11th International Workshop on Soft-
ference Extended Abstracts on Human ware Clones (IWSC), pages 1–7, Feb
Factors in Computing Systems, CHI 2017.
EA ’17, pages 2788–2795, New York,
NY, USA, 2017. ACM. [RMMH+ 09] Mitchel Resnick, John Maloney,
Andrés Monroy-Hernández, Natalie
[MLRRG15] Jesús Moreno-León, Gregorio Robles, Rusk, Evelyn Eastmond, Karen Bren-
and Marcos Román-González. Dr. nan, Amon Millner, Eric Rosenbaum,
Scratch: Automatic analysis of scratch Jay Silver, Brian Silverman, and
projects to assess and foster Computa- Yasmin Kafai. Scratch: Programming
tional Thinking. RED. Revista de Ed- for all. Commun. ACM, 52(11):60–67,
ucación a Distancia, 15(46), 2015. November 2009.
[MLRRG16a] Jesús Moreno-León, Gregorio Robles, [SF13] Linda Seiter and Brendan Foreman.
and Marcos Román-González. Com- Modeling the learning progressions
paring computational thinking devel- of computational thinking of primary
opment assessment scores with soft- grade students. In Proceedings of
ware complexity metrics. In 2016 the Ninth Annual International ACM
IEEE Global Engineering Education Conference on International Comput-
Conference (EDUCON), pages 1040– ing Education Research, ICER ’13,
1045, April 2016. pages 59–66, New York, NY, USA,
[MLRRG16b] Jesús Moreno-León, Gregorio Robles, 2013. ACM.
and Marcos Román-González. Exam-
[WDCK12] Linda Werner, Jill Denner, Shan-
ining the relationship between social-
non Campe, and Damon Chizuru
ization and improved software develop-
Kawamoto. The fairy performance
ment skills in the scratch code learn-
assessment: Measuring computational
ing environment. Journal of Universal
thinking in middle school. In Proceed-
Computer Science, 22(12):1533–1557,
ings of the 43rd ACM Technical Sympo-
2016.
sium on Computer Science Education,
[MR14] Jesús Moreno and Gregorio Robles. SIGCSE ’12, pages 215–220, New York,
Automatic detection of bad program- NY, USA, 2012. ACM.
ming habits in Scratch: A preliminary
study. In 2014 IEEE Frontiers in Ed- [WHC12] Amanda Wilson, Thomas Hainey, and
ucation Conference (FIE) Proceedings, Thomas Connolly. Evaluation of com-
pages 1–4, Oct 2014. puter games developed by primary
school children to gauge understanding
[RGMLR17] Marcos Román-González, Jesús of programming concepts. In European
Moreno-León, and Gregorio Robles. Conference on Games Based Learning,
Complementary tools for computa- page 549. Academic Conferences Inter-
tional thinking assessment. In CTE national Limited, 2012.
2017: International Conference on
Computational Thinking Education [WHT11] Ursula Wolz, Christopher Hallberg,
2017, page In press, July 2017. and Brett Taylor. Scrape: A tool for
visualizing the code of Scratch pro-
[RGPGJF16] Marcos Román-González, Juan-Carlos grams. In Poster presented at the 42nd
Pérez-González, and Carmen Jiménez- ACM Technical Symposium on Com-
Fernández. Which cognitive abilities puter Science Education, Dallas, TX,
underlie computational thinking? cri- 2011.
terion validity of the computational
thinking test. Computers in Human
Behavior, pages –, 2016.