Journal of Science Education and Technology

https://doi.org/10.1007/s10956-019-09772-0

Methodology and Epistemology of Computer Simulations

and Implications for Science Education
Maria Develaki 1

# Springer Nature B.V. 2019

Abstract
While computer simulations are a key element in understanding and doing science today, their nature and implica-
tions for science education have not been adequately explored in the relevant literature. In this article, (1) we
provide an analysis of the methodology and epistemology of computer simulations, aiming to contribute to a sound
and comprehensive account of the nature of computer simulations in science education, and (2) examine certain
implications for science education, particularly in terms of contemporary educational goals relating to scientific
literacy. We describe methodological elements relating to processes, techniques, and skills required for the construc-
tion and evaluation of scientific simulations, and we discuss epistemological views of their reliability and epistemic
status based on the relevant philosophical views. We then examine implications of these elements for the use of
simulations and especially for supporting scientific practices in the classroom and the corresponding educational
goals. Concretely, we compare educational simulations with those used in scientific research and with laboratory
experiments, we discuss the question of the reliability of simulations used in teaching or in public information, and
we give examples of their use for supporting NOS understanding and reasoning abilities. Finally, in the context of
the philosophical discourse about scientific realism, we examine implications of the epistemology of models that
concern the conception of the relation between scientific claims and the real world, which constitutes a fundamental
epistemological basis for teaching the nature of science.

Keywords Computer simulation models . Methodology of computer simulating . Epistemology of computer simulations .
Educational use of computer simulations . Nature of science . Scientific perspectivism . Model-based reasoning

Introduction Zeidler et al. 2005; NGSS Lead States 2013). Models

and modelling are crucial to the promotion of these
Basic aims of scientific literacy are the acquisition of a goals, since they are fundamental elements of both sci-
body of scientific knowledge, a set of scientific skills entific methods and scientific knowledge. Model-based
and ways of thinking, and an understanding of the na- teaching is the educational approach that was developed
ture of science (NOS), plus the ability to use them in in line with this thinking: that is, it bases the acquisition
addressing science-related problems of the age (e.g. of knowledge, abilities, and epistemologies on the fact
AAAS 1993; NRC 1996; Bybee 1997; Bybee 1997; that models are fundamental constituents of scientific
practice and scientific knowledge.
There is now a significant body of research on the
* Maria Develaki potential of model(ing)-based teaching and learning, on
[email protected]
the changes to curricula and teaching methods needed to
1
Educational Advice for Secondary Science Education, Hellenic
promote it (e.g. Halloun 2004; Clement and Rea-Ramirez
Ministry of Education, Athens, Greece 2008; Aduriz-Bravo and Izquierdo-Aymerich 2009;
 J Sci Educ Technol

Αduriz-bravo 2013; Develaki 2007, 2016; Gilbert and differential equations. Solving these equations yields the prog-
Justi 2016), and on the proper perceptions of models ress of the modelled real system, that is, the evolution of the
which must underlie these changes if they are to be effec- values of its variables in time and space—in other words, the
tive.1 Based on the model-based view2 in the philosophy model’s predictions. However, the precise mathematical solu-
of science, we might, in brief, say that models are ideal- tion, the analytical solution, can be found only for some forms
ized and perspectival conceptions of real systems and are of differential equation, the sort usually found in simple, high-
constructed and modified with the object of providing ly idealized models. More realistic models of phenomena
interpretations and predictions about phenomena that are (more dimensions of representation, more variables and pa-
compatible with existing theoretical knowledge and em- rameters, etc.), and especially models of very complex phe-
pirical data and that in order to fulfil this function, model- nomena and processes, such as astrophysical systems, the
ling is based on continuous interaction between theoriz- global climate or mental functions, are based on differential
ing, experimenting, and justificatory arguments. equations for which frequently an analytical solution cannot
Models are expressed and published in different forms (see, be found. This fact, for many years an impediment to the
e.g. Giere 1999; Gilbert and Justi 2016): there are, for exam- further practical utilization of theories and theoretical models,
ple, material models (scale models), visual models (pictures, has been addressed through approximative numerical solu-
sketches, diagrams), and more abstract models, such as theo- tions of the differential equations. This tackles the problem
retical models and computer simulation models. Theoretical of solving the differential equations by reducing them to,
models are closely associated with the fundamental theories and then solving, an immense number of simple equations, a
relevant to the modelled systems (Stöckler 1995; Giere 1999; computational task that, however simple the equations, is
Winsberg 2010), e.g. classical mechanics, quantum mechan- practically impossible with pencil and paper. The answer
ics, or the theory of evolution (and are thus mathematical if the came with computers (Gramelsberger 2010), which provided
theory is mathematicalized). This means that they fulfil the this computational capacity and thus made possible the radical
basic equations and principles of the theories but also addi- methodological advances in scientific modelling introduced
tional specific functions that specialize and adapt the general with computer simulation models beginning in the middle of
and abstract structure of the theories for application to specific the twentieth century. As we shall explain in greater detail in
categories of real systems and problems (see, e.g. Grandy ‘Methodology of Computer Simulations’, computer simula-
1992; Giere 1999).3 tions are basically theoretical models whose equations are
Theoretical–mathematical models, the most basic and use- solved numerically and which are thus converted into algo-
ful models in scientific research, are usually built on rithms and codes via a series of special practices and tech-
niques that allow them to be built, run, and processed on the
1
computer. Computer simulating thus permits the construction
The research on modelling-based teaching suggests that modelling activities
of more realistic models and models of very complex or inac-
can enhance the acquiring of knowledge, abilities, and epistemologies that
reflect real science (e.g. Schwarz & White 2005; Gilbert & Justi 2016) and cessible phenomena.
notes the conditions for the success of model-based teaching, such as the the Computer simulations now find application in an increas-
correct conception of models and modelling (e.g. Justi § Gilbert 2003; Oh & ing number of fields: astrophysics and nanophysics, quantum
Oh 2011) and the appropriate learning environments, e.g. the use of
simulation-based software (e.g. Dorn 1975; Andaloro et al. 1991; Mellar chemistry and molecular biology, meteorology/climatology
et al. 1994; Develaki 2017). and engineering, economics, and sociology. The increasingly
2
The model-based view (in its original version, the semantic view) based its widespread application of computer simulations to more and
analysis on the central role of models in scientific research and highlighted
many aspects of their nature and functions. It followed the statement view more different scientific fields and their effects on methodol-
predominating at the beginning of the twentieth century, which sees scientific ogy and the general culture of scientific practice make them a
theories as sets of lingustic statements (principles, axioms) that directly descibe key element in any understanding of contemporary science
the real world (see, e.g. Losee 1990; Giere 1999). The model-based view sees
scientific theories not as sets of statements but as sets of models that mediate
and a topic of interest in various science-related fields, e.g.
the application of the theories to complex real-world systems, which means the philosophy and sociology of science, science policy, and
that the theories are absolutely valid only for the models, the simplified ver- science education.
sions of the real systems, because the real systems are too complex and the
theories too general and abstract for immediate application to them. (Some
Computer simulations can also be a powerful teaching and
basic accounts of this view are given, e.g. by Suppe 1977; van Fraasen1980; learning aid if their design and use are linked to contemporary
Cartwright 1983; Morrison and Morgan 1999; Giere 1988, 1999; Knuuttila educational approaches and goals, in contrast to the initial use
2011.)
3 of simulation-based software, which focused mainly on the
Giere (1988, 1999) interpreted this conception for the case of classical me-
chanics: for example, the fundamental equation F = ma is valid for all the development of skills for its operation and on supporting the
theoretical kinematic models of Newtonian mechanics, but the function of the direct instruction model (de Jong 2006; Jimoyiannis 2010).
force is specialized differently in each model, that is, as F = ct, or F = k/r2, or F Indeed, one particularly interesting finding of the research
= − Dx for the models of rectilinear and curvilinear motion and harmonic
oscillation respectively. (F is the force, m is the mass, a is the acceleration, r on the impact of simulations on science teaching and learning
is the radial distance, and k and D are constants). is that the learning affordances of technology-based
J Sci Educ Technol

