You are on page 1of 12

Journal of Science Education and Technology (2018) 27:322–333

https://doi.org/10.1007/s10956-017-9726-9

Integrating Computational Science Tools into a Thermodynamics Course


Camilo Vieira 1 & Alejandra J. Magana 1 & R. Edwin García 2 & Aniruddha Jana 2 & Matthew Krafcik 2

Published online: 24 January 2018


# Springer Science+Business Media, LLC, part of Springer Nature 2018

Abstract
Computational tools and methods have permeated multiple science and engineering disciplines, because they enable scientists
and engineers to process large amounts of data, represent abstract phenomena, and to model and simulate complex concepts. In
order to prepare future engineers with the ability to use computational tools in the context of their disciplines, some universities
have started to integrate these tools within core courses. This paper evaluates the effect of introducing three computational
modules within a thermodynamics course on student disciplinary learning and self-beliefs about computation. The results suggest
that using worked examples paired to computer simulations to implement these modules have a positive effect on (1) student
disciplinary learning, (2) student perceived ability to do scientific computing, and (3) student perceived ability to do computer
programming. These effects were identified regardless of the students’ prior experiences with computer programming.

Keywords Computational science . Simulations . Thermodynamics . Self-beliefs . Conceptual change . Computational thinking

Introduction complex problems of a global society (ACM 2013; National


Science Foundation 2011; PITAC 2005; The Minerals,
Computational science and engineering (CSE) has been Metals, and Materials Society 2013). While graduate students
denominated the third pillar of science along with theoretical in engineering value the application of computational tools
and experimental approaches (PITAC 2005; Wofford 2009). and methods in their disciplines, they usually feel underpre-
CSE integrates multiple disciplines with mathematical and pared to use/develop those (Magana and Mathur 2012).
computational tools to enable scientists and engineers to solve To address this issue and at the same time integrate CSE into
complex problems (Shiflet and Shiflet 2006; Weintrop et al. undergraduate programs, several alternatives have been sug-
2016). The body of knowledge of the ACM Computer gested (Turner et al. 2002): creating concentrations in CSE,
Science (CS) curricula in 2013 defined CSE as Bthe applica- creating multidisciplinary project-based courses, enhancing
tion of computer science to solve problems across a range of existing courses with the use of CSE tools and methods, and
disciplines^ (The Joint Task Force on Computing Curricula— using computer visualizations as a vehicle for teaching and
ACM/IEEE-Computer Society 2013, p. 68). CSE involves learning disciplinary knowledge. Hence, some universities
several techniques such as modeling, simulation, visualiza- and colleges have started to integrate CSE tools and methods
tion, data processing, and numerical analysis. into their disciplinary engineering curricula (Magana et al.
Several reports and agencies highlight that engineering cur- 2013; Alabi et al. 2015), science courses (Wofford 2009), and
ricula need to keep the pace with advances in computational liberal arts environments (Turbak and Berg 2002). For instance,
science (Wofford 2009). Engineering graduates need to be the Department of Materials Science and Engineering at a mid-
prepared with the CSE knowledge and skills to address the Atlantic coast university created a curricular innovation that
involved one computation and programming course for mate-
* Camilo Vieira rials scientists and twelve computational modules integrated
cvieira@purdue.edu into six core courses (Magana et al. 2013; Magana et al.
2016). Similarly, the materials engineering department at
1
Department of Computer and Information Technology, Purdue
University of Illinois at Urbana-Champaign has integrated
University, 401 N. Grant Street, West Lafayette, IN 47907, USA computational modules throughout their undergraduate curric-
2
School of Materials Engineering, Purdue University, Neil Armstrong
ulum (Mansbach et al. 2016).
Hall of Engineering, 701 West Stadium Avenue, West This paper describes the implementation of three computa-
Lafayette, IN 47907-2045, USA tional modules as part of a Thermodynamics of Materials
J Sci Educ Technol (2018) 27:322–333 323

