Content uploaded by Pratim Sengupta
Author content
All content in this area was uploaded by Pratim Sengupta on Dec 05, 2019
Content may be subject to copyright.
Content uploaded by Pratim Sengupta
Author content
All content in this area was uploaded by Pratim Sengupta on Jan 31, 2018
Content may be subject to copyright.
Content uploaded by Pratim Sengupta
Author content
All content in this area was uploaded by Pratim Sengupta on Jan 28, 2018
Content may be subject to copyright.
Chapter 4
Toward a Phenomenology
of Computational Thinking in STEM
Education
Pratim Sengupta, Amanda Dickes, and Amy Farris
4.1 Introduction
In this chapter, we argue for an epistemological shift from viewing coding and
computational thinking as mastery over computational logic and symbolic forms to
viewing them as a more complex form of experience. Rather than viewing comput-
ing as regurgitation and production of a set of axiomatic computational abstractions,
we argue that computing and computational thinking should be viewed as discur-
sive, perspectival, material, and embodied experiences, among others. These expe-
riences include, but are not subsumed by, the use and production of computational
abstractions. We illustrate what this paradigmatic shift toward a more phenomeno-
logical account of computing can mean for teaching and learning STEM in K-12
classrooms by presenting a critical review of the literature, as well as by presenting a
review of several studies we have conducted in K-12 educational settings grounded
in this perspective.
Papert (1987) famously referred to technocentrism as the fallacy of referring all
questions about technology to the technology itself. A critical look at the history of
educational computing tells us that the research in this field has also been predom-
inantly technocentric in nature. Calls for taking into account the learners’experi-
ences as building blocks for deeper learning and the development of disciplinary
Partial support from NSF CAREER Award #115230 and the Imperial Oil Foundation is gratefully
acknowledged. All opinions are the author’s and not endorsed by funding agencies.
Version 5.0 (6 March 18)
P. Sengupta (*)
University of Calgary, Calgary, AB, Canada
e-mail: pratim.sengupta@ucalgary.ca
A. Dickes
Harvard University, Cambridge, MA, USA
A. Farris
The Pennsylvania State University, University Park, PA, USA
©Springer International Publishing AG, part of Springer Nature 2018
M. S. Khine (ed.), Computational Thinking in the STEM Disciplines,
https://doi.org/10.1007/978-3-319-93566-9_4
49
expertise in STEM certainly have been made (e.g., Papert 1980; DiSessa 2000).
However, the predominant effect of this call has also been technocentric in the sense
that it has resulted in the creation of a new genre of programming languages (e.g.,
LOGO, Scratch, NetLogo, StarLogo TNG, AgentSheets, ViMAP, CTSiM, etc.) and
microcontrollers (e.g., Arduino) designed to be easily usable for the “novice pro-
grammer.”The technocentric focus is also evident in the learning objectives and
assessment of computational thinking, which predominantly focus on the production
and use of computational abstractions (e.g., see the studies reviewed by Grover and
Pea (2013a)). Only a few, recent examples have focused on phenomenological
aspects of computational thinking, such as the centrality of discourse (Grover and
Pea 2013b; Farris and Sengupta 2014), the role of embodied reasoning (Francis et al.
2016), aesthetic experiences (Farris and Sengupta 2016) and the importance of
managing, rather than ignoring uncertainty (Farris et al. 2016) in the development
of computational thinking in STEM curricular contexts. And while recent arguments
have been made for an increased awareness for paying attention to sociological
dimensions of computing such as computing in public spaces (Sengupta and
Shanahan 2017), virtual communities (e.g., online Scratch communities) and
out-of-school, DIY makerspaces (Kafai and Burke 2013), our focus here is on the
K-12 public school classroom.
Our chapter is an argument for deepening and broadening the focus on the
phenomenology of computing and computational thinking in K-12 STEM curricular
contexts and classrooms. Our concerns are both epistemological and pedagogical
and are grounded historically as well as in the pragmatics of K-12 classrooms with
the focus on sustaining computing as a long-term practice. The first part of the
chapter presents a critical and synthetic review of the literature and argues for a
phenomenological approach toward developing an epistemology of computational
thinking that foregrounds the uncertainty and complexity in the experience of
computing and science, in professional practice and in STEM classrooms. The
second part of the chapter presents a set of pedagogical approaches for sustaining
computing and computational thinking through computational modeling in the
STEM classroom. This is presented in the form of a critical review of studies that
are conducted by our research group in K-12 classrooms in the USA, including
studies that were conducted in the form of partnerships with teachers.
4.2 The Need for a Phenomenology of Computational
Thinking
Since the phrase “computational thinking”has been popularized by Wing (2006),
there have been a plethora of studies on computational thinking in education. Yet,
beyond the early work on computational literacy by Papert (1980) and diSessa
(2001), the epistemology of computational thinking has received very little attention
in the literature. In this section, we examine core beliefs and assumptions about the
50 P. Sengupta et al.
nature of knowledge and knowing that are and should be involved in thinking
computationally, by adopting a historical perspective as well as by reviewing recent
research in and relevant to educational computing, from a phenomenological per-
spective. We highlight the importance of grounding computational thinking in
representational and epistemic practices that are central to knowing and doing in
science and, more broadly, in STEM education. The phenomenologist Merleau-
Ponty (1962)defined sense experience as “that vital communication with the world
which makes it present as a familiar setting of our life”(Merleau-Ponty 1962, pp 61).
We believe that thinking carefully in terms of these practices can help us understand
the materiality, uncertainty and subjectivity inherent in the students’and teachers’
sense experiences of computational thinking in STEM classrooms, for reasons we
explain in more detail next.
4.2.1 Inseparability of Abstractions and Practices
in Computing and Science
Citing a definition coined together with Jan Cuny of the National Science Founda-
tion and Larry Snyder of the University of Washington, Wing (2011)defined
“computational thinking”to indicate the “thought process involved in formulating
problems and their solutions so that the solutions are represented in a form that can
be effectively carried out by an information-processing agent [CunySnyderWing10]”
(Wing 2011, p 20). According to Wing, the “essence of computational thinking is
abstraction”(Wing 2008, pp 3717). She argued that computational thinking
involves dealing with abstractions in the following ways: (a) defining abstractions,
(b) working with multiple layers of abstraction, and (c) understanding the relation-
ships among the different layers (Wing 2008). Abstractions, according to Wing, give
computer scientists the power to scale and deal with complexity. She noted:
Abstraction is used in defining patterns, generalizing from instances, and parameterization. It
is used to let one object stand for many. It is used to capture essential properties common to a
set of objects while hiding irrelevant distinctions among them. (Wing 2011, p 20)
Wing’s conceptualization of abstraction, as the excerpt above shows, therefore,
emphasizes the notion of generalization. Abstractions, in her view, are generalized
computational representations that can be used (i.e., applied) in multiple situations or
contexts. In this sense, as Sengupta et al. (2013) pointed out, her definition of
abstraction is similar to Locke’s. In Locke’s view, abstraction is the process in
which “ideas taken from particular beings become general representatives of all of
the same kind”(Locke 1690/1979).
