=Paper=
{{Paper
|id=Vol-2294/DCECTEL2018_paper_25
|storemode=property
|title=Contextualized Instruction in Data Science and its Effect on Transfer of Learning
|pdfUrl=https://ceur-ws.org/Vol-2294/DCECTEL2018_paper_25.pdf
|volume=Vol-2294
|authors=Tomohiro Nagashima
|dblpUrl=https://dblp.org/rec/conf/ectel/Nagashima18
}}
==Contextualized Instruction in Data Science and its Effect on Transfer of Learning==
https://ceur-ws.org/Vol-2294/DCECTEL2018_paper_25.pdf
Contextualized Instruction in Data Science and its Effect
on Transfer of Learning
Tomohiro Nagashima
Carnegie Mellon University, Pittsburgh PA 15213, USA
tnagashi@cs.cmu.edu
Abstract. It is believed that designing “in-context” instruction for students so
that they can find a relationship between the instruction and their lives/back-
grounds is an effective instructional strategy. Educators have adopted this ap-
proach called contextualized instruction to make their instruction more meaning-
ful for students. Despite its popularity, however, little is known about the evi-
dence for the effectiveness of contextualized instruction on student learning. Par-
ticularly, although contextualization is thought to be deeply involved in the pro-
cess of transfer of learning, the impact of contextualization on transfer has not
been thoroughly explored. Through conducting an experiment in the domain of
data science and follow-up interviews, the proposed study attempts to provide
scientific evidence on whether and how contextualized instruction helps student
learning with a particular focus on its effect on transfer of concepts and skills in
data science.
Keywords: Contextualized Instruction, Contextualization, Transfer, Data Sci-
ence
1 Introduction
1.1 Contextualized Instruction
Integrating specific examples or stories related to students’ own lives and backgrounds
into instruction is believed to enhance student learning through making explicit con-
nections between the content of instruction and real-world contexts [7]. Educators in-
tentionally or unintentionally adopt this contextualization approach to their teaching as
it makes sense to think that instruction contextualized into students’ familiar fields/sit-
uations would better support student learning. For example, when teaching new vocab-
ulary in a language class, it is considered better to teach them using example sentences
rather than simply providing a list of new words and translations without no contextual
information on how the words should be used in real-world contexts [13]. The contex-
tualization approach and its implications have become increasingly important with the
prevalence of learning technologies which can personalize instruction to students’ per-
sonal preferences and situations [14].
Contextualization in teaching and learning has been adopted in various ways at dif-
ferent levels of education (K-12, higher education, and adult education) across many
2
subject areas (mathematics, basic skills education, language learning, and physics) un-
der different definitions and terms [5, 7, 9, 12, 13, 14]. One particularly important dif-
ference among these studies and practices is the difference in what they refer to as
“context”. For example, in the field of language learning, most studies compare an “in-
context” condition with a “no-context” condition (e.g. vocabulary in sentences versus
vocabulary as a list of words and translations) [5, 9]. Research on contextualization has
also examined whether teaching in a context relevant to students’ interests would be
better than simply providing a real-world context. For instance, in mathematics educa-
tion, students’ out-of-school personal interests, such as sports and movies, were incor-
porated into the instruction in an online learning environment and it was shown more
effective than the approach where no personal interests were incorporated [14]. At com-
munity colleges, researchers examine the effect of contextualization at an academic
domain level, finding that contextualizing teaching into students’ academic background
(e.g. contextualizing reading skill instruction into biology for students in a biology ma-
jor) is more effective than the “de-contextualized” approach where the students’ aca-
demic background was not considered [7].
Among different definitions and terms for the contextualization approach in these
studies, we decided to use the term “contextualized instruction” in this study as its def-
inition in [7] makes an explicit distinction from another popular term, “integrated in-
struction”, and it shares similar motivations and focuses of the strategy. According to
[7], contextualized instruction refers to an instructional strategy that integrates specific
contexts into the teaching of academic skills, such as reading, writing, and math skills,
whereas integrated instruction incorporates academic skills instruction into the teaching
of the content. In other words, the focus of instruction is on the teaching of academic
skills in contextualized instruction while the focus is on the teaching of the content in
integrated instruction [7].