course at a large Midwestern university. Each module activities, (4) scaffolding detailed mathematical activities,
consisted of a 50-min lab session in which students used the and (5) applying the course concepts into real life situations.
Virtual Kinetics of Materials Laboratory (VKML), hosted in However, when these simulations involve computer program-
nanoHUB.org (https://nanohub.org/tools/vkmllive). VKML is ming, students face challenges in mapping the conceptual
based on Gibbs (Cool et al. 2010), a set of Python-based model, the mathematical representation, and the algorithmic
libraries that enables students to represent complex thermody- representation (Vieira et al. 2017), and therefore, they may
namics phenomena using modeling techniques and simple benefit from additional support.
python programming. In order to assess this micro curricular This study explores the introduction of Gibbs, a Python-
innovation, the research team explored the following research based computer simulation software, into a course of thermo-
questions (RQ): dynamics of materials. During the implementation of three
computational modules, students had access to the Python
RQ1: What is the effect of introducing computational code that created the visualizations and simulations. In order
modules as part of a thermodynamics course on student to reduce the cognitive load from these activities, students
disciplinary learning? were provided with a worked example that was introduced
RQ2: What is the effect of introducing computational by the instructor. The meaningful exploration of these exam-
modules as part of a thermodynamics course on student ples was encouraged through self-explanation activities within
self-beliefs about computation? the homework assignments.
RQ3: What are the factors that affect changes on student
understanding of disciplinary phenomena after being ex-
posed to computational modules as part of a thermody- Constructivist Approaches to Conceptual
namics course? Change
RQ4: What are the factors that affect changes on student
self-beliefs about computation after being exposed to Conceptual change is the theoretical framework that guided
computational modules as part of a thermodynamics our investigation. Conceptual change has been broadly re-
course? ferred to as the process by which the individuals’ conceptual
frameworks are replaced by another set of beliefs which are
incompatible with the first (Posner et al. 1982). Researchers
have explained conceptual change as a the process of acquir-
Thermodynamics Teaching and Learning ing new knowledge based on models of philosophy of science
(Posner et al. 1982), as ontological shifts (Chi et al. 2012), or
Thermodynamics is an important course for several disci- as processes for restructuring the learners’ preinstructional
plines in engineering and the sciences (Probst and Zhang conceptual structures in order to allow understanding of sci-
2013), including mechanical, chemical, and materials engi- ence concepts (Duit and Treagust 2003). Although there is still
neering, as well as physics and chemistry. Yet, thermodynam- not a common agreement of the cognitive processes inherent
ics is a complex area that students struggle to understand and to conceptual change, there seems to be an agreement that
visualize (Mulop et al. 2012), and at the same time, instructors constructivist approaches to science learning may result in
usually struggle to find effective ways to teach it (Cobourn conceptual change (Duit 1999).
and Lindauer 1994; Probst and Zhang 2013). Some of the Conceptual change-based strategies focus on externalizing
common misunderstandings and difficult concepts in thermo- an existing the learner’s knowledge structure and then modi-
dynamics include (Mulop et al. 2012; Olds et al. 2004) work, fying it via instruction. Instructional elements for conceptual
heat, enthalpy, entropy, and equilibrium, as well as applying change include helping students to identify or externalize their
these concepts for problem-solving and visualizing these ab- own misconceptions, fragmented or incomplete conceptions;
stract phenomena. then provide instruction where students have an opportunity to
Researchers advocate for constructivist learning environ- apply an accurate scientific conception; and finally reinforce
ments to meaningfully engage students in the learning process the accurate conceptions via repetition (Anderson and Smith
of thermodynamics (Mulop et al. 2012). Along these lines, 1987).
computer simulations have been shown to be an effective Computer simulations have been used as a constructivist
approach to support this complex teaching and learning pro- approach for conceptual understanding in science (Khan
cess (Baher 1999; Cobourn and Lindauer 1994; Yang et al. 2011; Windschitl 1995; Wofford 2009). The use of computer
2012). Some of the affordances of these simulations in ther- simulations with proper instruction can facilitate the four con-
modynamics are (Baher 1999; Mulop et al. 2012; Khan 2011) ditions to promote conceptual change: (1) dissatisfaction with
(1) visualizing abstract concepts, (2) identifying relationships existing concepts, (2) the new concept should be explainable,
among variables, (3) supporting monitoring learning (3) the new concept should be intelligible, and (4) the new
324 J Sci Educ Technol (2018) 27:322–333

concept should be able to be extended (Posner et al. 1982). We this study was adapted to include two different areas in which
argue that in the same way computer simulations can promote the students could develop their ability to control the outcome:
conceptual change in science, they can also promote concep- designing and writing an algorithm or computer program and
tual change in engineering education. using computation for scientific purposes (e.g., visualizing a
phase diagram or solving a set of linear equations).
Throughout the entire semester, students attended regular
Methods bi-weekly lectures and one laboratory session were they
worked on the computational modules. Each computational
Participants module started with a 5-min pretest regarding THERMO con-
cepts that were explained during the lectures. Appendix 1
Students enrolled in a sophomore thermodynamics of mate- includes the pre/posttest instruments. After the pretest, stu-
rials course (herein called THERMO) were the participants for dents worked with the Gibbs software (Cool et al. 2010), as
this study. During the spring semester 2015, 62 students par- implemented in the VKML (Bartol et al. 2014).
ticipated in this study, while in the spring 2016 semester, 46 Three computational modules were introduced as part of
students participated too. At the beginning of the semester, the the THERMO course (see Table 2). All the modules materials
participants were asked about the number of programming and the worked examples were accessible to students through
courses they had taken before (either high school or college the VKML hosted on nanoHUB.org. The activities within
level). Twenty-two students had not taken any programming each module consisted of using VKML to represent the
courses before, 37 students had taken one course, 33 students thermodynamics phenomena. After getting familiar with the
had taken two courses, and 16 students had taken three or worked example, students were asked to complete a
more courses. Eighty-eight students mentioned to be familiar homework assignment to write in-code comments explaining
with at least one programming language, and 30 of these stu- what the code was doing. In 2015, students were allowed to
dents reported to be familiar with Python. This study was complete the assignment individually or in groups, and this
approved by the institutional review board (IRB) at the partic- assignment was extra-credit for the students. Since most of the
ipating institutions. students chose not to complete these extra-credit activities, the
instructor and the research team decided to make them indi-
Procedures and Data Collection vidually graded as part of the homework assignment in 2016.
One week after the computational module was delivered, a
Figure 1 summarizes the research procedures along with the posttest was administered, often together with the correspond-
data collection instruments for this study. The research process ing pretest for the following module.
started with a baseline survey that prompted students to report The first module was implemented as part of the topic of
their self-beliefs regarding their ability to write computer pro- state variables and state functions. A state function is Ba prop-
grams. The survey on self-beliefs was adapted from Magana erty of a material system that mathematically describes the
et al. (2016) as described in Table 1. The original survey only equilibrium state of the system independently on what process
included the following constructs: (1) control appraisals, de- the system followed to arrive at that condition^ (DeHoff
fined as the students’ Blevel of control of the outcome^ (p. 2006). These concepts had been introduced to the students
430); and (2) value appraisals, defined as the Bperceived im- during that week lecture sessions. The pretest and posttest
portance of achieving that outcome^ (p. 430). The survey for consisted of a single yes-or-no question that required students