However, a phenomenological interpretation of Wing’s notion of abstractions is
incomplete without a deeper understanding of the contextualization that necessitates
and grounds computational abstractions in professional practice. For example, the
computer scientist and software engineering researcher Douglas C. Schmidt (2006)
points out that software researchers and developers typically engage in creating
4 Toward a Phenomenology of Computational Thinking in STEM Education 51
abstractions that help them program in terms of their contextualized design goals –
e.g., the specific problem that they are solving, which is often in a different field
(domain) of professional practice. The abstractions that “need”to be created are
essential because the end user must be shielded from avoidable complexities, such as
the CPU, memory, and network devices, and, instead, interact directly with the
domain-specific problem (Schmidt 2006). Similarly, it is important to note that even
Wing (2011) acknowledges the complexity of computing systems as resulting
from the material and physical constraints underlying the information-processing
agent and its operating environment. She argues that while considering what com-
putational thinking is, we must also “worry about boundary conditions, failures,
malicious agents and the unpredictability of the real world”(Wing 2011, pp 20).
We therefore believe that the term “thinking”in computational thinking is a
semantic reduction of its intended meaning. Phenomenologically, computational
thinking involves both representational and epistemic work that are also grounded
disciplinarily and materially. It is in this light that Sengupta et al. (2013) argued that
when the notion of computational abstractions is grounded in use, it could be
understood as a practice that draws upon concepts that are fundamental to computing
and computer science, and it also includes practices such as problem representation,
abstraction, decomposition, simulation, verification, and prediction that are also
central to modeling, reasoning, and problem-solving in a large number of scientific,
engineering, and mathematical disciplines (National Research Council 2007; NGSS
2015).
Sociologists and philosophers of science have also identified the inseparability of
abstractions and practice in the work of scientists. It is rarely the case that the
transformation of an initial idea to a successful scientific experiment or a model is
a simple and linear process that relies on solely the invention and use of abstractions.
The philosopher Andrew Pickering pointed out that scientists are always enmeshed
in a “mangle of practice”(Pickering 1995). That is, scientists struggle continuously
in order to get theories and instruments on one hand and the natural world on the
other to perform in the ways that their investigations require. The creation of
scientific knowledge can therefore be understood as a dynamical process of interac-
tive stabilization of material and human agency –a process that Pickering termed as
the dance of agency (Pickering 1995; see also Lehrer 2009). Uncertainty, and
managing uncertainty are unavoidable aspects in this work, even though the most
popular image of scientific work tends to be one of the certitude of accurate pre-
dictions (Duschl 2008).
A central focus of the scientific work is the invention, reproduction, and modi-
fication of scientific inscriptions –such as graphs, equations, computer code, etc. –
which tend to amplify certain aspects of the phenomena under investigation while
reducing emphasis on other, less relevant aspects (Latour 1990). This is similar and
synergistic to the work of defining and using contextually relevant computational
abstractions, as we pointed out earlier. Additionally, computational models can also
bring to light new, unexpected ways of thinking about the phenomena by bringing
different disciplinary perspectives in contact with one another (MacLeod and
Nersessian 2015). The process of creation of these inscriptions –which are
52 P. Sengupta et al.
collectively termed “modeling”–involves both representational and epistemic work
in a deeply intertwined manner (Giere 1988; Pickering 1995; Lehrer 2009). This
perspective is known as the “science as practice”perspective and is now regarded as
a cornerstone of science education research (NGSS 2015). In the following sub-
sections, we consider the subjective and perspectival nature of the work involved in
modeling, and in particular, computational modeling.
4.2.2 Subjectivity in Representational Work
Studies of scientists and their production of scientific inscriptions reveal a rather
amorphous nature of scientific knowledge and work (Pickering 1995; Ochs et al.
1996; Latour 1999; Daston and Galison 2007). For Pickering (1995), as we
mentioned in the previous section (Sect. 4.2.1), subjectivity arises from the dance
of agency between theory formulation and the materiality of the physical world.
Latour (1999) argues that while a common image of science implies objectivity
and certitude, viewing science as research can help us see it as a much more
complex experience –one that is uncertain and subjective, and both human and
non-human. Ochs et al. (1996) highlighted the central role that interpretive work,
including negotiation between scientists, plays in dealing with uncertainty during a
research project. They also demonstrated that the interpretive nature and uncertainty
of this work –an epistemic phenomenon –are deeply tied to the representational
infrastructure (Ochs et al. 1996). This is echoed by Daston and Galison (2007), who
pointed out that as representational technologies evolve and new representational
technologies emerge, they necessitate new forms of uncertainty and
interpretive work.
Daston and Galison (2007) argued that with the introduction of photographic
technology and the printing press, the epistemic stance of scientific work shifted
from a falsely “objectivitist”stance to “trained judgment.”This was evident in their
comparison between the nineteenth century introduction of photographic technology
where the machinic nature of photography created an impression that scientist could
“get out of the way”and let the photograph produce what became perceived as bare,
uninterpreted, objective “facts.”In contrast, beginning in the early to mid-twentieth
century, with the advent of the printing press that in turn widened the audience for
scientific works such as atlases, the production of scientific images became neces-
sarily more interpretive on the part of the scientist, with a clear goal of enhancing the
communicativity of the images, which Daston and Galison (2007) termed “trained
judgment.”
Building on this work, Farris et al. (2016) have argued that the advent of
computing as a key mode and medium of scientific inquiry further amplifies this
epistemic stance of “trained judgment.”A case to point, they argued, is that recent,
long-term ethnographic studies of biomedical engineering labs illustrate how the
malleability and inherent interdisciplinary of the practice of computational modeling
results in new conceptual innovations in scientific practice (Nersessian 2012;
4 Toward a Phenomenology of Computational Thinking in STEM Education 53
Chandrasekharan and Nersessian 2015). Nersessian and colleagues showed that
computational modeling can be particularly helpful for creating new scientific
knowledge in the field of complex systems, by (a) bridging the gap between
theorization, dynamic visualization, and experimental work, (b) bringing together
multiple disciplinary perspectives, (c) using stochastic modeling techniques in cases
where clear mechanistic accounts are difficult to obtain, and (d) making it possible to
communicate directly with colleagues about complex, predictive visualizations of
the target phenomena.
4.2.3 Computational Modeling as Perspectival Work
In his seminal book, Mindstorms, Papert argued that working with the LOGO turtle
is a “model for what it is to get to know an idea the way you get to know a person”
(Papert 1980, pp 136). Papert argued that it involves getting to know the turtle,
through exploring what it can or cannot do. He cautioned that this should not mean
that all ideas be reduced to computational terms; rather, the early experience with
turtles is a good model of learning. That is, “... it is a good way to ‘get to know’
subject by ‘getting to know’its powerful ideas”(Papert 1980, p 138). As an
illustrative case, he noted that when children learn Newtonian mechanics using
LOGO, they do so through modeling changing velocities, i.e., by specifying how
fast the turtle should move. The propositional forms of these phenomena are
represented in the form of physical laws in the form of linear mathematical equa-
tions, and the fallacy of education is that these laws which are the products of
complex work (i.e., Pickering’s mangle of practice) in which qualitative thinking
that is less completely specified and seldom stated in propositional form play an
important role. Therefore, it is the qualitative experience of thinking like the turtle
and thinking with the turtle that makes the experience of learning a powerful and a
deep one and one that is quite antithetical to learning as usual in K-12 science (and
beyond). These forms of reasoning enable the learner to engage in embodied and
intuitive reasoning (Papert 1980; Wilensky and Reisman 2006; Dickes et al. 2016b;
Sengupta and Wilensky 2009).
The early success of LOGO has led to the development of several LOGO-like
programming languages and modeling environments such as NetLogo (Wilensky
1999), Scratch (Resnick et al. 2009), AgentSheets (Repenning and Sumner 1995),
CTSiM (Sengupta et al. 2013; Basu et al. 2016), and ViMAP (Sengupta et al.