environments can be enhanced, especially when they are used educational simulations and enhancing their pedagogical
in innovative didactical approaches (Webb 2005). In this re- use, e.g. for supporting scientific practices in the classroom
search, however, and in the science education literature on and the corresponding contemporary educational goals; with
models and modelling generally, there is a deficit as regards regard to the latter point, we give some examples of their use
the nature of computer simulation models and their implica- in supporting students’ NOS understanding and reasoning–
tions for science education, although they represent a major evaluating abilities. In ‘Epistemological Views Concerning
development in contemporary science and engineering and a the Relation of Scientific Knowledge to Reality—
powerful medium for supporting advanced educational goals. Implications from the Epistemology of Models’, we examine
Articles on the nature and the epistemology of simulations implications of the epistemology of models in relation to the
(e.g. Dorn 1975; Tala 2011; Greca et al. 2014) are rare, and epistemological conception of the relation between scientific
in general the topic of models is rarely addressed, and then claims and the real world in science education, which consti-
only summarily, in the NOS lists developed for teaching and tutes a primary epistemological basis for teaching and
discussing the NOS (see Tala and Vesterinen 2015), although discussing the NOS.
modelling is now defined as one of the core skills that science In this article, we have connected the methodology, episte-
teaching must aim to develop in students (NGSS Lead States mology and educational use of computer simulations with the
2013). Computer simulations especially are almost entirely basic goals of understanding and carrying out scientific prac-
absent from the NOS lists, where more traditional aspects of tices in science education. The article is addressed to science
scientific methods are represented. Computer simulating is a educators, science and technology teachers, and designers of
relative new field, and so the nature and epistemology of com- educational software who are interested in using simulations
puter simulations have only recently begun to be studied (e.g. in support of those objectives in high school and teacher train-
in the philosophy of science). ing classes.
In this context, we (1) present some basic elements of the
methodology and epistemology of computer simulations with
the object of contributing to the formation of a coherent and Methodology of Computer Simulations
comprehensive account of the nature of computer simulations
in science education and (2) discuss certain implications of Numerical Solutions of Differential Equations—the
these elements for an understanding of contemporary scientif- Basic Idea and Structure
ic practices and for the use of simulations for doing science in
the classroom. We will begin by outlining the basic logic and structure un-
Concretely, in ‘Methodology of Computer Simulations’, derlying the numerical solution of differential equations,
we describe processes and particular techniques, skills, and deeming that an elementary description of these is important
innovations required for constructing and evaluating comput- for an understanding of the methodological and epistemolog-
er simulation models (algorithms and codes) and for tackling ical characteristics of computer simulations. (Since our focus
the problems peculiar to computer simulating. In is on a conceptual understanding of the process, we will limit
‘Epistemology of Computer Simulations’, we describe basic the technical part as far as possible.)
epistemological issues concerning the reliability and epistemic The process of computer simulating is largely based on the
status of simulations, through a brief reconstruction of the numerical solution of differential equations. The basic idea
related philosophical discussion. The text focuses mainly on behind numerical solution is the conversion of differential
the features of physics simulations, but most of it applies equations, which contain derivatives of the variables of the
equally to simulations in the other natural sciences, even if system (e.g. dy/dt), into difference equations, which contain
these have their distinctive features and epistemological is- quotients of differences (e.g. Δy/Δt) and can thus be solved
sues. In ‘Research on the Impact of Simulations on Science algebraically with the ordinary arithmetical operations of ad-
Teaching and Learning’, we summarize basic findings and dition and multiplication (discretization process).
proposals for improving the use of simulations in science We give, schematically, as an example of the numerical
teaching and learning through a concise overview of related solution of differential equations, the solution of the differen-
research papers. In ‘Nature, Possibilities, and Reliability of tial equation of the simple theoretical–mathematical model in
Simulations in the Educational Context—Implications of the classical mechanics for the motion of a body with constant
Methodology and Epistemology of Computer Simulations’ acceleration (in one dimension); it also has an analytical solu-
and ‘Use of Simulations for Promoting Contemporary tion, which permits comparison of the two. This equation is
Educational Goals—Implications of the Methodology and the fundamental equation of Newtonian mechanics F = ma
Epistemology of Simulations’, we discuss implications of with F = const (F is the acting force and m and a are the mass
the methodology and epistemology of computer simulations and the acceleration of the body). Since in classical mechanics
in relation to fostering an understanding of the nature of velocity v is defined by the derivative of position y with
 J Sci Educ Technol

respect to time (v = dy/dt) and acceleration by the derivative of reality is not true and introduces errors, which is why the nu-
velocity v with respect to time (a = dv/dt), the equation F = ma merical solutions are approximative compared to the analytical
in its differential form is F = ma = md2y/dt2 → d2y/dt2 = F/m, solutions. A question here is how large a number of difference
and finally d2y/dt2 = C (where F/m = const = C). equations there should be for the most accurate numerical so-
The analytical solution of the equation gives, for the vari- lution possible. The number of difference equations of the nu-
ables velocity (v) and position (y), the continuous functions v = merical solution is determined by the modellers depending on
v0 + at and y = 1/2at2 + v0t + y0, from which it is possible to the nature of the system being studied and the desired degree of
calculate their values for any time point, if the initial conditions, precision in a specific study of the system. In practice, the
i.e. the initial velocity and position v0 and y0, and the parameters number of difference equations, and therefore of corresponding
F and m of the system, are given. On the other hand, a numer- calculations, is immense (e.g. for the most realistic models or
ical solution of the equation, by using for example the simplest for models of spatially or temporally vast phenomena, e.g.
forward Euler discretization method,4 equates to an infinite atmosphere–climate or the evolution of planetary systems or
number of difference equations (the set of Eqs. (1) and (2)): organisms that have a large number of variables and parame-
ters). They are, however, in a form that can be translated into
Δv ¼ aΔt→viþ1 −vi ¼ aΔt→viþ1 ¼ vi þ aΔt ð1Þ
commands for the computer and processed by it.
v1 ¼ v0 þ aΔt ð1:1Þ Obviously, the greater the number of discrete values calcu-
v2 ¼ v1 þ aΔt ð1:2Þ lated, i.e. the smaller the calculation step Δt (a denser grid),
the more closely the numerical solution will approach the
v3 ¼ v2 þ aΔt analytical solution and the graph of discrete values given by
ð1:3Þ
::: and so on the numerical solution will tend to converge with the corre-
sponding graph of the analytical solution (Fig. 1). However,
Δy ¼ vΔt→yiþ1 −yi ¼ vΔt→yiþ1 ¼ yi þ vi Δt ð2Þ
the larger the number of difference equations, the more values
y1 ¼ y0 þ v0 Δt ð2:1Þ the computer has to calculate and therefore the more time and
y2 ¼ y1 þ v1 Δt ð2:2Þ computer memory is required, in which case, with the limited
computational power of computers, the equations of the
y3 ¼ y2 þ v2 Δt models may become computationally intractable (see
ð2:3Þ
::: and so on ‘Construction, Control, and Validation of Computer
Simulations and Peculiar Problems of Computer Simulating’).
where i = 1, 2, 3, …, n, Δt = ti + 1 − ti is the time interval The mathematical process for the numerical solution of
between two successive discrete time points ti and ti + 1, Δy = differential equations described above is a basic component
yi + 1 − yi is the movement of the body, and Δv = vi + 1 − vi the of computer simulations, but this does not mean that their role
change of its velocity over the time interval Δt. is confined to the solution of the differential equations of the
From these difference equations, it is possible to calculate, fundamental theories (see in this regard Hughes 1999;
step-by-step, any number of discrete—but only discrete— Lehnard 2007; Humphreys 2004; Winsberg 2010). As we will
values of the variables at the corresponding time points, but explain in the next section, computer simulating involves all
not in the intermediate interval, unlike the analytical solutions. the usual elements of constructing theoretical models (basic
Concretely, given the initial conditions v0 and y0 at the initial theories and experimental data, proper idealizations and
time point t0 (i = 0), the Eqs. (1.1) and (2.1) can be used to choices of theoretical approaches for the systems modelled)
calculate the values of the velocity and position v1 and y1 at but also requires additional methods and skills for turning the
the next time point t1, and then in the same way, the values v2 simulation models into tractable algorithms.
and y2 at the next time point t2 can be calculated from v1 and y1,
and so on. (These temporal points, and/or spatial points in other Construction, Control, and Validation of Computer
models, define the grid of the computer simulation model.) Simulations and Peculiar Problems of Computer
The numerical calculation of the discrete values of the var- Simulating
iables is based on the assumption that during the intermediate
interval Δt the variables v and y remain constant, which in The difference equations of the numerical solution, together
with the determined initial and boundary conditions and pa-
4 rameters set, constitute the core of the simulation algorithm,
The core logic of forward Euler discretization is in general outline also
followed, although with mathematically more advanced and complicated the computer simulation model per se (Winsberg 2010). The
discretization schemes, in the solution of the very complex differential equa- algorithm of a computer simulation, which at this point is still
tions (which contain derivatives of the variables also with respect to space, i.e. a concept on paper, is then translated into the program or code
they are partial differential equations) contained in the models of complex
states/phenomena, such as turbulent flows in the atmosphere or the astrophys- of the simulation, using a programming language (e.g.
ical plasma in solar flares or the progress of a forest fire. FORTRAN). The program can then be run on a computer
J Sci Educ Technol