Fig. 1 Intervention in the SAMPLE MODULE


thermodynamics course. 1 SELF-EXPLAINING
ONE EXAMPLE
Baseline survey on self-beliefs, 2 PRETEST Students write POSTTEST
Instructor walks
three computational modules Thermo
students through
comments within Thermo
including pretest/posttest and Concepts the code to self- Concepts
the example
student explanations of a worked explain it
example, 3 project interview, 4
end-of-the-semester interview
1. BASELINE

3. PROJECT

SURVEY
3. EXIT

Start MODULE 1 MODULE 2 MODULE 3 End

2. COMPUTATIONAL MODULES

SPRING 2015, 2016


J Sci Educ Technol (2018) 27:322–333 325

Table 1 Self-belief survey items as adapted from Magana et al. (2016)

Original Adapted

Construct Question Construct Question

Control I have the ability to design an algorithm. Programming abilities Q1. I have the ability to design an algorithm.
I have the ability to write a computer program. Q2. I have the ability to write a computer
program.
I have the ability to visualize data using a Scientific computing abilities Q3. I have the ability to visualize data using
computer. a computer.
I have the ability to implement a graphical Q4. I have the ability to build phase diagrams
user interface. using computation.
N/A Q5. I have the ability to use a computer to solve
a set of linear equations.
N/A Q6. I have the ability to implement a numerical
model based on a simple partial differential
equation
Value I feel computation (data visualization, modeling Perceived value of computation Q7. I feel computation (data visualization,
and simulation algorithm design) will be modeling, and simulation algorithm design)
useful in my studies. will be useful in my studies.
I feel computation (data visualization, modeling Q8. I feel computation (data visualization,
and simulation algorithm design) will be modeling, and simulation algorithm
useful in my career. design) will be useful in my career.
I intend to purposefully seek courses that will Q9. I intend to purposefully seek courses that
allow me to increase my knowledge about will allow me to increase my knowledge
computation. about computation.
I intend to use computation (data visualization, Q10. I intend to use computation (data
modeling, and simulation algorithm design) visualization, modeling, and simulation
in my future career. algorithm design) in my future career.

to show their procedures (see Appendix 1). Students were Then, the instructor guided the implementation along with a
given five extra-credit points for completing each of these worked example of this program. The student assignments for
instruments correctly, and the teaching assistant (TA) was in the following week included commenting the code to describe
charge of the scoring process. each step and modify the program so that it evaluated a dif-
The session continued with the course instructor describing ferent function. The worked example for this module is pre-
simple actions in VKML such as create a new file, declare sented in Fig. 2. The code first declared the thermodynamic
variables, write a Bhello world^ program, write a for-loop, variables to create the equation A on section 3. On section 5,
execute a program, import Python packages, and insert com- the program computed the partial derivatives of the equation,
ments within the code. The instructor also highlighted the which were differentiated again on section 7. On section 9, the
importance of commenting the code to be able to collaborate worked example verified whether both derivatives were the
and better understand a program. The next part of the module same or not to identify equation A as a state function.
was intended to write a program in VKML that would identify The other two computational modules followed the same
whether a given equation was a state function or not. First, the procedures as in the first one, but with subsequent topics from
group discussed the activity conceptually and mathematically. the course. Module II focused on stability of binary solutions

Table 2 Learning outcomes of


the three computational modules Module Learning outcome

Module [Disciplinary] Learners will identify the conditions that define thermodynamic stability,
1 metastability, and instability[Computational] Learners create, execute, and interpret python
programs using VKML
Module [Disciplinary] Learners will identify the free energy curve for eutectic, peritectic, and lenticular
2 (isomorphous) phase diagrams[Computational] Learners will create binary phase diagrams using
VKML
Module [Disciplinary] Learners will identify the phases of the peritectic phase diagram and the solubility
3 limit of a given component[Computational] Learners will create phase diagrams using VKML
326 J Sci Educ Technol (2018) 27:322–333

Table 3 Pearson correlation among survey questions

Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10

Q1 1
Q2 0.53 1
Q3 0.34 0.49 1
Q4 0.38 0.44 0.61 1
Q5 0.29 0.24 0.41 0.47 1
Q6 0.46 0.34 0.44 0.46 0.44 1
Q7 0.36 0.27 0.39 0.43 0.21 0.29 1
Q8 0.27 0.34 0.4 0.42 0.18 0.33 0.67 1
Q9 0.22 0.31 0.25 0.33 0.19 0.17 0.59 0.56 1
Q10 0.27 0.4 0.36 0.51 0.21 0.28 0.58 0.6 0.6 1