2015b). Computational models developed in such languages are more generally
known as agent-based models (ABMs). When users develop ABMs, they construct
programs by providing simple rules to a computational object or agent (e.g., the
sprite in Scratch, the turtle in LOGO, etc.), which then enacts the rules through
movement in computational space. These agent-level actions are repeated over time
and/or across multiple agents. In the former case, it enables learners to generate
models of continuous movement (Newtonian mechanics) from temporal aggrega-
tions of discrete actions (Sengupta and Farris 2012; Sengupta et al. 2012). In the
54 P. Sengupta et al.
latter case, it enables learners to model dynamical systems (e.g., ecological
interdependence) in which multiple agents are simultaneously interacting with
each other (Dickes and Sengupta 2013; Dickes et al. 2016b).
Because the agent-level interactions, attributes, and behaviors are often body-
syntonic (i.e., can be explained and understood through simple embodied actions of
the child), young children can model complex scientific phenomena using such
forms of computing (Papert 1980; Danish 2014; Dickes et al. 2016b; Levy and
Wilensky 2008). As Dickes et al. (2016b) demonstrated, by engaging in agent-based
modeling, even young learners can investigate and develop explanations of system-
level, emergent behaviors from the perspective of agents within the system. They
key argument supported by these studies is that thinking like the agent provides
learners an intuitive pathway in exploring emergent outcomes of the system
(Wilensky and Reisman 2006; Levy and Wilensky 2008). Evelyn Fox Keller’s
biography of the biologist Barbara McClintock supports this claim, citing evidence
that thinking like the agent (e.g., a chromosome) enabled McClintock to make
significant advances in her research on human genetic structures (Keller 1984).
Similarly, Ochs et al. (1996) also identified that scientists’sensemaking in the
domain of physical sciences also involves such mental projections of the self into
the phenomenon of inquiry.
4.3 Phenomenological Approaches for Sustaining
Computing in STEM Classrooms
What does the theoretical review in the preceding section mean for the praxis of
computing in STEM education? We argue that the experience of coding in STEM,
from the perspective of the learners and teachers, especially over a long period of
time, is inherently heterogeneous. That is, dealing with computational abstractions in
the context of STEM disciplinary contexts and classrooms involves engaging with
multiple forms and genres of representations beyond coding, and often translating
between these representations requires interpretive judgments. This stands in con-
trast to the views that have been more traditionally supported by educational
researchers, where the goal is to “apply”algorithmic thinking and computational
abstractions to determine the correct answer. This complexity is left out in
technocentric images of coding, even when they apparently focus on computational
productions by participants.
In the remainder of this section, we propose some phenomenological approaches
that can help us address these issues in the K-12 STEM classroom. We will review a
set of studies conducted in partnership with K-12 teachers and students. Participants
in these studies used coding in order to design and develop models in science and
math on a long-term basis, throughout the academic year. We present a close
examination of the nature of the experience through which teachers and students
4 Toward a Phenomenology of Computational Thinking in STEM Education 55
appropriated coding and computational thinking as the language of doing scientific
work in their classrooms. We begin with an argument for adopting a particular genre
of programming and modeling (agent-based programming and modeling) for
modeling across disciplines, which is essential for long-term curricular integration.
We then suggest a set of pedagogical guidelines for integrating programming in the
K-12 STEM curricula, grounded in the perspectives of teachers and learners in K-12
classrooms.
4.3.1 Agent-Based Computational Modeling
as a Transdisciplinary Practice
Scientific practices like modeling develop only over the long term, both historically
within the sciences and ontogenetically within the lifetime of individuals. This is
because modeling is a rather nuanced and complex form of epistemology, even
though most educational texts and curricula do not directly address these complex-
ities (Lehrer 2009). The yearlong science classroom is a better context for engaging
children in such extended forms of practice, rather than the predominant tradition in
educational research to conduct intervention studies where children engage in
modeling (including computational modeling) spanning a few hours to a few days.
But, in order to support such long-term curricular integration, we must take into
consideration how to integrate computational modeling and programming across
disciplinary contexts.
Different forms of phenomena lend themselves to different forms of modeling
(Lehrer and Schauble 2007), and we have found that at the elementary, middle, and
high school levels, the categories of linear continuity and emergent aggregation can
be helpful guides for us in selecting scientific phenomena across disciplines that can
lend themselves well to computational modeling and programming. An example of
modeling linear continuity would be modeling motion as a continuous change in
position, where the behavior of a single “agent”(e.g., a ball rolling on a ramp) can be
modeled as a temporal series of changes of position and/or other variables such as
speed and acceleration that obey linear mathematical relationships (Sherin et al.
1993; Sengupta and Farris 2012). An example of modeling emergent aggregation
would be modeling ecological interdependence, where multiple agents simulta-
neously interact with each other and the environment, which in turn result in
aggregate-level outcomes, e.g., the dynamical relationship between the predator
and prey populations in an ecosystem (Wilensky and Reisman 2006; Dickes and
Sengupta 2013; Wagh et al. 2017). Such aggregate-level behaviors or outcomes are
known as emergent, because although linear relationships between individual agents
(objects) produce these behaviors, these behaviors are not apparent in the description
of either the individual objects or the relationships (Lehrer and Schauble 2007;
Wilensky and Resnick 1999). Other examples of emergent phenomena that have
been successfully adopted by teachers and students through the use of agent-based
56 P. Sengupta et al.
computational modeling and programming include electrical conduction (Sengupta
and Wilensky 2011), crystallization (Blikstein and Wilensky 2009), molecular
chemistry (Stieff and Wilensky 2003), evolution (Dickes and Sengupta 2013;
Wagh et al. 2017), ethnocentrism (Hostetler et al. 2018), etc. This suggests that
adopting agent-based modeling and programming as the form of computing can
make it possible for educators to use the same genre of modeling and programming
across multiple disciplines.
It has also been argued that students’conceptual difficulties in understanding both
linear continuity and both emergent aggregation have similar origins (Reiner et al.
2000; Chi 2005). For example, Reiner et al. (2000) argued that physics novices tend
to use substance-based knowledge when reasoning about concepts like force, heat,
light, and electric current (e.g., force as a property of an object). For example, the
misconception that continuing motion implies a continued force in the direction of
the movement is generated from a more primitive idea (called phenomenological
primitives or p-prims) called “continuous force,”which can be abstracted from
common everyday experiences of needing constant effort to keep an object in motion
(DiSessa 1993). Note that these novice intuitive ideas about physics have an
underlying structure of a direct schema –one that involves an agent either acting
on another agent or an agent being acted upon by an impetus (Talmy 1983). On the
other hand, an expert-like understanding of kinematics involves being able to
conceptualize a situation in terms of more complex interactions –e.g., situations
involving lack of motion, or constant speed could be conceptualized as forms of
dynamic equilibrium between interacting systems (Clement 1993; Greeno and Van
De Sande 2007). Similarly, in the domain of ecology, researchers have argued that
commonly noted misconceptions are indicative of direct schema or event schema,
which imply a direct cause-effect relationship (such as “A”causes “B”) or an event
that has a finite duration of time (as opposed to being continuous), whereas the
expert conception of ecological phenomena involves a more complex cognitive
structure involving the dynamic and decentralized nature of emergent phenomena
in terms of a myriad of simultaneous interactions (Chi 2005). However, studies have
also shown that pedagogical approaches based on agent-based models and modeling
can act as productive learning environments, using which novice learners can
develop deep understandings of dynamic, aggregate-level phenomena by
bootstrapping, rather than discarding their agent-level intuitions (Dickes and
Sengupta 2013; Dickes et al. 2016b; Wilensky and Reisman 2006; Levy and
Wilensky 2008).