1.2 Problem in the field
One topic that is still yet to be examined in the discussion of contextualization is how
contextualizing instruction can contribute to transfer of learning [12]. Defined as “the
ability to extend what has been learned in one context to a new context”, transfer is
considered as one of the most important goals of education [2]. As contextualized in-
struction involves the strategic use of contexts to help students learn, which is one of
the keys for successful transfer [2], several advocates argue that transfer can be mean-
ingfully achieved through the contextualization approach if designed carefully [3, 4, 7,
14]. There is evidence supporting this argument, demonstrating that high school stu-
dents taught algebra using personalized content according to students’ out-of-school
interests in an online learning environment successfully transferred the knowledge to
more complex problem-solving activities better than students in the de-contextualiza-
tion group [14]. However, to the best of the author’s knowledge, no prior studies have
rigorously tested the transfer effect of what we mean by “contextualized instruction”
[7]. Given the potential benefit of contextualized instruction on transfer of learning, it
is worthwhile to investigate its effect. Note that in our study, transfer of learning refers
specifically to the transfer of concepts and skills from one context where those are
3
learned, to another, or a more abstract context. Also, as it is suggested that transfer can
be achieved in a certain situation but not in general [2], it is important to explore when
and how contextualized instruction can contribute to transfer, which would then pro-
vide implications on the specific designs of contextualized instruction.
1.3 Data Science as the Target Domain
For the proposed study, we chose data science as our target domain. The identification
of the target domain was based on our realization that current undergraduate data sci-
ence education does not necessarily provide a learning experience in the meaningful
context for students. Although real-world data and problems may be integrated into its
curricula in many university programs [11], data science education does not usually
address students’ interests and academic backgrounds. In particular, despite the fact
that the high demand in the field requires university graduates in a variety of fields to
acquire basic data science skill, regardless of their academic backgrounds [6], typical
undergraduate data science instruction assumes that students have sufficient back-
ground knowledge in data science-related fields, including mathematics, statistics,
and/or computer science [1]. This could potentially prevent students with little
knowledge in such fields from learning data science effectively as those students would
have a hard time connecting the concepts taught with what they know in their familiar
contexts and fields. In fact, qualitative research repeatedly suggests that the lack of
connections between students and data can negatively impact how students make sense
of data and their engagement with data [8, 10]. This implies that it is meaningful to test
the approach of contextualized instruction in the domain of data science.
1.4 Research Questions
To improve data science instruction through contextualization, the proposed study in-
vestigates whether and how contextualized instruction in an online environment can
enhance the understanding and transfer of data science concepts and skills among stu-
dents having backgrounds in non-data science fields. We are particularly interested in
understanding whether contextualizing instruction into a relevant context to students is
better than a generic approach where students’ familiar contexts are not considered. We
will create online data science activities using examples and datasets from the field of
chemistry for students coming from the academic background in chemistry. The pro-
posed study, which is part of the dissertation study, examines the following research
questions:
- Will non-data science students taught using contextualized data science activi-
ties in their academic background perform better on transfer assessments than
those taught using de-contextualized data science activities?
- How will contextualized (and de-contextualized) activities contribute to stu-
dent’s understanding of data science concepts and skills?
4
2 Method
2.1 Participants
Over 100 subjects will be recruited from several introductory data science and chemis-
try courses at Carnegie Mellon University in the United States. The target population
is undergraduate students majoring in chemistry, which will be identified using data
provided by the university. The age range will be from 18 to 23.
2.2 Materials
Instructional Activities. Data science instructional activities will be created on Open
Learning Initiative (OLI: www.oli.cmu.edu), a widely-used online learning platform
developed at Carnegie Mellon University. Two types of activities will be designed for
an intervention. One of them will be contextualized into the field of chemistry (contex-
tualized activities) and the other is “de-contextualized” activities, where the activities
are not contextualized into a specific domain or example while they still ask same con-
cepts and skills as contextualized activities. For contextualized activities, relatively
easy examples in chemistry will be chosen from an introductory chemistry textbook so
that the context will be easily understood by every participant in the study.
Assessments. We will develop two types of data science problem-solving assessment
problems where one of them is contextualized into chemistry and the other is de-con-
textualized. These problems will involve the same concepts and skills as the instruc-
tional activities used in the intervention but use different examples and values. We will
use these assessment problems to examine the effect of contextualized instruction on
the transfer of learning.
2.3 Study Design and Procedure
We will conduct an experiment where we vary the type of instructional activities (con-
textualized or de-contextualized: independent variable). Our dependent variable is the
two types of assessment problems (contextualized or de-contextualized). Half of the
students in the contextualized instruction group will receive contextualized assessment
problems and the other half will receive de-contextualized assessment problems. Sim-
ilarly, half of the students in the de-contextualized instruction group will work on con-
textualized assessment problems and the other half will be given de-contextualized as-
sessment problems. Participants will be randomly assigned to either of these four
groups: 1) contextualized instructional activities and contextualized assessment prob-
lems, 2) contextualized instructional activities and de-contextualized assessment prob-
lems, 3) de-contextualized instructional activities and contextualized assessment prob-
lems, 4) de-contextualized instructional activities and de-contextualized assessment
problems.
They will be first asked to solve either contextualized or de-contextualized instruc-
tional data science activities depending on the group they belong to. Two weeks later,
5
they will be told to work on the data science assessment problems. All the procedure
will be conducted on OLI and students will individually access the materials.