Fig. 1 Comparison of the graphs of the numerical solution and the analytical solution for free fall with Modellus, (a) when Δt = 1 s and (b) when Δt =
0.2 s. Initial conditions and parameters used in the program: y0 = 0, v0 = 0, g = 9.81 m/s2, and m = 1 kg

and yield the results of the simulation. These results are ini- creation of multi-scale or hybrid models, which use both more
tially a huge set of numbers, which are then processed into and less analytical theories at the same time in the model (see,
more familiar and transparent forms of representation (graphs, e.g. Lehnard 2006; Winsberg 2010).
diagrams, films, etc.) so as to be more interpretable and com- All these elements—the complexity of the algorithms, the
parable with the experimental and theoretical data that exist errors resulting from the numerical schemes, the tractability
for the system being studied.5 restrictions, and the relatively new, less straightforward tech-
Simulators are faced with certain particular problems inher- niques used to address these problems—complicate the con-
ent in computer simulation construction methods and the use trol of the simulation and thus create concerns regarding the
of computers. For instance, instabilities may appear during the reliability of its results (Winsberg 2010). In order to evaluate
running of a simulation: that is, after a while, the values of the the reliability of computer simulation models, apart from car-
variables become infinite, rendering the program useless. The rying out the basic control of the validity of the model, typi-
instabilities may be due in the main either to the numerical cally by comparing its results with the empirical data, scien-
approximations or to an erroneous or overly idealized theoret- tists also perform additional mathematical checks. Thus, com-
ical approach at the modelling stage. These problems may be puter simulation control processes are of two kinds, which
addressed by choosing a finer grid (in the first case) or a more scientists have labelled verification and validation:
analytical theoretical approach (in the second), but this could verification, for checking whether the numerical solutions of
increase immensely the number of calculations and lead to the model’s equations are mathematically correct, which aims
intractable algorithms and codes. Among the strategies that at identifying and removing errors in the numerical methods
have been developed to deal with this problem are that could be responsible for altering the image of a physically
parameterization (introducing sub-routines into the algorithm correct simulation, and validation, for checking whether the
because these are computationally less costly than a more simulation model is a satisfactory conception of the real sys-
analytical but intractable theoretical approach (see, e.g. in tem being studied (whether it fits the empirical data), that is,
Winsberg 2010; Gramelsberger 2010; Tala 2011)) and the whether the model’s theoretical equations were properly se-
lected and adapted to the real system.
5
Scientific computer simulations are essentially written in the mathematical A number of strategies and criteria have been developed for
form described in the previous subsection ‘Numerical Solutions of Differential the verification of the simulations. Thus, if the algorithm of a
Equations—the Basic Idea and Structure’, which is very productive because
simulation is correct, it should, when applied to any system/
the model can thus easily be improved or expanded with further processing.
The mathematical form of programming (as in Figs. 1, 2, and 3) is more state described by equations that have an analytical solution
appropriate for senior high school and college level students. For younger (e.g. by the harmonic oscillator), reproduce the results of the
students, graphically oriented computer-based programming software has
analytical solutions. The results of the algorithm can also be
been developed, which provides a microworld environment and a program-
ming language that is either in text form (the program is then a set of textual compared to those of another known code constructed for the
orders) or in graphical form (in which case the program is a set of behaviour same problem and based on the same initial theoretical equa-
rules for the objects/items designated to represent the phenomenon or system tions describing the system and the same initial and boundary
modeled. For a comparative and research-based study of these programming
environments (characteristics, ways of use, possibilities and limitations), see, conditions. If the results of these or any number of similar
e.g. in Louca 2004; Luca & Zacharia 2008; Sherrin et al. 1993. codes match, this is a strong indicator of the reliability of the
 J Sci Educ Technol

code in question. As regards the validation of the simulation, simulations, namely the view that running a simulation is
the most reliable and typical control process is the comparison equivalent to carrying out an experiment and by extension that
of its results (its predictions) with the empirical data for the a good simulation has the same epistemic value as a good
real system being simulated, which a good simulation ought to laboratory experiment, that simulations are experiments
reproduce if the whole mathematical process of the numerical (hence the emergence of the terms computer experiments,
solution of its equations is correct. numerical experiments, dry experiments, and in silico exper-
iments that scientists sometimes use for computer simula-
tions), and the view that simulations are not equivalent to
Epistemology of Computer Simulations traditional experiments, that there are differences between
them that make their epistemological value different.
For the reasons outlined above, the reliability of the results The view that computer simulations are equivalent to
yielded by simulations raises particular considerations, both in experiments, or that at least there is no essential difference
science, as we have seen in the previous section, and in the between them, is based mainly on the multiple methodo-
philosophy of science. logical analogies between computer simulating and tradi-
In the philosophy of science, it is no accident that the de- tional experimenting, as, e.g. the possibility for repeats
bate on the epistemology of computer simulations hinges on and the similar techniques of data analysis (see, e.g.
their relation to theory and experiment, given the established Humphreys 2004; Hughes 1999; Norton and Suppe
position of the basic theories and experiments as sources of 2001; Parker 2009; Morrison 2009; Giere 2009;
reliable knowledge. The debate focuses mainly on the com- Winsberg 2010). Indeed, with simulations, as with labo-
parison of simulation to experiment, on the one hand because ratory experiments, it is possible to run many repeats with
the practice of computer simulating is to a considerable extent different parameter values or with different boundary and
analogous to traditional experimentation, and on the other initial conditions, with altered numerical schemes, etc.,
because experiments have an ancient tradition in the develop- and there is also intense analysis of the output data from
ment of methods for ensuring the reliability of their results. the simulation, with techniques similar to those used in
As a framework for the comparison of simulation and ex- experiments (handling errors and uncertainties, eliminat-
periment, we note some basic elements of the process of ex- ing artefacts, etc.). This view also points out that, like
perimentation. Very briefly, one might say that an experiment traditional experiments where the target system is repre-
studies properties, structures and behaviours of a specific sys- sented by another system, in computer experiments the
tem (or a series of like systems) that constitutes the object of target system is represented by the simulation program/
interest, or the target of investigation. The experimental inter- code created for it (e.g. systems such as the galaxies, solar
vention can be applied to the target system itself, or to another flares, or nuclear radiation are represented by and inves-
system, the object of investigation, which is considered a sat- tigated via simulation programs for these systems) (e.g.
isfactory representation of the target system, e.g. experiments Norton and Suppe 2001; Winsberg 2010). In sum, it
using a mechanically vibrated system to draw conclusions might be said that the key point in the framework of this
about electrical oscillation (Norton and Suppe 2001; Hughes view is that the traditional experiment is not intrinsically
1999), or using pea plants to draw conclusions about the he- and a priori epistemologically more valid than a computer
redity of characteristics in all organisms (Giere 2006), or using simulation (e.g. Parker 2009) and that the obstacle to
a gas to draw conclusions about black holes or plasma in eliminating the distinction between simulation and tradi-
astrophysics (Winsberg 2010). Other typical characteristics tional experiment may be an intuitive psychological pre-
of an experiment are the possibility of repeating the experi- disposition, based on the ancient and fundamental role of
mental process and the extensive processing of the experimen- the experiment in the history of scientific methodology
tal data, which involves a series of assumptions and theories (Winsberg 2010).
for the statistical analysis of the experimental results, the han- The key argument in the other view, which distin-
dling of errors of measurement, the elimination of external guishes simulations from traditional experiments, is that
factors/noise, etc. traditional experiments can work not only with the tar-
The relation of computer simulations to theory and exper- get system but also with another representative system,
iment, and especially their comparison with experiment, has which, however, is also a material system that in its
served as the basis for various views and arguments properties and structure has obvious material similarities
concerning their nature and epistemological status. Among with the target of investigation, while in computer ex-
these are the view that computer simulating is a new scientific periments, the manipulations and calculations are carried
practice, both theoretical and experimental (e.g. Dowling out by a simulation program that bears no material re-
1999; Galison 1997; Tala 2011), and the two opposing views, semblance to the target system, so that in the final anal-
which are more directly related to the issue of the reliability of ysis, the experimentation is with a model of the
J Sci Educ Technol