This body of research also provides useful guidelines for the sequence of learning
activities in each domain, and our general pedagogical approach explicitly adopts the
perspective that expert-like scientific knowledge can result through building upon
and refining existing naive intuitive knowledge (Dickes et al. 2016b; Danish 2014;
Sengupta et al. 2015b). For example, the initial learning activities leverage a naive
conceptualization of the domains and progressively scaffold them toward refine-
ment. In kinematics, learners begin by inventing representations of motion in terms
of measures of speed (how fast an object is moving) and inertia (innate tendency of
an object to continue its current state of rest or motion, which often takes an
4 Toward a Phenomenology of Computational Thinking in STEM Education 57
anthropomorphic form in novice reasoning), and gradually move to a force-based,
more canonical description of motion in subsequent activities (Sengupta and Farris
2012; Farris et al. 2016). Similarly, in ecology, students begin with programming the
behavior of single agents in the ecosystem and gradually develop more complex
programs for modeling the behavior and interaction of multiple species within the
ecosystem (Wilensky and Reisman 2006; Danish 2014; Sengupta et al. 2013; Dickes
et al. 2016b).
4.3.2 Framing Programming as Designing Mathematical
Measures of Change
Our studies have demonstrated that framing programming as “mathematizing”in the
science classroom can serve as a productive pedagogical approach for integrating
programming in the K-12 science classroom (Sengupta et al. 2013, Sengupta et al.
2015a,b,2018; Dickes et al. 2016a; Farris et al. 2016). In this approach, program-
ming is used in the context of creating computational models of scientific phenom-
ena through designing discrete mathematical representations of units of change, for
representing change over time. That is, the computational code created by students
serve to define a “unit”of measurement, which would then get repeated as the
program was “run”to produce the desired motion.
From the perspective of praxis in the K-12 science classroom in North American
public schools, this form of activity is of critical importance for classroom integra-
tion of computational modeling and programming. Teachers in US and Canadian
public schools who we have worked with have reported that interpreting and
constructing mathematical measures (e.g., units of measurement and graphs) are
areas where most of their students experience difficulties (e.g., see Sengupta et al.
2018). This is also of importance for US and Canadian public schools because
manipulating units is emphasized in standardized assessments (in the USA) and
the program of studies (in Canada), and therefore, teachers acknowledge this as an
important learning goal in their regular science classroom.
We see this as a great opportunity for integration of computational modeling and
programming in K-12 science classrooms. Our studies show that agent-based pro-
gramming and modeling can help students overcome conceptual challenges in
understanding linear continuity (e.g., kinematics; see Sengupta and Farris 2012)
and emergent aggregation (e.g., ecology; see Dickes et al. 2016b), through the
iterative design of measures of change over time. This is because the activity of
programming the behavior of agents requires the learners to define the event in
discrete measures (Sengupta et al. 2015b). The state of the simulation, at any instant,
represents a single event in the form of spatialized representations of agent actions
and interactions. To “run”the simulation, these events are repeated a number of
times specified by the user. By engaging in iterative cycles of building, sharing,
refining, and verifying computational models, students refine their understanding of
58 P. Sengupta et al.
what actions and interactions of agents represent an “event,”which are then
displayed on graphs. This enables students to define and explore different kinds of
units and see their simulation measured in those units (Farris et al. 2016) and even
merge computational modeling with artistic design (Sengupta et al. 2012).
4.3.3 Supporting Perspectival Work Through Embodied
Modeling
Research in science education suggests that the integration of ABMs in elementary
classrooms also benefits greatly from the use of other synergistic forms of modeling
such as embodied and physical modeling. Programming an agent involves learning
to think like the agent, because it can help students understand the relationship
between their code and the simulated output. In our studies, all teachers saw
embodied modeling as a valuable activity for teaching students how to think like
an agent. Embodied modeling introduces the students to the relevant computational
rules represented by the agent-based programming commands through embodied
interactions with the material world, and in doing so, helps them debug their
programs and deepens their understanding of the graphs in the simulations (Dickes
et al. 2016a,b).
Why are these different forms of modeling necessary? Science educators and
cognitive scientists have argued that embodied thinking is central to the development
of agent-based thinking and representational practices (Papert 1980; Goldstone and
Wilensky 2008; Wilensky and Reisman 2006). For example, in a recent study
conducted in a third-grade classroom, students began with an embodied modeling
activity of foraging behavior, followed by the generation of mathematical inscrip-
tions based on their embodied actions, and finally, conducted further inquiry of
interdependence in an ecosystem using two separate ABMs (Dickes et al. 2016b).
We found that the students recalled and built upon their embodied modeling
experiences as butterflies foraging for nectar (see Fig. 4.1), during their subsequent
interactions with the agent-based simulation of a butterfly-bird-flower ecosystem
(see Fig. 4.2). We also found that creating mathematical inscriptions (bar graphs) to
represent the data collected during the embodied modeling activity provided a
representational continuity between the embodied modeling activities and the
ABMs, as well as with previous representational forms that students used and
developed in their science and math classes prior to the study. And finally, we also
found that embodied modeling activities, especially in the case of modeling inter-
actions between different types of agents, must be designed so that students are able
to take on the perspectives of different types of agents, rather than prompting
students to take on the perspective of only one type of agent.
As students engaged iteratively in cycles of embodied modeling and graphing by
taking on the perspective of the agents in the system, and then modeled the same
phenomena using multi-agent-based NetLogo simulations, we found that they were
4 Toward a Phenomenology of Computational Thinking in STEM Education 59
able to develop progressively more complex forms of mechanistic explanations of
emergence. Mechanistic explanations focus on the processes that underlie cause-
effect relationships and thereby take into account how the activities of the constituent
components affect one another (Russ et al. 2008). In particular, we found that
learners were able to engage in a particular form of mechanistic reasoning that
Russ et al. (2008) termed chaining. During chaining, learners use knowledge
about the causal structure of the phenomena to make claims about what must have
happened previously to bring about the current state of things (backward chaining)
or what will happen next given that certain entities or activities are present now
(forward chaining).
This is an important finding from the perspective of computational thinking in the
context of science education, because this suggests that event-based programming
and modeling can support children in developing deep conceptual understandings of
complex scientific phenomena. Furthermore, this also suggests that focusing on
supporting the growth of students’mechanistic reasoning through modeling may
Fig. 4.1 Students participating in phase I’s embodied modeling activity
Fig. 4.2 Screenshots of the predator ABM (left) and watched energy ABM (right). Both models
were designed to actively recruit students’previous embodied modeling experiences shown in
Fig. 4.1
60 P. Sengupta et al.
be helpful for integrating computational thinking in science classrooms. As
Sengupta et al. (2013) identified, mechanistic reasoning in the domain of science
education is well aligned with algorithm design and complexity analysis in the
domain of computational thinking.