All the correlation values were statistically significant: p value < 0.05.
All italicized entries have a p value < 0.001

Fig. 2 Worked example for module no. 1: identifying whether a given


function is a state function or not
Our factor analysis showed similar results compared to
Magana et al. (2016 regarding the perceived value. The
construct scientific computing abilities also shared the
item Q2 with the construct on programming abilities.
with phase diagrams while module III introduced phase dia-
Table 4 also shows the internal consistency measure for
gram analysis of composition range(s). These two modules
all the constructs. Cronbach’s alpha (Cronbach 1951) was
included a graphical user interface where students could
computed for the constructs scientific computing abilities
change parameter values to identify their effect on the phase
and perceived value, but not for programming abilities.
diagrams. Table 2 summarizes the learning outcomes of the
Cronbach’s alpha is not recommended for constructs with
modules. All the python codes are accessible through VKML
only two items (Eisinga et al. 2013), and therefore, we
(Bartol et al. 2014).
report the Spearman-Brown coefficient as the reliability
measure for this construct. An alpha of 0.7 or higher has
Validity and Reliability been suggested to be acceptable in the social sciences
(Kline 1999).
In order to validate the survey items and constructs, we The content validity of the pretest/posttest instruments
computed the Pearson correlation coefficient among the is given by experts iteratively designing and reviewing
ten questions using students to the survey instruments. these instruments. First, the course instructor and the
The number of complete observations that were used for teaching assistant met to discuss what type of questions
this analysis was 198. Table 3 shows the Pearson correla- would be relevant for a given module. Then, the teaching
tion values among all the survey questions. All the corre- assistant created sample pretest and posttest questions,
lation values were statistically significant (i.e., p value < which the course instructor made comments on. The teach-
0.05). The item Q4 (I have the ability to build phase ing assistant refined the instruments until agreement was
diagrams using computation) showed a moderate correla- reached. The scoring process was also discussed between
tion to several items within the perceived value constructs the course instructor and the teaching assistant until agree-
(e.g., Q7, Q8, and Q10). Since students were part of a ment was reached.
Materials Science and Engineering program, this relation-
ship makes sense: students who know how to visualize
phase diagrams using computation, also, see the value of Data Analysis
using computation in their career.
We then performed a factor analysis using the maxi- The pretest/posttest instruments, assessing conceptual
mum likelihood factoring method to validate the three knowledge, were scored by the TA considering not only
constructs (see Table 4). Since the factors were correlated the students’ correct response but also their procedure or
to each other, we used oblimin as the rotation method. explanation. Each of these pretest-posttest pairs from the
J Sci Educ Technol (2018) 27:322–333 327

Table 4 Factor analysis and


Cronbach’s alpha for the three Construct Internal consistency Item Factor 1 Factor 2 Factor 3
constructs
Programming abilities 0.69* Q1 0.99
Q2 0.31 0.35
Scientific computing abilities 0.77** Q3 0.74
Q4 0.75
Q5 0.62
Q6 0.52
Perceived value 0.86** Q7 0.77
Q8 0.77
Q9 0.8
Q10 0.7
Eigenvalue 2.44 2.07 1.29
% of variance 24% 21% 13%

***Spearman-Brown formula value; Cronbach’s alpha

four computational modules were also compared using a modules within a thermodynamics course, a materials engi-
paired t test, aimed at identifying statistically significant neering core course. We first present the descriptive statistics
differences in the students’ disciplinary understanding af- for each of the measures and basic inferential statistics to
ter being exposed to the computational modules (RQ1). identify potential changes pretest to posttest. The next step is
The assumptions of normality and homogeneity of vari- then to identify whether the students’ prior experiences with
ance were satisfied. In order to identify the factors that programming or the graded self-explanation activities had an
influence these changes (RQ3), we used a one-way effect on these changes.
ANOVA with the gain from pretest to posttest within each
module as a dependent variable. The independent vari- Learning Gains
ables that we evaluated are (1) the course offering, 2015
or 2016; (2) the number of programming courses the stu- Table 5 lists the average score for the pretest and posttest
dents were previously exposed to; and (3) whether the instruments in the three computational modules. These
student had prior experience with Python or not. instruments were scored in a scale from zero to five: a
We computed the three constructs from the survey (i.e., low score is considered to be below two, a moderate score
programming abilities, scientific computing abilities, and is between two and 3.5, and a high score is considered to
perceived value of computation) as a linear combination be higher than 3.5. Overall student performance on the
of the items that resulted from the factor analysis, using pretest can be considered from low to moderate, except
the corresponding factor loadings. We then identified sig- for module two where on average the student’s performed
nificant changes between pre and posttest for each of over 50%. Student performance on the posttest is higher
these constructs using a paired t test (RQ2). The assump- for all the modules compared to the pretest. Modules one
tions for the parametric test were satisfied for the three and two moved towards a high score, while module three
constructs. Finally, we compared the gain from pretest to resulted now as moderate.
posttest within each construct using one-way ANOVA The scores for pretest and posttest within each compu-
with the same independent variables as before (RQ4). tational module were compared using a paired t test. We
The purpose of this analysis was to identify whether any used Bonferroni correction to protect from type I error.
of these variables had an effect on the changes of the Our initial alpha level of 0.05 was divided by the three
students’ self-beliefs. tests that we conducted for the three modules. Thus, in
order to be a significant result, the p value must be below
0.0167. The three modules showed positive significant
Results differences from pretest to posttest, suggesting that stu-
dent exposure to these modules actually increased their
This study explores the changes on student learning and self- conceptual understanding. The effect sizes for these dif-
beliefs after being exposed to three computational science ferences were computed using Cohen’s d (Cohen1988).
328 J Sci Educ Technol (2018) 27:322–333