phenomenon rather than with the phenomenon itself the simulations take the place of experiments and
(Guala 2002, Morgan 2003). Consequently, one cannot observations—and so they are trusted even in circumstances
draw conclusions for the system being studied directly where they cannot be evaluated by comparing their results
from the results of the simulation, or according to with the world’ (p. 120).
Μorgan (2003), the experiment is epistemically more Winsberg also thinks that the essential difference be-
powerful precisely because of the interaction with the tween simulations and traditional experiments, from the
system of interest itself (material similarity argument).6 epistemological point of view, is the different argumenta-
Supporters of the first opposing view, however, challenge tion used by scientists to affirm the reliability of their
the obviousness and importance of the material similarity be- results in the two cases. Arguments for the reliability of
tween the object and the target of investigation, noting that in experimental results may be based on the materiality of
practice, the determinant reliability criterion, both for experi- the interactive process with the physical systems, while
ments and simulations, is not the similarity, respectively ma- those for the reliability of simulation results may be based
terial or formal, between the target and the object of investi- (apart from the fact that they can make predictions that
gation but rather other factors, such as which and how many agree with actual data) on the fundamental principles un-
factors were taken into account in the model or experiment, derlying the simulations and the various reliable tech-
and the reliability of the experimental methods and devices niques applied, e.g. the discretization schemes, the param-
used in the experiment or the methods used in the simulation eterizations, and the mathematical checks (see also Giere
(the validity of the underlying theory, the quality of the nu- 2009). Winsberg does not agree, however, that experi-
merical schemes, the fineness of the grid, etc.) (Parker 2009). ments are inherently more epistemically powerful than
In relation, moreover, to the argued epistemological superior- simulations, arguing that this claim is affected by the
ity of the laboratory experiment, Parker (2009) also notes long-standing tradition and status of laboratory experi-
instances where a good simulation, based on solid theories mentation in scientific methodology, compared to the re-
and numerical techniques, yields much more reliable results cent methodologies and techniques of computer simula-
and calculations, and permits inferences for the system under tion modelling, although these are evolving and have al-
consideration, than any corresponding experiment (such as ready developed their own tradition, thus increasing the
simulations based on Newton’s gravitational theory for the epistemological power of simulations. ‘There may have
calculation of planetary orbits and other related phenomena). been a time in the history of science, perhaps before
Within this framework of comparison between simulation Newton, perhaps even earlier, when we did not have suf-
and traditional experiment, attention is also drawn to certain ficient systematic knowledge of nature—enough of a
epistemological differences between them: specifically the in- toolkit of trustworthy model-building principles—to cre-
ability of simulations to answer questions of an ontological ate a simulation that could ever be as reliable a source of
nature, e.g. about the existence of some particles in the frame- knowledge as even the crudest experiment, but that time
work of a theory, and that while an experiment can sometimes has long passed’ (Winsberg 2010, p. 71) (see also
play a crucial role in the validation/selection of models and Gramelsberger 2010).
hypotheses (empirical confirmation of predictions), a simula-
tion alone cannot do the same thing, except where it is possi-
ble to compare the simulation results with actual data (Hughes Context-Depended Practices and Epistemologies
1999; Giere 2009). Giere (2009) states that ‘Computer exper-
iments not connected to actual data, do not confer any addi- The cases typically studied and used to extract or exemplify
tional confidence in the fit (or lack thereof) between the sim- the above views in the philosophy of science are simulations
ulation model and the target system’ (p. 62). In relation to this from the theories of fluid mechanics for the study of astro-
point, Winsberg (2010) notes additionally, however, that physical or meteorological/climatological phenomena (e.g.
‘Indeed researchers run simulations of systems about which Lehnard 2007; Gramelsberger 2010; Winsberg 2010), from
data from real experiments is difficult or impossible to get— the nanosciences/nanotechnology (e.g. Lehnard 2006;
Winsberg 2010), from psychology-neurophysiology (e.g.
6
It has though been argued (e.g. Morrison 2009) that, like traditional exper- Suppes 1997), biology (e.g. Knuuttila and Loettgers 2013),
iments on material systems, computer simulations also have materiality, in that and economics (e.g. Morgan 1998). These views, in other
during the running of a simulation there is experimentation with a material
words, developed in the context of a contemporary philosophy
entity, that is, with a programmed computer (the object of investigation in this
case) that undergoes the intervention of the simulation program and falls into of science that based its positions on scientific practice
different states during the running of the simulation, yielding information (naturalized philosophy, see, e.g. Giere 1999), a trend that
about the evolution of the target system. (For more details as regards the has developed in response to the criticisms of the history and
explanations, the objections and the arguments relating to this view, see, e.g.
Morrison 2009; Norton & Suppe 2001; Hughes 1999; Giere 2009; Winsberg sociology of science that the traditional philosophy of science
2010). is cut off from and does not reflect real scientific practice.
 J Sci Educ Technol