4.3.4 Refining Computational Modeling Through
Disciplinarily Grounded Classroom Norms
Our studies also illustrate that emphasizing mathematizing and measurement as key
forms of learning activities can help teachers meaningfully integrate programming as
a“language”of science (Dickes et al. 2016a; Sengupta et al. 2018). Long-term
studies of classroom integration of computional modeling and programming in the
science curricula has further shown that teachers can seamlessly accomplish this by
supporting the development of disciplinarily grounded classroom norms for devel-
oping and refining mathematical measures (Dickes et al. 2016a). Science educators
have shown that the iterative design of mathematical measures can result in deep
conceptual growth of students in elementary science, especially when these activities
are integrated throughout the curriculum over several months (Lehrer 2009). Lehrer
and colleagues have also shown that asking the question what counts as a “good”
model also needs to be established in classroom instruction as a norm, in order to
deepen students’engagement with scientific modeling in elementary grades. Fur-
thermore, similar to Cobb and his colleagues’work in the mathematics classroom
explained in the next paragraph (McClain and Cobb 2001; Yackel and Cobb 1996;
Cobb et al. 1992), these norms also follow shifts toward deeper disciplinary warrants
over time (Lehrer and Schauble 2006; Ford and Forman 2006; Lehrer et al. 2008). In
such classrooms, mathematical modeling becomes a meaning-making lens through
which the natural world can be systematized and described (Lehrer et al. 2001).
The specific genre of classroom norms that we have found to be at work in our
studies has been termed sociomathematical norms (McClain and Cobb 2001; Yackel
and Cobb 1996; Cobb et al. 1992). In a recent paper, we outlined and demonstrated
how the emphasis on developing and refining sociomathematical norms pertaining to
the design of mathematical measures of motion can help teachers seamlessly inte-
grate programming with science education in a third-grade classroom and how they
are taken up in students’work (Dickes et al. 2016b). Sociomathematical norms differ
from general social norms that constitute the classroom participation structure in that
they concern the normative aspects of classroom actions and interactions that are
specifically mathematical. These norms regulate classroom discourse and influence
the learning opportunities that arise for both the students and the teacher. As in the
work of Cobb and his colleagues (Yackel et al. 1991; Cobb et al. 1992), we also
found that it was the classroom teacher who initiated and guided the development of
4 Toward a Phenomenology of Computational Thinking in STEM Education 61
these norms in order to foster and sustain classroom microcultures characterized by
explanation, justification, and argumentation.
In our study (Dickes et al. 2016b), an important and rather fundamental
sociomathematical norm that began as the central guiding question posed by the
teacher at the beginning of the class was “what counts as a good model.”Similar to
Yackel and Cobb (1996), we found that this norm typically originated as a socially
defined norm and shifted over time to a more sociomathematically defined norm. That
is, students’initial warrants were decided on the basis on how many of their peers
“liked”a particular model during class discussion and sharing of models rather than
thinking more deeply about how their ViMAP code represented the relevant phenom-
enon they were modeling. However, over time, these warrants became progressively
more grounded in the mathematically warrants of how representative their code were
of the relevant phenomena being modeled. The class jointly took up normative ways
of thinking about and representing motion (walking) through designing and refining
approximate and predictive measures of change over time, using embodied modeling
activities, drawings of their embodied modeling activities that represented “step-
sizes”, and their ViMAP code and graphs (Dickes et al. 2016b).
Overall, we found that students’use of the ViMAP programming commands
became increasingly sophisticated as they held their models accountable to the
sociomathematical norms (Dickes et al. 2016a). Over a 6-week period, we scored
each student’sfinal ViMAP model at the end of each class period in terms of whether
they used appropriate computational abstractions identified by Sengupta et al. (2013)
as being relevant to computational thinking such as variables, loops, and initializa-
tion. Students’code was scored on the appropriate and non-redundant use of vari-
ables and loops in their models and whether their graphs represented appropriate
element(s) of the phenomenon being simulated using their ViMAP code. The growth
in students’computational fluency is evident in Fig. 4.3. For example, a score of zero
meant none of the variables used were appropriate, whereas a score of 3 meant no
3.000
2.500
2.000
1.500
1.000
0.500
0.000
5-Feb 10-Feb 15-Feb
Variables Accurate Graphs
20-Feb 25-Feb 2-Mar 7-Mar 12-Ma
r
Fig. 4.3 Improvement in computational thinking supported by sociomathematical norms
62 P. Sengupta et al.
use of redundant or incorrect variables. The accuracy of the graphs in students’later
models were indicative of the appropriate use of the “repeat”command (i.e., loops)
and order of placement of the “place measure”command. This in turn relied on their
conceptual understanding of when to initialize the measurement (i.e., initialization)
and how often the desired measurement had to be repeated in order to generate the
graph (loops).
4.3.5 Framing Coding as Designing for Authentic Use
In a study conducted in a fourth-grade classroom in a low-income (90% free lunch),
public charter school in Nashville, we investigated how collaboratively designing
computational machines for authentic users could support the integration of coding
in STEM education (Sengupta et al. 2015b). The first phase of the study focused on
introducing students to agent-based programming through creating geometric shapes
(e.g. squares, circles, spirals) using the ViMAP programming language (Sengupta
et al. 2015b). ViMAP uses the NetLogo modeling platform (Wilensky 1999) as its
simulation engine and enables learners to design, program and graph NetLogo
simulations using both programming blocks and text-based programming (see
Fig. 4.6). This phase lasted for eight class periods. For the next 18 class periods,
students worked in dyads on a STEM design challenge (capstone activity), i.e.,
constructing mathematical machines and user guides for generating geometric
shapes using a distributed computing infrastructure.
During the capstone learning activity, learners worked in dyads and constructed a
mathematical machine for generating geometric shapes. Each machine consists of
two components: virtual and physical. The virtual component was a ViMAP pro-
gram that learners constructed using visual programming primitives selected from
the ViMAP programming library. The physical component consisted of two physical
control interfaces, each designed to control the reading on one of the distance
sensors. Each sensor controlled a distinct turtle variable (e.g., color, speed, rotation).
This was an activity that required intersubjective collaboration (Sengupta et al.
2015b), because while each member of the dyad independently designed one of
these physical control structures using Lego bricks, the dyad was responsible for
jointly designing the ViMAP program. Figure 4.4 shows an example of student
work.
We specified that other fourth-grade teachers in Nashville would use these
machines, so that students had a specific image of user(s) in mind. To ensure
authenticity of the users, we also invited three graduate students in education with
prior math teaching experience in elementary grades, but unaffiliated with our study,
to serve as “users.”The user testing took place twice: first in mid-March (user testing
1) and in late April (user testing 2). During both the user testing events, each user
interacted with a dyad’s machine for about 20 min and provided them written and
verbal feedback. After user testing 1, students improved their machines and user
4 Toward a Phenomenology of Computational Thinking in STEM Education 63
guides in order to address the issues highlighted in the feedback. User testing 2 was
also the capstone activity.
We compared the work of each dyad at two stages: user testing 1 (UT1) and user
testing 2 (UT2). In terms of children’s mechanistic explanations (Russ et al. 2008),
we found that compared to UT1, attending to what the user needs to know resulted in
improving greatly the quality of students’mathematical explanations during UT2.
Their explanations, as evident both in their user guides and verbal explanations
during the user testing process, made explicit the mathematical relationships
between algorithmic elements (e.g., number of loops in their ViMAP program)
and variables in their ViMAP programs, and the actions of the turtle in every step
(e.g., right turn), which in turn was directly effected by the users’actions (e.g.,
sensor reading generated by the user). The greater emphasis on identifying and
representing the relationships between computational abstractions (algorithms and
variables), mathematical relationships, and the mechanics of the physical setup
resulted from the need to create designs that were more communicative (Sengupta
et al. 2015b). A sample comparison is shown in Fig. 4.5a.