Table 5 Descriptive and


inferential statistics of pretest and Module no. Pretest Posttest t test
posttest instruments for the three
computational modules Mean SD N Mean SD N t DF p value d

1 2.7 2.13 107 3.84 1.92 105 − 5.37 103 < 0.01* 0.56
2 3.22 2.01 105 4.01 1.47 102 − 3.52 98 < 0.01* 0.45
3 2.18 1.85 107 3.43 1.96 106 − 5.96 104 < 0.01* 0.66

*p value < 0.0167, significant difference with Bonferroni correction

These values can be interpreted as follows (Rubin 2012): show a significant effect from pretest to posttest. The effect
weak—|d| < 0.2; weak to moderate—0.2 < |d| < 0.4; mod- size was weak to moderate for the programming abilities and
erate—0.4 < |d| < 0.65; moderate to strong—0.65 < |d| < strong for the scientific computing abilities.
0.8; and strong—0.8 < |d|. Modules one and two showed
a moderate effect size, while module three showed a mod-
erate to strong effect size. Controlling Factors

Self-Beliefs In order to identify factors that could affect student conceptual


change or changes in student self-beliefs, the research team
The descriptive statistics for the three self-belief con- carried out a one-way ANOVA using the following indepen-
structs assessed in the beginning and the end of the dent variables: (1) course offering: 2015 or 2016, (2) number
semester survey are presented in Table 6. Since these of previous programming courses, and (3) prior Python
scores were computed using the factor loadings derived experience.
from the factor analysis process, we normalized their The students’ performance between pretest and posttest
resulting value so that they are comparable to each oth- did not show any significant differences for module one
er. Values in the range 0–30% were considered negative, regarding the course offering (F(95 ,1) = 2.98, p value =
the range 31–70% was considered neutral, and anything 0.09), previous programming courses (F(95, 6) = 1, p val-
above 70% was considered positive self-beliefs. The re- ue = 0.09), or prior Python experiences (F(95, 1) = 2, p
sults suggest that the two constructs related to ability value = 0.16), neither did they show for module two:
increased, although the programming abilities remained course offering (F(95, 1) = 1.07, p value = 0.31), previous
neutral. The scientific computing abilities moved from programming courses (F(95, 6) = 1.21, p value = 0.31), or
neutral to positive, and the perceived value of computa- prior Python experiences (F(95, 1) = 1.07, p value = 0.31);
tion remained positive in spite of a small negative and module three: course offering (F(95, 1) = 3.9, p val-
change from pretest to posttest. ue = 0.05), previous programming courses (F(95, 6) =
These three constructs were then compared from pretest to 0.73, p value = 0.63), or prior Python experiences (F(95,
posttest using paired t tests. The two constructs related to 1) = 0.05, p value = 0.8). These results suggest that the
student ability to control certain outcome showed significant students’ previous experiences with programming did
positive differences after students had been exposed to the not affect their learning gains regarding the disciplinary
computational modules. Only the perceived value did not knowledge in any of the modules.

Table 6 Descriptive and inferential statistics of pretest and posttest instruments for the three computational modules

Construct Pretest Posttest t test

Mean (%) SD (%) N Mean (%) SD (%) N t DF p value d

Programming abilities 56.53 22.95 108 64.37 22.95 91 − 2.66 90 < 0.01* 0.34
Scientific computing abilities 61.73 16.88 108 75.25 16.17 92 − 6.49 91 < 0.01* 0.82
Perceived value 74.59 18.34 107 71.96 18.59 92 0.87 90 0.39 0.14