Research in the history and sociology of science empha- literature (see ‘Use of Simulations for Supporting NOS
sizes, for example, the different cultural traditions and Understanding’).
mindsets in the various disciplines (biology, chemistry, phys-
ics, geology), and the resulting varying epistemological views
(as regards, e.g. data collection or the evaluation of evidence) Research on the Impact of Simulations
among scientists of those different disciplines, as well as on Science Teaching and Learning
among the sub-communities within each discipline, e.g. ex-
perimentalists, theorists-modellers, engineers and mathemati- Education policy in many countries has clearly seen the class-
cians, novices, and experts, that co-exist especially in the com- room possibilities of the new technologies and the need for
plex, computer simulation-based research of contemporary students to become familiar with their use. Information and
science (e.g. in biochemistry, biophysics, and environmental communication technology (ICT) has given rise to a variety of
sciences) (see, e.g. Galison 1997; Tala 2013). Galison (1997), new teaching tools, such as recourse to the Internet, e-books,
for example, describes the interactions, co-operation and ne- video games, multimedia of virtual worlds and environments,
gotiations that take place in the ‘trading zone’, e.g. for the and computer simulations. Among them, computer simula-
evaluation of the fit of simulation models and for the general tions stand out as the most interactive and participative digital
processing of such interdisciplinary projects, while Tala teaching tools and the most appropriate means for supporting
(2013) explains that this trading process contributes to the model-based inquiry learning (see in Scalise et al. 2011;
enculturation of novices in a specific field and argues in fa- Smetana and Bell 2012). Simulation-based software provides,
vour of supporting the enculturation process with explicit among other things, tables of values and graphs of the vari-
epistemological discussions and reflections. ables, visualizations of mechanisms and processes, and inter-
The subject of discipline-specific aspects of science has active virtual laboratories.
also been addressed in science education. Several studies The impact of simulations on science teaching and learning
(e.g. Wong and Hodson 2009; Irzik and Nola 2011; has in recent decades been the subject of many research pa-
Matthews 2012) have criticized the lists of NOS items pers, the majority of which report positive results for the im-
created for teaching the nature of science (see, e.g. in pact of simulations, while noting their limitations and neces-
Abd-El-Khalick et al. 1998; Lederman 2006) as present- sary preconditions and the improvements still needed. This
ing an unchanging and far too generalized picture of sci- research suggests and provides evidence of many possible
ence, ignoring the discipline-specific and the context de- benefits of computer simulations, among them that they can
pendency of scientific work. Their recommendations in- be incorporated into different kinds of teaching and can foster
clude taking scientists’ views into account in the concep- students’ understanding of scientific content, their inquiry
tualization and teaching of the nature of science and en- skills (see, e.g. in Smetana and Bell 2012; de Johng and van
couraging contacts between schools and scientists/science Joolingen 1998) and metamodelling knowledge (e.g. Schwarz
centres in order to provide students with an authentic and and White 2005; Gobert et al. 2011; Webb 2005), and
updated picture of scientific enterprise. teachers’ education and professional development (e.g.
Students also develop their own epistemologies, e.g. Zacharia 2005); that variables can easily be changed or added,
through engagement in laboratory or computer-based experi- making teaching and learning more applicable to real-world
ments, (and also through the influence of the media). How far problems and, by giving direct feedback for these changes,
students’ epistemologies are more or less adequate depends enable the learning process to be individualized and personal-
among things on the epistemologies underlying such activi- ized (e.g. Zacharia 2005; Linn 2003); that they make it possi-
ties: when, for example, in teaching (or in the media), the ble to recreate real-world perspectives which would otherwise
predominant view is that in science, the checking of hypoth- be too complex, costly, time-consuming, or risky to handle in
eses and models is based solely on the conducting of experi- a conventional classroom setting (e.g. Lunetta and Hofstein
ments, or that the result of a verification experiment can, in 1981; de Jong et al. 2013); that by simplifying the complex
every case, decisively determine the validity or otherwise of real-world environments, they make it possible to reduce ex-
the scientific hypothesis, this implies an oversimplified and traneous cognitive load, allowing the students to focus their
empiricist view that the methodology of empirical sciences attention on studying specific target topics and relations
is governed by indisputable empirical proofs and unambigu- (Smetana and Bell 2012; Zacharia 2005); that they make it
ous decisions. This is far from the reality of scientific practice, possible to create more comprehensive and interdisciplinary
with its complex validation processes (see ‘Use of Simulations programs for working on specific science topics, e.g. projects,
for Supporting Students’ Reasoning and Evaluating and for addressing socio-scientific issues (e.g. Lunetta and
Abilities’), and various ways to address such misconceptions Hofstein 1981; Linn 2003; Webb 2005; Barab and Dede
and promote effective teaching and learning about the nature 2007); that they provide an opportunity for involving students
of science have been proposed and studied in the related in the processes of scientific inquiry and reasoning and for
J Sci Educ Technol

developing students’ mathematical modelling skills at high research projects, expert-like modelling and reasoning activities
school and university level (e.g. Wu 2010; Lunetta and or distance and non-formal learning, and for addressing socio-
Hofstein 1981; Andaloro et al. 1991); and that they can enrich scientific issues) (Lunetta and Hofstein 1981; Wu 2010; Webb
the social dimension of learning (discussion, collaborative 2005; Linn 2003; Barab and Dede 2007; Anneta 2012).
cognition and activities) (Roth et al. 1996; Hmelo and Day The implications of the methodology and epistemology of
1999). computer simulations discussed below focus mainly on this
There are also reviews of sets of such research papers that last area of research.
give a more general overview of the situation (e.g. Smetana
and Bell 2012; Rutten et al. 2012), provide instructional de-
sign principles and criteria for the design and evaluation of Nature, Possibilities, and Reliability
virtual labs and simulations (e.g. Scalise et al. 2011), or ana- of Simulations in the Educational
lyse how the learning affordances of ICT-based environments Context—Implications of the Methodology
can be enhanced, especially when they are used in innovative and Epistemology of Computer Simulations
didactical approaches such as conceptual change, formative
assessment, or NOS-oriented curricula (Webb 2005). For ex- Educational Simulations Compared to Simulations
ample, Smetana and Bell’s 2012 review of empirical research in Scientific Research
papers shows that as regards the impact of simulations on the
promotion of conceptual understanding (conceptual change) Computer simulating, as already noted, combines all the usual
and process skills, simulations can be as or more effective than elements and strategies of modelling (involvement of theories
traditional instructional practices and tools (lectures, text- and empirical data, idealizing assumptions, etc.), but also re-
books, laboratory experiments) and that the factors that in- quires additional particular techniques and skills for the fur-
crease their effectiveness include teacher assistance for fuller ther development of the simulation algorithms and codes. It is
exploitation of their possibilities (e.g. multiple representations important to note here that there are certain differences be-
of the phenomena, immediate access to data and other infor- tween educational simulations and those used in scientific
mation), specific design features (rapid easy repeats of proce- research. In science, the modellers are free to modify and
dures, application of knowledge to new situations, etc.), and manipulate the simulations that they are constructing as they
combination/integration of simulations with other traditional choose; they themselves, in other words, decide on the theo-
teaching practices and learning experiences. This last point, on retical approach and degrees of freedom of their model (its
the effectiveness of combining simulation with traditional equations, variables, and parameters), the kind of numerical
teaching practices and tools, is also a focal research question solution of the equations, and the modifying interventions
in Rutten et al. (2012), where the papers reviewed provide applied for purposes of adjustment, improvement, and tracta-
evidence that computer simulations can enhance traditional bility. This is not, however, the case with the simple use of
instruction (they improve the cognitive and affective aspects prepared computer simulations, as, e.g. with the use of an
of the learning process) when they are used, e.g. either to ordinary educational simulation, where those using it may be
supplement or replace parts of the curriculum (see, e.g. able to alter the values of the parameters and initial conditions
Zacharia 2007; Zacharia et al. 2008). This research result is but not its underlying equations (the theoretical model) which
important because it suggests an effective and realistic way to have been defined by its constructor.
generalize the use of computer simulations in teaching, which Moreover, simulations in scientific research seek solutions
does not require specific or major changes to current curricula to as yet unanswered questions about the studied phenomena,
and current teaching methods and means. whereas educational simulations are usually ready-to-use pro-
The research also looks into how far computer simulations grams on topics for which a solid body of scientific knowl-
can meet the demands of innovative teaching approaches and edge already exists. This knowledge may be reflected to a
help with the handling of the associated problems, e.g. in the greater or a lesser degree in the educational software at the
case of inquiry learning by providing students with scaffolds discretion of the creator, who decides, in other words, how
and cognitive tools (pre-defined hypotheses, background in- realistic the underlying theoretical model of the simulation
formation, hints for efficient experimentation, etc.) (de Jong should be and whether it will be based on, e.g. a deterministic
and van Joolingen 1998; de Jong 2006). or a more stochastic approach: ‘The author of the simulation
While most existing research focuses on the impact of simu- decides which elements to include and which to omit on the
lations on classical science education goals, science content basis of instructional goals and the nature of the system being
learning, and the development of process skills, there are also simulated’ (Lunetta and Hofstein 1981, p. 243; see also Dorn
an increasing number of studies that examine and recommend 1975). This means that a degree of suggestive direction and
the use of simulations for promoting newer aims and orientations control is introduced into the simulation, so that the user can
(e.g. for promoting an understanding of the nature of science, only choose between predetermined variables and parameters
 J Sci Educ Technol