The phenomenological lesson here is that when coding is embedded in an
authentic design activity intended for and tested by an authentic audience, paying
attention to the needs and the perspective of the user can deepen the coders’
conceptual understanding of the relationship of computational abstractions with
disciplinarily grounded knowledge and representations.
4.3.6 Support Transition from Visual to Text-Based
Programming
Another important issue for sustaining programming in K-12 STEM classrooms,
especially in the higher grades (middle school or high school), is that although visual
Fig. 4.4 (a) (Left) Jerry’s pulley mechanism for controlling turn of the turtle via 1. (b) (Middle)
Chuck’s machine for controlling the speed of the turtle via sensor 2. (c) (Right) is a screenshot of
their ViMAP program for generating a square, and our annotation makes explicit the multiplicative
reasoning involved in generating angles and sides of the square
64 P. Sengupta et al.
programming lowers the barrier for entry into programming, learners who intend to
pursue careers in computing may find the drag-and-drop nature of visual program-
ming inauthentic or find it difficult to transition to text-based programming (DiSalvo
2014). In a recent study conducted in an eighth-grade classroom, we investigated this
issue (Sengupta et al. 2015b). We used ViMAP, because ViMAP is a dual-mode
programming language that enables users to engage in both blocks and text-based
programming. Visual programming commands in ViMAP are defined as short
NetLogo procedures (see Fig. 4.6), which students can easily access and modify
using text-based NetLogo code. In our study, after engaging in visual programming
with ViMAP for approximately 2 months to build simulations of interdependence in
ant ecosystems, the teacher and the students wanted to make deeper changes in the
underlying text-based NetLogo code. But, given the limited instructional time, the
teacher found it challenging to help students create new simulations in NetLogo
Fig. 4.5b A schematic for mechanistic explanations used by all groups in user testing 2
Fig. 4.5a Jacinda and Tom’s user guides in user testing 1 (left) and user testing 2 (right). We
annotated their user guides using the schematic shown in Fig. 4.5b
4 Toward a Phenomenology of Computational Thinking in STEM Education 65
using text-based programming. This required a lot of “overhead,”because the
language syntax was often disconnected from the relevant scientific concepts.
To address this issue, the teacher then decided to return to the ViMAP-Ants unit
(see www.vimapk12.net for the curricular activities) and asked the students to work
in small groups to create new ViMAP commands by modifying and extending the
underlying NetLogo code. For the eighth graders, this work was motivated by a
capstone project of designing and creating a version of ViMAP-Ants in order to
teach fourth graders about food webs, which is a required curricular standard in
fourth grade. The teacher introduced the students to relevant “chunks”(procedures)
in the NetLogo code pertaining to specific ViMAP commands they were already
familiar with. She led class discussions in which the students collaboratively
interpreted and explained the significance of the computational abstractions in
NetLogo code in terms of relevant scientific concepts represented in the ViMAP
commands. Learning the syntax and new forms of abstractions (such as classes) in
text-based programming therefore became deeply intertwined with the relevant
concepts in ecology (e.g., hierarchy of organisms in food webs). While students’
previous work using visual programming introduced them to computational abstrac-
tions such as loops, variables, and conditionals, text-based programming further
deepened their computational thinking because it involved creating computational
objects or classes, declaring new local variables, creating and modifying condi-
tionals, editing and repurposing lists, and using random numbers. Students’growth
in computational thinking was further evident in a post-assessment activity, in which
they successfully created new commands for a NetLogo simulation of a different
ecosystem without teacher-provided assistance (Sengupta et al. 2015b).
Fig. 4.6 ViMAP-Ant-Foodweb simulation and programming commands developed by eighth
graders. Popped-out images show NetLogo procedures underlying the ViMAP commands created
by the students. Left to right, graphs of population and energy, library of ViMAP commands, a
sample ViMAP program, and the NetLogo simulation controlled by the ViMAP program
66 P. Sengupta et al.
4.4 Discussion: Computational Thinking as Experience
in K-12 STEM
In Quest for Certainty, John Dewey argued against empiricist ontology that sub-
stitutes data for objects (and inquiry). Data, he argued, signifies a phenomenon for
further inquiry; but instead, empiricism often represents data as being self-evident
(Dewey 1929 (1984)). In a similar vein, the persistent fallacy of the predominant
epistemology in educational computing, especially as it pertains to computational
thinking in education, is the normative notion that knowledge is some antecedent
reality (Dewey 1929 (1984)), reified in terms of learners’use of computational
abstractions used commonly by professional coders. That is, for researchers, the
experience of learners is substituted by canonical assessments of the “computational
abstractions”that the learners used in their computer programs. Certainly, there are
efforts, especially by constructionist scholars, to demonstrate how computing can
take on diverse and personally meaningful forms (Resnick et al. 2000; Farris and
Sengupta 2016), but the hallmark of the experience of coding, as reported in nearly
all research articles on computational thinking (including some of our own previous
work), remains the deft use of computational abstractions by learners who haven’t
had much prior experience with coding. This is the danger of technocentrism (Papert
1987) realized –where the questions about technology are being answered by
referring the questions to the technology itself.
In this chapter, we have argued for paradigmatic shift in the epistemology and
pedagogy of computing and computational thinking, especially for K-12 STEM
education. Our position is that we must shift away from empiricist ontology that
Dewey argued against (Dewey 1929 (1984)), and technocentrism that Papert argued
against (Papert 1987), toward more phenomenological perspectives, in terms of
trying to both understand and support the development of computational thinking
as experience in the context of K-12 STEM education. Epistemologically, we have
argued that computational thinking must be reconceptualized more appropriately as
an intersubjective experience, as opposed to more cognitivist and technocentric
images of learning and reasoning that can be assessed through the production of
symbolic code. Contextualizing computational abstractions in K-12 science class-
rooms is a complex experience that can be better understood as a “phenomenal field”
(Merleau-Ponty 1962), rather than by simply focusing on a cognitivist image of
“thinking”. This experience is rife with uncertainty and involves significant instruc-
tional work. For example, even in short-term studies, Sengupta et al. (2013) and
Basu et al. (2016) acknowledge and highlight the importance of extensive scaffold-
ing provided by facilitators in order to help students in overcoming challenges in
designing and using the necessary computational abstractions for modeling kine-
matics and ecology. However, despite such efforts, a commonly used approach of
assessing computational thinking relies primarily (and in many cases, only) on eval-
uating structural elements of learners’computer programs (e.g., Grover et al. 2018;
Weintrop et al. 2018; Dasgupta et al. 2016). Pedagogically, we have argued that
addressing this issue necessitates careful considerations of the complexities of K-12
4 Toward a Phenomenology of Computational Thinking in STEM Education 67
classrooms, without ignoring teachers’and students’experiences in which comput-
ing and coding are situated. In particular, we have proposed the following pedagog-
ical guidelines for sustaining computational thinking in the K-12 classroom:
1. Reframing programming and coding as “modeling,”in particular, as the design of
mathematical units of measurement of change over time, for the K-12 science
classroom;
2. Highlighting transdisciplinary representational and epistemic practices such as
design and modeling to support continuity in learning experiences across
disciplines;
3. Designing complementary activities that use embodied modeling and
non-computational materials as representational and cognitive amplifications of
computational code;
4. Focusing on disciplinarily grounded, normative instructional approaches (e.g.,
sociomathematical norms) during classroom instruction for refining computa-
tional modeling;
5. Reframing coding and modeling as designing for an authentic audience; and
6. Using both visual (block-based) and text-based programming languages for
longer-term curricular integration.