We hypothesize that the contextualized instructional activities will have a positive
effect on the performance on the assessment problems, therefore expecting that groups
1 and 2 will perform better than groups 3 and 4 on the assessment problems. We also
hypothesize that group 1 will perform better than group 2 and group 4 will perform
better than group 3 because students in the groups 1 and 4 will not have to transfer the
learned concepts and skills into a more abstract context or a more concrete context,
respectively. For those groups involving transfer of knowledge from one context to
another, we expect to see that group 2 will perform better than group 3 on the assess-
ment problems because of the expected benefit of deep connection making for the con-
textualized instruction group.
Following the experiment, 3 to 5 students in each of the conditions will be randomly
selected and invited for follow-up interviews where the students will be asked how they
approached the activities, whether there was any difficulty in solving problems, and
how contextualized (or de-contextualized) activities helped them understand data sci-
ence concepts and skills.
3 Progress so far
We have identified past literature on the topic of contextualized instruction and de-
signed the study as discussed above, based on and improving upon the study designs
employed in the prior work. We are currently developing the instructional activities and
the assessment activities. We will conduct the experiment and the interviews in 2019.
4 Contributions
The proposed study will make contributions to the TEL field on the following two as-
pects:
- The study will provide evidence on the effectiveness of contextualized instruc-
tion in data science, particularly on whether it contributes to transfer of learn-
ing
As discussed above, the effect of contextualization on transfer is not yet established.
We expect that providing scientific evidence on the transfer effect would be helpful in
the research community and would produce meaningful research questions. For exam-
ple, future research can test the transfer effect of contextualized instruction in other
academic domains. Future studies can also examine the approach of providing multiple
opportunities for initial learning in contextualized instruction and its effect on transfer
[4].
- The study will suggest how contextualization should be designed, especially
with the use of learning technologies.
6
In terms of the practical use of the approach of contextualization, the complexity of
the topic and the lack of consistent evidence on its effectiveness we currently see pre-
vent educators from designing effective contextualized learning activities. We aim to
suggest several design implications on contextualized instruction to enhance transfer of
knowledge among students not only in data science but also in other related domains,
based on the findings from the experiment and the interviews. Particularly, we will
provide implications on the use of learning technologies because adaptive learning
technologies can promote the practices of contextualized instruction in effective and
efficient ways [14]. Such implications on how to design contextualized instruction can
foster meaningful communications between researchers and practitioners and can even-
tually improve classroom practices.
References
1. Anderson, P.: Bowring, J.: McCauley, R.: Pothering, G.: Starr, C.: An undergraduate degree
in data science: curriculum and a decade of implementation experience. In: Proceedings of
the 45th ACM Technical Symposium on Computer Science Education, pp. 145-150. ACM,
New York (2014).
2. Bransford, J.: Brown, A.: Cooking, R.: How people learn: brain, mind, experience, and
school. National Academy of Science, Washington (2000).
3. Cognition and Technology Group at Vanderbilt. Anchored instruction and its relationship to
situated cognition. Educational Researcher, 19(6), 2-10 (1990).
4. Gick, M.: Holyoak, K.: Schema induction and analogical transfer. Cognitive Psychology,
15(1), 1-38 (1983).
5. Gillam, S.: Gillam, R.: Reece, K.: Language outcomes of contextualized and decontextual-
ized language intervention: results of an early efficacy study. Language, Speech, and Hear-
ing Services in Schools, 43, 276-291 (2012).
6. Patil, D.: Building data science teams: The skills, tools, and perspectives behind great data
science groups. O’Reilly Media, Sebastopol (2011).
7. Perin, D.: Facilitating student learning through contextualization. Community College Re-
search Center Working Paper (2011).
8. Philip, T.: Olivares-Pasillas, M.: Rocha, J.: Becoming racially literate about dat and data-
literate about race. Cognition and Instruction, 34(4), 361-388 (2016).
9. Prince, P.: Second language vocabulary learning: the role of context versus translations as
function of proficiency. The Modern Language Journal 80(4), 478-493 (1996).
10. Rubel, L.: Lim, V.: Hall-Wieckert, M.: Sullivan, M: Teaching mathematics for spatial jus-
tice: an investigation of the lottery. Cognition and Instruction, 34(1), 1-26 (2016).
11. Song, I.: Zhu, Y.: Big data and data science: what should we teach? Expert Systems, 33(4),
364-373 (2016).
12. Taasoobshirazi, G.: Carr, M.: A review and critique of context-based physics instruction and
assessment. Educational Research Review 3, 155-167 (2008).
13. Uswatunnisa. Contextualized and decontextualized vocabulary learning to enhance vocabu-
lary understanding. PhD Dissertation, Diponegoro University.
14. Walkington, C.: Using adaptive learning technologies to personalize instruction to student
interests: The impact of relevant contexts on performance and learning outcomes. Journal of
Educational Psychology 105(4), 932-945 (2013).