and simply vary their values so as to reach the desired conclu- activities, or even by scientists for topics where the research
sions. (Essentially the users re-discover the properties of the is still ongoing. When simulations are constructed in class-
model underlying the (educational) simulation; see de Jong room activities, validation of the simulation models has to
and van Joolingen 1998; de Jong 2006.) According to Dorn be one of their basic objectives, and the best and standard
(1975), although there is also a specific amount of direction in strategy for checking computer simulation models and their
the conventional laboratory, the possibilities and challenges underlying hypotheses is comparing their results with empir-
are much greater with computer simulations. Where the stu- ical data, e.g. with the results of laboratory experiments. The
dents themselves are constructing a computer simulation mod- absence of such discussion and clarification concerning the
el, however, or essentially reconstructing a given simulation, epistemological features of simulations may lead to miscon-
e.g. with an advanced modelling software, their activity is ceptions regarding their reliability (see Greca et al. 2014), for
closer to scientific simulating, although they are obviously example, to the idea that simulations are simply games played
working with simpler models and as a rule re-discovering with the values of the variables or initial conditions until the
known knowledge and solutions. simulation matches the measured data or to the opposite idea
that their reliability can be taken for granted.
The Reliability of Simulations
Computer Simulations Compared to Experiments
Another important point concerns the perceptions that stu-
dents may have of the reliability of simulations when they In science education, the comparison between computer sim-
are simply used with no explicit discussion of this aspect. As ulations and experiments mainly concerned comparing the
we noted in ‘Epistemology of Computer Simulations’, the teaching and learning benefits of laboratory experiments with
philosophy of science view that simulations can have the same those of virtual experiments of simulation-based software (see
epistemic value as experiments is qualified: it applies to sim- in this regard, e.g. Lunetta and Hofstein 1981; Zacharia 2007;
ulations based on well-established theories and tested comput- Zacharia et al. 2008; de Jong et al. 2013).
er simulation methods and techniques. It is also recognized In the context of the aims of this paper, we will focus here
that simulations on their own, that is, without the possibility of on comparing simulations and laboratory experiments with
evaluation by comparison of their results with the real world, regard to how far computer simulations promote students’
cannot have the same weight as experiments for the validation inquiry and reasoning abilities compared to laboratory
and selection of claims and hypotheses. Giere (2009), for ex- experimentation.
ample, notes that with the view that computer experiments are Two of the basic activities that are traditionally promoted
in general equivalent to traditional experiments there is a risk, by classroom laboratory work are inductive and deductive
especially with non-experts, of overestimating the reliability processes and reasoning (see, e.g. Popper 1959). In the first
of the simulations, a risk that ‘…can mislead consumers of case (inductive reasoning), students are encouraged to draw,
simulation results (and maybe even some practitioners) into from particular observations and measurements, general con-
thinking that their simulation models are epistemologically clusions about the phenomenon they are studying: e.g. they
better founded than they in fact are.’ (p. 62). This mispercep- infer Hooke’s law or the laws of gases from some relevant
tion can have serious consequences, for example, in the case observations and measurements with some springs or gases.
of assessing, and therefore adopting or rejecting, simulations In the second case (deductive reasoning), they check the va-
that are associated with important matters in private life and in lidity of general scientific principles or models, or of their own
the social-political domain (health, socio-scientific issues, hypotheses, by checking how far the predictions that are de-
etc.) duced from these principles, models, and hypotheses agree
The implication is that the question of the reliability of with the experimental and theoretical data: for example, in
simulations in teaching has to be handled in a balanced and order to examine the validity of Newton’s gravitational theory
critical manner. Although educational simulations constructed which predicts the association of the oscillation period with
for the study of common curricular contents can be presented gravitational acceleration, they calculate the value of gravita-
in the classroom as reliable, since they are based on tional acceleration based on measurements of the period of a
established knowledge, students must not be left with false pendulum, and compare it to the known value.
impressions as regards the question of the reliability of simu- The philosophical discussion on the epistemology of com-
lations in general. It must be made clear to them that we can puter simulations focuses especially, as we saw in
have confidence in the results of ready-to-use educational ‘Epistemology of Computer Simulations’, on their compari-
simulations because these are based on well-established theo- son with experiments. Many of the arguments advanced in this
ries that are customarily taught, such as classical mechanics or debate tend towards the view that the running of computer
electrodynamics, but not a general confidence in simulations, simulations is equivalent to carrying out an experiment, and
e.g. when these are constructed by students in classroom that the traditional experiment is not, a priori and in every
J Sci Educ Technol

case, epistemologically superior to simulation. In a number of need in order to cope with the demands and the socio-
science education articles on the subject, computer simula- scientific issues of the age (see, e.g. Hodson 2014; Clough
tions are also understood and treated as computer experi- and Olson 2008). Science educational research has produced
ments, as media that promote analogous reasoning processes a significant body of material and research on the content of
and goals to laboratory experiments (see, e.g. Dorn 1975; NOS teaching, i.e. which ideas correctly reflect basic aspects
Lunetta and Hofstein 1981; Andaloro et al. 1991). Thus, as of science and should be taught (e.g. Osborne et al. 2003;
with experiments, so with simulations, one can in general Lederman 2006; McComas 2008; Irzik and Nola 2011;
discern two main ways of using them in classroom activities: Wong and Hodson 2009), as well as on how to teach it. As
students either run a ready-to-use simulation program in regards the ways for the teaching and learning of the nature of
which they change only some input data, and from the results science, two basic approaches have been developed’: the im-
given for particular cases are led to general conclusions or plicit approach, which argues that students acquire an under-
principles for the whole class of similar systems; or they con- standing of NOS from their participation in the usual teaching
struct their own simulation program (or reconstruct a prepared and classroom activities, and the explicit approach, which
program) and examine its validity by comparing the results to holds that this is insufficient, and that the learning efficacy
empirical data. With computer modelling, they can also mod- of NOS instruction is increased by explicitly addressing par-
ify (by changing hypotheses, parameters and initial condi- ticular NOS ideas contextualized in scientific content (see, e.g.
tions) and run the program quickly, getting immediate feed- Khishfe and Abd-El-Khalick 2002; Clough and Olson 2008;
back for the modified models. Computer simulations thus al- McComas 2008; Develaki 2012; Gilbert and Justi 2016).
low students to become involved in reasoning and inquiry Modelling-based activities can be an effective means for
processes (see ‘Use of Simulations for Supporting Students’ both implicit and explicit teaching and learning aspects of
Reasoning and Evaluating Abilities’) similar to those of labo- NOS concerning, e.g. the practices of scientific experimenta-
ratory experiments, which is important especially when deal- tion and theorizing; this applies especially to computer
ing with microscopic and macroscopic phenomena, which are simulation-based activities, which are particularly suitable
difficult to study in laboratory conditions and which scientists for carrying out and exemplifying scientific practices in the
use simulations to study: ‘…simulations are often constructed classroom.
precisely because data about the systems they are designed to Below, we give an example of the use of simulations for
study are sparse. In these circumstances, simulations are explicit teaching and understanding of NOS ideas contextual-
meant to replace experiments and observations as sources of ized in curricular science content. The example is based on
data about the world’ (Winsberg 2010, 121). simple simulation models for the fall of bodies from a great
height.
When friction is ignored in the model (the free fall model),
Use of Simulations for Promoting running the corresponding program shows that two bodies of
Contemporary Educational different mass have the same velocity (v) at each moment and
Goals—Implications of the Methodology therefore reach the ground at the same time (Fig. 2(a)); the
and Epistemology of Simulations graphs of the velocity of the two bodies coincide). This model
renders the phenomenon satisfactorily only in specific condi-
We have already mentioned that one important research find- tions which in practice make the friction of the air negligible in
ing concerns the enhanced learning affordances of ICT-based comparison with the weight of the falling bodies. (In the free
environments, especially when these are used in innovative fall model: F = W = mg = ma, where F is the acting force, m, W
didactical approaches and for innovative educational goals. and a are the mass, weight and acceleration of the falling body,
In this section, we shall discuss, with examples, the possibil- and g is the gravitational constant.)
ities offered by simulations for supporting students’ NOS un- In the model that takes the friction of the air Ffr into account
derstanding and inquiry and reasoning abilities. The examples (F = W + Ffr, and let Ffr = −bv), the bodies achieve a final
are based on simple simulations of high school mechanics velocity that is greater in the case of the heavier body because
topics using Modellus. attained later (Fig. 2(b)). The heavier body thus falls to the
ground sooner than the lighter. Thus, Newtonian theory in-
Use of Simulations for Supporting NOS cludes Aristotle’s conception of the fall of bodies as one of
Understanding its models.
The possibility of constructing and immediately comparing
Teaching the nature of science (the nature of scientific alternative models and hypotheses in the above manner can be
methods and knowledge and the socio-cultural dimension of very useful for supporting students’ conceptual change and for
science) contributes to a meaningful science learning, and to highlighting and discussing elements of the nature of science.
the acquirement of skills and epistemologies that students will The specific example demonstrates the unifying potential of
 J Sci Educ Technol