This list is far from exhaustive. However, given the context in which most of our
studies have been carried out –high-poverty, predominantly nonwhite classrooms in
public schools with limited resources –we believe that these guidelines can help us
focus our attention on issues that can make a difference in terms of helping teachers
and students adopt computing and computational thinking as a “language”of STEM,
especially on a long-term basis.
References
Basu, S., Biswas, G., Sengupta, P., Dickes, A., Kinnebrew, J. S., & Clark, D. (2016). Identifying
middle school students’challenges in computational thinking-based science learning. Research
and Practice in Technology Enhanced Learning, 11(1), 13.
Blikstein, P., & Wilensky, U. (2009). An atom is known by the company it keeps: A constructionist
learning environment for materials science using agent-based modeling. International Journal
of Computers for Mathematical Learning, 14(2), 81–119.
Chandrasekharan, S., & Nersessian, N. J. (2015). Building cognition: The construction of compu-
tational representations for scientific discovery. Cognitive Science, 39, 1727–1763.
Chi, M. T. (2005). Commonsense conceptions of emergent processes: Why some misconceptions
are robust. The Journal of the Learning Sciences, 14(2), 161–199.
Clement, J. (1993). Using bridging analogies and anchoring intuitions to deal with students’
preconceptions in physics. Journal of Research in Science Teaching, 30(10), 1241–1257.
Cobb, P., Wood, T., Yackel, E., & McNeal, B. (1992). Characteristics of classroom mathematics
traditions: An interactional analysis. American Educational Research Journal, 29(3), 573–604.
Danish, J. A. (2014). Applying an activity theory lens to designing instruction for learning about the
structure, behavior, and function of a honeybee system. The Journal of the Learning Sciences,
23(2), 100–148.
68 P. Sengupta et al.
Dasgupta, S., Hale, W., Monroy-Hernández, A., & Hill, B. M. (2016, February). Remixing as a
pathway to computational thinking. In Proceedings of the 19th ACM Conference on Computer-
Supported Cooperative Work & Social Computing (pp. 1438–1449). ACM.
Daston, L., & Galison, P. (2007). Objectivity. New York: Zone Books.
Dewey, J. (1929 (1984)). The later works of John Dewey 1929: The quest for certainty. SIU Press.
Dickes, A. C., & Sengupta, P. (2013). Learning natural selection in 4th grade with multi-agent
based computational models. Research in Science Education, 43(3), 921–953.
Dickes, A. C., Farris, A. V., & Sengupta, P. (2016a). Integrating agent-based programming with
elementary science: The role of socio-mathematical norms. In: Proceedings of the 12th Inter-
national Conference of Computers in Education (ICCE 2016), pp 128 –138.
Dickes, A., Sengupta, P., Farris, A. V., & Basu, S. (2016b). Development of mechanistic reasoning
and multi-level explanations in 3rd grade biology using multi-agent based models. Science
Education, 100(4), 734–776.
DiSalvo, B. (2014). Graphical qualities of educational technology: Using drag-and-drop and text-
based programs for introductory computer science. IEEE Computer Graphics and Applications,
34(6), 12–15.
DiSessa, A. A. (1993). Toward an epistemology of physics. Cognition and Instruction, 10(2–3),
105–225.
DiSessa, A. A. (2001). Changing minds: Computers, learning, and literacy. Cambridge, MA: MIT
Press.
Duschl, R. (2008). Science education in three-part harmony: Balancing conceptual, epistemic, and
social learning goals. Review of Research in Education, 32(1), 268–291.
Farris, A. V., & Sengupta, P. (2014). Perspectival computational thinking for learning physics: A
case study of collaborative agent-based modeling. In: Proceedings of the 12th international
conference of the learning sciences (pp. 1102–1107).ICLS. 2014.
Farris, A. V., & Sengupta, P. (2016). Democratizing children's computation: Learning computa-
tional science as aesthetic experience. Educational Theory, 66(1–2), 279–296.
Farris, A. V., Dickes, A. C., Sengupta, P. (2016). Development of disciplined interpretation using
computational modeling in the elementary science classroom. In Proceedings of the 13th
international conference of the learning sciences (pp. 282–290).ICLS. 2016.
Ford, M. J., & Forman, E. A. (2006). Redefining disciplinary learning in classroom contexts.
Review of Research in Education, 30(1), 1–32.
Francis, K., Khan, S., & Davis, B. (2016). Enactivism, spatial reasoning and coding. Digital
Experiences in Mathematics Education, 2,1–20.
Giere, R. N. (1988). Explaining science: A cognitive approach. Chicago: University of Chicago
Press [RNG].
Goldstone, R. L., & Wilensky, U. (2008). Promoting transfer by grounding complex systems
principles. The Journal of the Learning Sciences, 17(4), 465–516.
Greeno, J. G., & Van De Sande, C. (2007). Perspectival understanding of conceptions and
conceptual growth in interaction. Educational Psychologist, 42(1), 9–23.
Grover, S., & Pea, R. (2013a). Computational thinking in K–12: A review of the state of the field.
Educational Researcher, 42(1), 38–43.
Grover, S., & Pea, R. D. (2013b). Using a discourse-intensive pedagogy and android’s app inventor
for introducing computational concepts to middle school students. In Proceedings of the 44th
ACM technical symposium on computer science education (SIGCSE ‘13). New York: ACM.
Grover, S., Basu, S., & Schank, P. (2018, February). What we can learn about student learning
from open-ended programming projects in middle school computer science. In Proceedings of
the 49th ACM Technical Symposium on Computer Science Education (pp. 999–1004). ACM.
Hostetler, A., Sengupta, P., & Hollett, T. (2018). Unsilencing critical conversations in social-studies
teacher education using agent-based modeling. Cognition and Instruction, 36(2), 139–170.
4 Toward a Phenomenology of Computational Thinking in STEM Education 69
Kafai, Y. B., & Burke, Q. (2013). The social turn in K-12 programming: Moving from computa-
tional thinking to computational participation. In Proceeding of the 44th ACM technical
symposium on computer science education (pp. 603–608). ACM.
Keller, E. F. (1984). A feeling for the organism, 10th aniversary edition: The life and work of
Barbara McClintock. London: Macmillan.
Latour, B. (1990). Drawing things together. In M. Lynch & S. Woolgar (Eds.), Representation in
scientific practice (pp. 19–68). Cambridge: MIT Press.
Latour, B. (1999). Pandora’s hope: essays on the reality of science studies. Cambridge: Harvard
University Press.
Lehrer, R. (2009). Designing to develop disciplinary dispositions: Modeling natural systems.
American Psychologist, 64(8), 759.
Lehrer, R., & Schauble, L. (2006). Cultivating model-based reasoning in science education. In R. K.
Sawyer (Ed.), The Cambridge handbook of the learning sciences (pp. 371–387). New York:
Cambridge University Press.
Lehrer, R., & Schauble, L. (2007). Contrasting emerging conceptions of distribution in contexts of
error and natural variation. In M. Lovett & P. Shah (Eds.), Thinking with data (pp. 149–176).
Mahwah: Lawrence Erlbaum Associates.
Lehrer, R., Schauble, L., Strom, D., Pligge, M. (2001). Similarity of form and substance: Modeling
material kind. In Cognition and instruction: Twenty-five years of progress (pp. 39–74).
Lehrer, R., Schauble, L., & Lucas, D. (2008). Supporting development of the epistemology of
inquiry. Cognitive Development, 23(4), 512–529.