*p value < 0.0167, significant difference with Bonferroni correction


J Sci Educ Technol (2018) 27:322–333 329

The changes from the pretest survey to the posttest that for the three modules showed a positive significant differ-
were identified in the programming ability construct did ence between pretest and posttest measures. The results
not depend on the course offering (F(82, 1) = 0.16, p val- suggest that the constructivist approach (Mulop et al.
ue = 0.69), the previous programming courses (F(82, 6) = 2012) using the VKML computer simulation supported
0.92, p value = 0.49), or the students’ prior Python expe- the students’ conceptual change (Duit and Treagust
riences (F(82, 1) = 3.85, p value = 0.05). Although the 2003; Pyatt and Sims 2012) in the context of thermody-
overall perceived value of computation did not change namics. Having students reflect on how the examples
significantly from pretest to posttest, we still conducted worked, and use the interactive visualizations, may have
this analysis to identify whether certain groups of students helped them to better understand abstract concepts taught
(e.g., students with certain previous experiences) had had during the lecture sessions (Magana et al. 2016; Anderson
specific changes. However, the results were not signifi- and Smith 1987; Weintrop et al. 2016). In a way, the pretest
cant for any of the three independent variables as related instruments might have also helped students to identify their
to the students’ perceived value of computation. Only the fragmented or incomplete conceptions, while the computer
changes in scientific computing ability showed a signifi- simulations helped students to repair these misconceptions
cant difference between students who had prior experi- through the visualization of abstract phenomena (Magana
ence with Python and those who did not have them et al. 2016; Anderson and Smith 1987).
(F(83, 1) = 6.157, p value = 0.02). Students who had prior
experience with Python showed a smaller increase RQ2: What is the effect of introducing computational
(mean = 5.12%, SD = 17.88%) as compared to those stu- modules as part of a thermodynamics course on student
dents who did not have this previous experiences (mean = self-beliefs about computation?
16.58%, SD = 20.10%).
Student self-beliefs about computation also changed as a
result of the computational modules. The construct that
Discussion had the largest positive effect was scientific computing
abilities, which was aimed to assess the students’ ability
This study explored the effect of introducing computa- to use computation to represent disciplinary phenomena.
tional modules into a thermodynamics course on student The programming ability showed a weak to moderate pos-
learning and student self-beliefs. Three computational itive change, suggesting that even though the modules did
modules employed the VKML computer simulations to not require students to create programs, these still helped
represent thermodynamics abstract phenomena and to ex- students to increase their perceived ability to do computer
pose students to computational tools that can be useful in programming. Finally, the construct assessing the per-
their careers. The next subsections explore:(1) changes on ceived value of computation did not show a significant
student learning, (2) changes on student self-beliefs, (3) change. However, this construct had shown a positive
factors affecting student learning, and (4) factors affecting perceived value from the pretest, which suggests that stu-
student self-beliefs. dents already found computation to be an important ele-
ment in their careers. These findings are aligned to other
RQ1: What is the effect of introducing computational studies showing that supplementing disciplinary courses
modules as part of a thermodynamics course on student with computational modules have a positive effect on
disciplinary learning? the students’ perceived abilities with computation, but
not on the already positive perceived value of computa-
Students attended the computational modules after being ex- tion (Magana et al. 2016; Mansbach et al. 2016).
posed to the classroom lectures. The modules started with a
pretest assessing student understanding of the concepts that RQ3: What are the factors that affect changes on student
were going to be represented using VKML computer simula- understanding of disciplinary phenomena after being
tion. Students then interacted with the computer simulation exposed to computational modules as part of a thermo-
using a worked example that helped them represent such ab- dynamics course?
stract concepts. One week later, when the next computational
module was starting, students completed the corresponding Prior research has shown that computational modules involv-
posttest. ing computer programming may require to have students be
The results from the pretest/posttest analysis suggest exposed to prior programming experiences in order to have a
that students benefited from the computational modules positive effect on disciplinary learning (Magana et al. 2016).
to increase their understanding about thermodynamics In order to identify whether these prior experiences have an
concepts taught during the lecture time. Student scores effect on student learning, the research team conducted one-
330 J Sci Educ Technol (2018) 27:322–333

way ANOVA controlling for the following variables: (1) nanohub.org/tools/vkmllive). VKML is based on Gibbs
course offering: 2015 or 2016; (2) number of previous pro- (Cool et al. 2010), a set of Python-based libraries. Students
gramming courses; and (3) prior Python experience. The rel- were provided with a worked example code that created a
evance of including the course offering lies on the fact that the representation of a given phenomenon, and they could either
students’ self-explanation activities were graded for the sec- interact with the code directly, or just with the user interface to
ond offering (2016) of the course, but not for the first one modify parameters.
(2015). The results from this study suggest that the construc-
The results from this study suggest that the students’ tivist approach using computer simulations can help stu-
conceptual change after being exposed to the computa- dents to better understand abstract disciplinary concepts.
tional modules was not affected by their prior experiences At the same time, student self-beliefs about computation
nor by the course offering. Although the computational changed towards positive ones, especially as related to
modules allowed students to access the Python code, this ability to do scientific computing. Moreover, the fact that
was not a requirement for students to take advantage of some students did not have prior experiences with com-
the computational visualizations. Also, providing students puting did not prevent them from learning and having this
with the worked examples may have been enough scaf- positive effect in their self-beliefs. In fact, students with
folding for students to be able to focus on the interpreta- no experiences in Python actually showed a larger posi-
tion of the output instead of actually building the tool. tive effect in their perceived ability to do scientific
Even in 2016, when students were asked to write com- computing.
ments within the code as a self-explanation strategy, the While this study evaluated the effect of the modules on
learning gains for disciplinary knowledge were not signif- student disciplinary learning and self-beliefs of computa-
icantly different when compared to 2015. tion, we still do not know how much students actually
learned about computation, and how is that related to dis-
RQ4: What are the factors that affect changes on student ciplinary learning or student self-beliefs. Some students
self-beliefs about computation after being exposed to used the VKML to complete their course projects, but this
computational modules as part of a thermodynamics was not a requirement for the project itself. Consequently,
course? the next steps in this research will look into changes on
student learning of computation and how these can trans-
A similar analysis was conducted to identify whether fer to other disciplinary contexts.
these factors had an effect on the changes of the students’ An additional limitation of this study is the lack of a
self-beliefs after participating in the computational mod- control group to compare this constructivist approach to
ules. In this case, the results suggested significant differ- more traditional approaches (e.g., classroom lectures).
ences for changes in scientific computing ability when Since students participating in this study spent additional
students had prior experience with Python. Students with- time working on these concepts after the lectures, it is
out this experience showed a larger positive effect from possible that such time on task had an effect on the out-
the computational modules in their ability to use it for comes of this study as related to student learning. A future
scientific computing when compared to students with this study may look at comparing two groups that spend the
experience. This result suggests that the approach evalu- same time working on these concepts to identify the
ated here (i.e., VKML simulation tool + worked examples affordances of using computational models and simula-
in thermodynamics) helped students to increase their self- tions for thermodynamics concepts.
perceived ability, even if they had no experience with the
programming language. Acknowledgements The research reported in this paper was supported in
part by the US National Science Foundation under the awards no.
EEC1449238 and no. EEC1329262. The content is solely the responsi-
bility of the authors and does not necessarily represent the official views
Conclusion, Limitations, and Next Steps of the National Science Foundation.