Fig. 2 Simulations with Modellus for the fall of two bodies of different conditions and parameters used in the program are as follows: m1 = 1 kg,
weights with the corresponding v(t) graphs: (a) without air friction (the m2 = 3 kg, g = 9.81 m/s2, b = 1.50, y0 = 0, v0 = 0 (Δt = 0.1 s)
graphs coincide) and (b) with air friction (Ffr = −bv). The values of initial

the fundamental theories, as here Newtonian theory incorpo- NOS aspects, and that this requires explicit discussion about
rates the Aristotelian view. It also helps demonstrate a basic these NOS aspects alongside and in relation to the modelling
characteristic of models, namely that there can be more than activities carried out, suggesting thus a combination of the
one model for the same system or phenomenon: models that implicit and explicit approach can best enhance the effective-
are more or less realistic (e.g. models for motion with or with- ness of NOS instruction.
out friction), which are selected and used according to how
well the scientist thinks they cover the points of interest and
precision requirements of a specific study. Use of Simulations for Supporting Students’
Furthermore, the very fact of participating in modelling Reasoning and Evaluating Abilities
activities gives students a chance to experience and engage
with aspects of the nature of science. Modelling activities We have argued that simulations can promote reasoning ac-
familiarize them with the practice and process of scientific tivities on a par with experiments. Now, we shall look at how
modelling, i.e. with the building, presenting, testing, revising, computer simulations can support activities that promote ad-
evaluating, and applying of models. In addition, taking part in vanced reasoning and specifically deductive reasoning as
the various stages of modelling involves them in other basic interpreted in the model-based view.
scientific practices that are necessary for effectuating those In the model-based view, deductive reasoning is
stages, such as reasoning and evaluating (see ‘Use of interpreted not formally, as in Logical Empiricism,7 but
Simulations for Supporting Students’ Reasoning and more pragmatically, by taking into account the practical
Evaluating Abilities’). For example, participating in the crea- and conventional criteria that scientists (in addition to
tion and expression/presentation of a model gives them an empirical and logical tests) usually apply when evaluating
opportunity to deal with typical NOS features like the selec- the validity of their models (e.g. the simplicity or the
tion and organization of data, the formulation of hypotheses, interpretive and predictive power of models, as well as
and the use of arguments and of different ways of presenting socio-psychological criteria, including social traditions
and communicating ideas and conceptions, while engaging in and relations in their scientific community and profession-
the testing and evaluation of models helps them learn such al goals/interests) (Lakatos 1974; Kuhn 1989; Giere
NOS features as the weight of (empirical) evidence, argumen- 1999). This means, first of all, that the evidence yielded
tative discussions, and conventions and practical criteria in the by comparison of the model’s predictions with the data
testing and evaluation of scientific models and discover pro- does not necessarily lead to unequivocal conclusions
fessional and social parameters of science, such as collabora-
7
tion and competition in scientific research and the significance Logical Empiricism interprets deductive reasoning in a formal way, consid-
ering that the agreement or non-agreement of the predictions derived from
of material tools and funding (see Gilbert and Justi 2016). hypotheses and theories with the empirical data leads to the unequivocal ac-
Based on the results of their empirical research on this topic, ceptance or rejection of the theory. This is a strong scheme in logic and
Gilbert and Justi (2016) note however that students’ experi- mathematics but is insufficient to interpret in all cases the complexity of judg-
ing and choosing the empirical theories of science (see, e.g. Duhem 1978;
ence does not of itself lead them to awareness of the intended Lakatos 1974; Kuhn 1989; Giere 2001).
J Sci Educ Technol

about the validity of the model (see Giere 2001): is to involve students in more advanced, more scientific, ways
Scientists are critical and cautious whether the evidence of reasoning, evaluation, and argumentation, so as to help
for the model is negative (non-agreement between the them acquire a more critical stance and a solid basis for
predictions and empirical and theoretical data), since this assessing the credibility of science-related information.
can be due not to the model itself but to various other As a final observation, we want to point out that the sub-
factors, (e.g. error in the derivation or experimental veri- jects treated in this article, namely the methodology, episte-
fication of the predictions), or positive, because positive mology, and pedagogical use of simulations, are important
evidence does not mean that the model under examination elements in the design and implementation of activities that
is the best one, unless it has already been compared with help students master scientific practices and ways of thinking
other possible alternatives. Positive evidence for a and thus develop epistemologies and skills that they will need,
model’s prediction is of critical importance for the validity as future citizens and/or future scientists, in addressing
or selection of the model only if the possible alternative science-related problems of their age.
models do not yield the same prediction as the model
under examination; when the alternative models also give
the same prediction, the evidence is not conclusive, and Epistemological Views
scientists base the evaluation or choice of models on prag- Concerning the Relation of Scientific
matic criteria, and/or seek to acquire more conclusive data Knowledge to Reality—Implications
and evidence. from the Epistemology of Models
Computer simulations can in particular support classroom
activities that can engage students in advanced, scientific ways In the philosophy of science, the main question concerning the
of reasoning and argumentation, since they permit the quick relation between scientific claims and the real world is how far
running of, and immediate feedback for, numerous modified scientific theories, and especially their theoretical, non-
or alternative simulation models, and thus enable the compar- observable entities and properties (such as photons, quarks,
ative evaluation of these models. In this context we shall refer spin), are related to something real or whether their function is
briefly to an example of the use of simulations for model- exhausted in the internal structures and aims of the theories.
based deductive reasoning which is presented analytically in The two main opposing views are those of scientific real-
another paper (Develaki 2017). The case concerns the evalu- ism and those of empiricism and social constructivism. The
ation of alternative models (for the fall of parachutes) for strong version of scientific realism holds the view that we can
which there is positive or negative or inconclusive evidence. perceive the world as it is and thus supports the existence of
In these models air friction is taken into account, because the theoretical entities and argues that axioms are universal laws
free fall model that ignores friction does not predict the termi- of the world. The moderate version recognizes instances of
nal asymptotic velocity that the parachutes acquire during theoretical entities to which an instrumentalistic character only
their fall and with which they reach the ground but merely a can be attributed, that is, it accepts the postulation of theoret-
continuously accelerated velocity. Thus, the different hypoth- ical terms and entities for purposes of developing and fleshing
eses for the equation for the value of the air friction lead to out the theory, which are later abandoned because they have
alternative models and corresponding simulation programs no realistic counterpart in nature (such as the ether or phlogis-
that can be run to check whether they predict the terminal ton), but holds that scientific theories and laws can, at least
velocity and with the correct value. Running the program sometimes, reveal truths about the natural world (see, e.g. in
shows that the model in which friction is assumed to be con- Bunge 1970; Popper 1959; Boyd 1983; Devitt 1991). The
stant during fall does not predict the final velocity (negative empiricists, on the other hand, perceive theories (especially
evidence) (Fig. 3(a)). The model that assumes that friction is those concerning the microscopic world) essentially as useful
proportional to the velocity of the parachute does predict a research instruments, e.g. for providing successful predictions
final asymptotic velocity (positive evidence) and with an ac- and technically applicable knowledge (see, e.g. in Burian
ceptable value when the choice of constant of proportion C 1980; van Fraasen 1980), while the constructivists argue that
between friction and velocity is appropriate (Fig. 3(b)): the the impact of internal and external social influences on the
diagram shows the graphs for the velocities of the falling methodology of science (methodological traditions, conven-
parachute for 4 values of C). But the model that assumes that tions, hierarchies, etc., and external social and political fac-
friction is proportional to the square of the velocity of the tors) is sufficiently powerful to affect the evaluation of scien-
parachute also predicts an acceptable value for the final veloc- tific knowledge in terms of truth and objectivity.
ity (positive evidence) with a suitable value for C (Fig. 2(c)). The response of scientific realism, particularly to the im-
These data are therefore not conclusive for the choice between portant constructivist arguments and accounts, was the devel-
the two last alternative models, and the matter requires further opment of more realistic and sophisticated conceptions, which
research, e.g. using the relevant literature. But the object here recognize the constructive element of scientific activity and
 J Sci Educ Technol