Levy, S. T., & Wilensky, U. (2008). Inventing a “mid level”to make ends meet: Reasoning between
the levels of complexity. Cognition and Instruction, 26(1), 1–47.
Locke, J. (1690/1979). An essay concerning human understanding. New York: Oxford University
Press.
MacLeod, M., & Nersessian, N. J. (2015). Modeling systems-level dynamics: Understanding
without mechanistic explanation in integrative systems biology. Studies in History and Philos-
ophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical
Sciences, 49,1–11.
McClain, K., & Cobb, P. (2001). An analysis of development of sociomathematical norms in one
first-grade classroom. Journal for Research in Mathematics Education, 32(3), 236–266.
Merleau-Ponty, M. (1962). Phenomenology of perception. New York: Routledge.
National Research Council. (2007). Taking science to school: Learning and teaching science in
grades K-8. Washington, DC: National Academies Press.
Nersessian, N. J. (2012). Modeling practices in conceptual innovation: An ethnographic study of a
neural engineering research laboratory. In U. Feest & F. Steinle (Eds.), Scientific concepts and
investigative practice (pp. 245–269). Berlin: DeGruyter.
Next Generation Science Standards. (2015). www.nextgenscience.org/next-generation-science-
standards
Ochs, E., Gonzales, P., & Jacoby, S. (1996). “When I come down I’m in the domain state”:
Grammar and graphic representation in the interpretive activity of physicists. Studies in Inter-
actional Sociolinguistics, 13, 328–369.
Papert, S. (1980). Mindstorms: Children, computers, and powerful ideas. New York: Basic Books.
Papert, S. (1987). Information technology and education: Computer criticism vs. technocentric
thinking. Educational Researcher, 16(1), 22–30.
Pickering, A. (1995). The mangle of practice: Time, agency, and science. Chicago: University of
Chicago Press.
Reiner, M., Slotta, J. D., Chi, M. T., & Resnick, L. B. (2000). Naive physics reasoning: A
commitment to substance-based conceptions. Cognition and Instruction, 18(1), 1–34.
Repenning, A., & Sumner, T. (1995). Agentsheets: A medium for creating domain-oriented visual
languages. Computer, 28(3), 17–25.
70 P. Sengupta et al.
Resnick, M., Berg, R., & Eisenberg, M. (2000). Beyond black boxes: Bringing transparency and
aesthetics back to scientific investigation. The Journal of the Learning Sciences, 9(1), 7–30.
Resnick, M., Maloney, J., Monroy-Hernández, A., Rusk, N., Eastmond, E., Brennan, K., Millner,
A., Rosenbaum, E., Silver, J., Silverman, B., & Kafai, Y. (2009). Scratch: programming for all.
Communications of the ACM, 52(11), 60–67.
Russ, R., Scherr, R., Hammer, D., & Mikeska, J. (2008). Recognizing mechanistic reasoning in
student scientific inquiry: A framework for discourse analysis developed from philosophy of
science. Science Education, 93, 875–891.
Schmidt, D. C. (2006). Model-driven engineering. Computer, 39(2), 25–31.
Sengupta, P., & Farris, A. V. (2012). Learning kinematics in elementary grades using agent-based
computational modeling: a visual programming-based approach. In Proceedings of the 11th
international conference on interaction design and children (pp. 78–87). ACM.
Sengupta, P., & Shanahan, M. C. (2017). Boundary play and pivots in public computation:
New directions in STEM education. International Journal of Engineering Education, 33(3),
1124–1134.
Sengupta, P., & Wilensky, U. (2009). Learning electricity with NIELS: Thinking with electrons
and thinking in levels. International Journal of Computers for Mathematical Learning, 14(1),
21–50.
Sengupta, P., & Wilensky, U. (2011). Lowering the learning threshold: Multi-agent-based models
and learning electricity. In Models and modeling (pp. 141–171). Cham: Springer.
Sengupta, P., Farris, A. V., & Wright, M. (2012). From agents to continuous change via aesthetics:
Learning mechanics with visual agent-based computational modeling. Technology, Knowledge
and Learning, 17(102), 23–40.
Sengupta, P., Kinnebrew, J., Basu, S., Biswas, G., & Clark, D. (2013). Integrating computational
thinking with K12 science education using agent-based modeling: A theoretical framework.
Education and Information Technologies, 18, 351–380.
Sengupta, P., Krishnan, G., Wright, M., & Ghassoul, C. (2015a). Mathematical machines &
integrated STEM: An intersubjective constructionist approach. Communications in Computer
and Information Science, 510, 272–288.
Sengupta, P., Dickes, A. C., Farris, A. V., Karan, A., Martin, K., & Wright, M. (2015b). Program-
ming in K12 science classrooms. Communications of the ACM, 58(1), 33–35.
Sengupta, P., Brown, B., Rushton, K., & Shanahan, M. C. (2018). Reframing coding as “Mathe-
matization”in the K12 classroom: Views from teacher professional learning. Alberta Science
Education Journal, 45(2), 28–36.
Sherin, B., diSessa, A. A., & Hammer, D. (1993). Dynaturtle revisited: Learning physics through
collaborative design of a computer model. Interactive Learning Environments, 3(2), 91–118.
Stieff, M., & Wilensky, U. (2003). Connected chemistry—Incorporating interactive simulations
into the chemistry classroom. Journal of Science Education and Technology, 12(3), 285–302.
Talmy, L. (1983). How language structures space. In H. Pick & I. Acredelo (Eds.), Spatial
orientation: Theory, research, and application. New York: Plenum Press.
Wagh, A., Cook-Whitt, K., & Wilensky, U. (2017). Bridging inquiry-based science and
constructionism: Exploring the alignment between students tinkering with code of computa-
tional models and goals of inquiry. Journal of Research in Science Teaching, 54(5), 615–641.
Weintrop, D., Afzal, A., Salac, J., Francis, P., Li, B., Shepherd, D. & Franklin, D. (2018).
Evaluating CoBlox: A Comparative Study of Robotics Programming Environments for Adult
Novices. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems
(CHI’18). pp. 366:1–12. Montreal QC, Canada: ACM Press.
Wilensky, U. (1999). NetLogo. http://ccl.northwestern.edu/netlogo/. Center for Connected Learn-
ing and Computer-Based Modeling, Northwestern University, Evanston.
Wilensky, U., & Resnick, M. (1999). Thinking in levels: A dynamic systems approach to making
sense of the world. Journal of Science Education and technology, 8(1), 3–19.
4 Toward a Phenomenology of Computational Thinking in STEM Education 71
Wilensky, U., & Reisman, K. (2006). Thinking like a wolf, a sheep, or a firefly: Learning biology
through constructing and testing computational theories –an embodied modeling approach.
Cognition and Instruction, 24(2), 171–209.
Wing, J. M. (2006). Computational thinking. Communications of the ACM, 49(3), 33–35.
Wing, J. M. (2008). Computational thinking and thinking about computing. Philosophical Trans-
actions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 366(1881),
3717–3725.
Wing, J. (2011). Computational thinking––What and why? The Link: News from the School of
Computer Science., 6,20–23.
Yackel, E., & Cobb, P. (1996). Sociomathematical norms, argumentation, and autonomy in
mathematics. Journal for Research in Mathematics Education, 22(4), 390–408.
Yackel, E., Cobb, P., & Wood, T. (1991). Small-group interactions as a source of learning
opportunities in second-grade mathematics. Journal for Research in Mathematics Education,
22(5), 390–408.
72 P. Sengupta et al.