This study explored the effect of introducing three computa- Compliance with Ethical Standards All procedures in this study that in-
tional modules as part of a thermodynamics course on student volved human participants were in accordance with the ethical standards
disciplinary learning and student self-beliefs. The computa- of the institution and were approved by the Institutional Review Board
(IRB) before the implementation of the research procedures. All proce-
tional modules were taught after students attended the class-
dures performed in studies involving human participants were in accor-
room lectures, and these were designed to support the stu- dance with the ethical standards of the institutional research committee
dents’ understanding of abstract thermodynamics phenomena. and with the 1964 Helsinki Declaration and its later amendments or com-
The modules used the Virtual Kinetics of Materials parable ethical standards. For this type of study, formal consent is not
Laboratory (VKML), hosted in nanoHUB.org (https:// required.
J Sci Educ Technol (2018) 27:322–333 331

Appendix 1

Pretest Module I

Q: Is the following f = ( )

(a) Yes (b) No

[Encircle the correct answer. Show all your working steps below. Use the back side of this page if necessary.]

Module I

Q: Is the following f = /

(a) Yes (b) No

[Encircle the correct answer. Show all your working steps below. Use the back side of this page if necessary.]

Pretest Module II
332 J Sci Educ Technol (2018) 27:322–333

Pretest Module III

Posest Module III


J Sci Educ Technol (2018) 27:322–333 333

References National Science Foundation. (2011). Empowering the nation through


discovery and innovation - NSF strategic plan for fiscal years (FY)
2011–2016. Washington, DC: Author. Retrieved from http://www.
Alabi, O., Magana, A.J. and Garcia, R.E. (2015). Gibbs, computational nsf.gov/news/strategicplan/nsfstrategicplan_2011_2016.pdf.
simulation as a teaching tool for students’ understanding of thermo- Olds, B. M., Streveler, R. A., Miller, R. L., & Nelson, M. A. (2004).
dynamics of materials concepts. Journal of Materials Education, Preliminary results from the development of a concept inventory in
37(5-6), 239-260. thermal and transport scienceage, 9, 1.
Anderson, C., & Smith, E. (1987). Teaching science. In V. Richardson- PITAC. (2005). Computational science: ensuring America’s
Koehler (Ed.), The educator’s handbook: a research perspective (pp. competitiveness. Retrieved from http://www.nitrd.gov/pitac/
84–111). New York: Longman. reports/20050609_computational/computational.pdf
Baher, J. (1999). Articulate virtual labs in thermodynamics education: a Posner, G. J., Strike, K. A., Hewson, P. W., & Gertzog, W. A. (1982).
multiple case study. J Eng Educ, 88(4), 429–434. https://doi.org/10. Accommodation of a scientific conception. Toward a theory of con-
1002/j.2168-9830.1999.tb00470.x. ceptual change. Sci Educ, 66(2), 211–227. https://doi.org/10.1002/
Bartol, A., Cool, T., & García, R. E. (2014). The virtual kinetics of materials sce.3730660207.
laboratory. Retrieved from https://www.nanohub.org/tools/vkmllive/. Probst, D. K., &Zhang, Y. (2013) A gentle bridge between dynamics and
Chi, M. T. H., Roscoe, R. D., Slotta, J. D., Roy, M., & Chase, C. C. (2012). thermodynamics. In proceedings of the 2013 ASEE Annual
Misconceived causal explanations for emergent processes. Cogn Sci, Conference & Exposition, Atlanta, Georgia. https://peer.asee.org/
36(1), 1–61. https://doi.org/10.1111/j.1551-6709.2011.01207.x. 19062
Cobourn, W. G., & Lindauer, G. C. (1994). A flexible multimedia instruc- Pyatt, K., & Sims, R. (2012). Virtual and physical experimentation in
tional module for introductory thermodynamics. J Eng Educ, 83(3), inquiry-based science labs: attitudes, performance and access. J
271–277. https://doi.org/10.1002/j.2168-9830.1994.tb01115.x. Sci Educ Technol, 21(1), 133–147. https://doi.org/10.1007/s10956-
Cohen, J. (1988). Statistical power analysis for the behavioral sciences 011-9291-6.
(pp. 20–26). Hillsdale: Lawrence Earlbaum Associates. Rubin, A. (2012). Statistics for evidence-based practice and evaluation
Cool, T., Bartol, A., Kasenga, M., Modi, K., & García, R. E. (2010). (research, statistics and program evaluation) (3rd ed.). Nashville:
CALPHAD: computer coupling of phase diagrams and thermo- Cengage Learning.
chemistry Gibbs: phase equilibria and symbolic computation of ther- Shiflet, A. B., & Shiflet, G. W. (2006). Introduction to computational
modynamic properties. CALPHAD: Computer Coupling of Phase science: Modeling and simulation for the sciences. Princeton:
Diagrams and Thermochemistry, 34(4), 393–404. https://doi.org/10. Princeton University Press.
1016/j.calphad.2010.07.005. The Joint Task Force on Computing Curricula—ACM/IEEE-Computer
Cronbach, L. J. (1951). Coefficient alpha and the internal structure of Society. (2013). Computer Science Curricula 2013: Curriculum
tests. Psychometrika, 16(3), 297–334. https://doi.org/10.1007/ Guidelines for Undergraduate Degree Programs in Computer
BF02310555. Science. Practice.https://doi.org/10.1145/2534860.
DeHoff, R. (2006). Thermodynamics in materials science (2nd ed.). Boca The Minerals, Metals & Materials Society. (2013). Integrated
Raton; Taylor & Francis. Computational Materials Engineering (ICME): implementing
Duit, R. (1999). Conceptual change approaches in science education. In ICME in the aerospace, automotive, and maritime industries.
W. Schnotz, S. Vosniadou & M. Carretero (Eds.), New perspectives Warrendale. Retrieved from www.tms.org
on conceptual change (pp. 263–282). Oxford: Elsevier Science Turbak, F., & Berg, R. (2002). Robotic design studio: Exploring the big
Duit, R., & Treagust, D. F. (2003). Conceptual change: a powerful frame- ideas of engineering in a liberal arts environment. J Sci Educ
work for improving science teaching and learning. Int J Sci Educ, Te c h n o l , 11 ( 3 ) , 2 3 7 – 2 5 3 . h tt p s : / / d o i . o rg / 1 0 . 1 0 2 3 / A :
25(6), 671–688. https://doi.org/10.1080/09500690305016. 1016376818781.
Eisinga, R., Grotenhuis, M.. & Pelzer, B. (2013) Int J Public Turner, P. R., Cunningham, S., Phillips, A. T., Claire, E., Shiflet, A. B.,
Health 58(4), 637–642. https://doi.org/10.1007/s00038-012-0416-3 Stewart, K., …Shiflet, A. (2002). Undergraduate Computational
Khan, S. (2011). New pedagogies on teaching science with computer Science and Engineering Programs and Courses. In SIGCSE (pp.
simulations. J Sci Educ Technol, 20(3), 215–232. https://doi.org/ 96–97). Covington, Kentucky. https://doi.org/10.1145/563340.
10.1007/s10956-010-9247-2. 563374
Kline, P. (1999). The handbook of psychological testing (2nd ed.). Vieira, C., Magana, A. J., Falk, M. L., & Garcia, R. E. (2017). Writing in-
London: Routledge. code comments to self-explain in computational science and engi-
Magana, A. J., & Mathur, J. (2012). Motivation, Awareness and neering education. ACM Transactions on Computing Education,
Perceptions of Computational Science. Computing in Science and 17(4), 1–21. https://doi.org/10.1145/3058751.
Engineering (CiSE). IEEE Computer Society, 14(1), 74–79. Weintrop, D., Beheshti, E., Horn, M., Orton, K., Jona, K., Trouille, L., &
Magana, A. J., Falk, L. M., & Reese, J. M. (2013). Introducing Discipline- Wilensky, U. (2016). Defining computational thinking for mathe-
Based Computing in Undergraduate Engineering Education. ACM matics and science classrooms. J Sci Educ Technol, 25(1), 127–147.
Transactions on Computing Education, 13(4), 1–22. https://doi.org/10.1007/s10956-015-9581-5.
Magana, A. J., Falk, M. L., Vieira, C., & Reese, M. J. (2016). A case Windschitl, M. A. (1995). Using computer simulations to enhance con-
study of undergraduate engineering students' computational literacy ceptual change: the roles of constructivist instruction and student
and self-beliefs about computing in the context of authentic prac- epistemological beliefs. (Doctor of Philosophy), Iowa State
tices. Computers in Human Behavior 61, 427–442. https://doi.org/ University Ames, Iowa.
10.1016/j.chb.2016.03.025. Wofford, J. (2009). K-16 computationally rich science education: A ten-
Mansbach, R., Ferguson, A., Kilian, K., Krogstad, J., Leal, C., Schleife, year review of the journal of science education and technology
A., et al. (2016). Reforming an undergraduate materials science (1998–2008). J Sci Educ Technol, 18(1), 29–36. https://doi.org/10.
curriculum with computational modules. J Mater Educ, 38(3–4), 1007/s10956-008-9127-1.
161–174. Yang, D., Streveler, R. A., Miller, R. L., Slotta, J. D., Matusovich, H. M.,
Mulop, N., Yusof, K. M., & Tasir, Z. (2012). A review on enhancing the & Magana, A. J. (2012). Using computer-based online learning
teaching and learning of thermodynamics. Procedia-Social and modules to promote conceptual change: helping students understand
Behavioral Sciences, 56, 703–712. https://doi.org/10.1016/j. difficult concepts in thermal and transport science. Int J Eng Educ,
sbspro.2012.09.706. 28(3), 686.

You might also like