Fig. 3 Simulations with Modellus for the fall of a parachute. (a) Ffr = asymptotic velocity. (F is the acting force, Ffr is the friction, W, m, v
const = Β, F = W-Ffr = mg-B. The model does not predict asymptotic and a are the weight, the mass, the velocity and the acceleration of the
velocity (vas). (b) Ffr = Cv, F = W-Ffr = mg-Cv. The model predicts parachute, and C is the constant of proportionality between Ffr and v.)
asymptotic velocity. The graphs show the velocities for C = 25, 50, 90, (Initial conditions and parameters used in all the three cases are as
and 130. (c) Ffr = Cv2; F = W-Ffr = mg-Cv2. The model predicts follows: m = 85 kg, g = 9.81 m/s2, y0 = 0, v0 = 0.)

accepts the existence of constructivistic episodes in the history other things, that even the most correct scientific methods
of science (e.g. the disputable determination of IQ or of the cannot eliminate the crucial influence of social and psycho-
taxonomy of the higher species, or when racist theories ap- logical factors and that if scientific approaches change, it is
peared as biologically conditioned) (see in Boyd 1992; Devitt more because the new approaches better represent the spirit of
1991). It argued, however, that science has developed mech- their times or give new impetus to research activity and em-
anisms to ensure objectivity in scientific knowledge, using as ployment than because they offer some new advance in the
arguments the fact that scientists modify or radically change knowledge of reality (Kuhn 1989).
their theories in order to bring them into accordance with the In this context, we find it useful to refer to the scientific
outer world, and that the view that scientific theories are social perspectivism developed by Ronald Giere (see, e.g. Giere
constructs cannot explain science’s successful predictions and 2006), using as a starting point the perspectival nature of col-
technological applications.8 Constructivists reply, among our vision in humans and of the observation tools used in
scientific research. Giere argues that ‘the grand principles
and axioms that objectivists cite as universal laws are better
8
Such successful applications and predictions, e.g. when they are based on understood as defining highly idealized models that character-
theories about entities of the microscopic world, would then appear as a mir-
acle if the existence of theoretical entities is not accepted: the non-miracles ize a theoretical perspective’ (p. 14) and that the ‘… general
argument (see in Boyd 1983; Devitt 1991). principles by themselves make no claims about the world, but
J Sci Educ Technol

more specific models constructed in accordance with the prin- for adequately preparing them to play a role in addressing
ciples can be used to make claims about specific aspects of the socio-scientific issues relating to, e.g. environmental degrada-
world.’ (p. 15). In his view ‘… perspectival realism provides tion, health, nutrition, and social and political imbalances. In
us a genuine alternative to both objectivist realism and social the context of this article, for example, using computer simu-
constructivism’ (2006, p. 14), since it both ‘makes room for lations in a framework of scientific perspectivism means that
constructivist influences in any scientific investigation’ (p. students are helped to interact with simulation in order to
15), to the extent and degree warranted in each case, and ‘... develop or compare and contrast models with empirical data
supports the rejection of all claims to absolute truths’ (p. 16) of and also to become aware of the limits to claims of absolute
strongly objectivist realism. Giere also makes it clear that per- truth in scientific matters and more generally.
spectival realism does not mean relativism, in the sense that
‘every perspective is regarded as good as any other’ (p. 13)
and that a comparative evaluation of perspectives is therefore Summary and Conclusions
meaningless.
Scientific perspectivism focuses and is based on two fun- Computer simulations constitute a major advance in contem-
damental characteristics of scientific models: their porary scientific research and are also an important education-
perspectival–conceptual nature, i.e. that a model reflects the al tool. In science education, there is an important body of
specific perspective of its underlying theory (e.g. the classical research on the educational possibilities of simulations, their
mechanical or electromagnetic perspective in physics, or ‘an impact on science teaching and learning, and the factors that
evolutionary perspective’ in biology), and their abstractive— affect their classroom effectiveness. A concise overview of
idealized nature, i.e. that a model deals with only some aspects this research is given in ‘Research on the Impact of
of the real systems and processes and only to a limited degree Simulations on Science Teaching and Learning’. In this re-
of accuracy. search, however, and also in the science education literature
The questions and constraints on the reliability of computer on models and modelling generally, there is a gap with regard
simulation models discussed in ‘Epistemology of Computer to the particular nature of computer simulations (their meth-
Simulations’ further strengthen the scientific perspectivism odology and epistemology) and its implications for science
approach, raising to scientific realists additional epistemolog- education. One reason for this may be that in most research
ical questions about computer simulating. Winsberg, for ex- work the simulations are studied mainly as tools for promoting
ample, proposes the notion ‘reliability without truth’ as a re- specific educational goals, that is, the focus is primarily on
placement for the ‘success implies truth’ of the scientific real- their instrumental aspect. In this article, we examine computer
ists, based on the fact that counterfactual acceptances are often simulations both as teaching tools and as cognitive outcomes,
introduced into simulation models and programs, such as the a differentiation also noted, e.g. in Louca (2004) and Luca and
artificial viscosity, an unphysically large value for viscosity, Zacharia (2008). In other words, we give weight not only to
that is broadly and successfully employed in order to dampen the educational aspect of simulations but also, and chiefly, to
out instabilities and thus achieve computationally tractable approaching them as cognitive results, i.e. as models that ex-
simulations in the field of computational fluid dynamics. ‘I plain the studied phenomena and enable their further investi-
argue that these kinds of model-building techniques, therefore, gation. Viewing simulations as cognitive results and as
are counter examples to the doctrine that success implies intended learning outcomes in science education requires a
truth—a doctrine at the foundation of scientific realism.’ … sound and coherent background in the methodology and epis-
‘The practice of using fictions in building credible simulations temology of simulations, which are a particular kind of model
is worthy of closer scrutiny by philosophers of science inter- as regards both structure and potential.
ested in the various arguments for and against realism.’ In this paper, we analysed the methodology and epistemol-
(Winsberg 2010, p. 121; see also in Lehnard 2007). ogy of computer simulations, with a view to fostering a fuller
The similar questioning of the fundamental epistemologi- and sounder understanding of modern scientific modelling,
cal conception of the relation of scientific claims to the real and discussed some of their implications for science educa-
world and the related arguments that have developed in sci- tion. We described methods, processes, and skills used in the
ence education usually reject extreme social-constructivist po- construction and validation of computer simulations and
sitions in favour of more moderate but sophisticated realistic discussed epistemological views concerning their reliability
views (e.g. Matthews 1994; Koponen 2007; Tala 2011; Wong and epistemic status. These elements help expand our concep-
and Hodson 2009). In our opinion, scientific perspectivism tions of scientific modelling and strengthen the educational
can meet the considerations and objectives of science educa- use of computer simulations, as required for teaching contem-
tion for an epistemologically solid and pedagogically ade- porary scientific methodology and for cultivating in high-
quate basis for supporting an understanding of NOS and fa- school students an up-to-date scientific and technological lit-
miliarizing students with scientific practices, in other words, eracy, which must include some understanding and experience
 J Sci Educ Technol

of the basic elements of computer simulating (e.g. the elemen- The analyses and implications discussed in this article re-
tary structure and numerical basis of simulation algorithms late more directly to model-based teaching and learning.
and programs and their possibilities and limits). Nonetheless, they also apply to all other, more or less tradi-
We then examined certain implications of that methodolo- tional, didactic approaches that already make use of or would
gy and epistemology, relating to the shaping of correct episte- be interested in introducing some modelling activities in order
mological perceptions of scientific practice and to the utiliza- to enrich the lesson or facilitate the teaching of specific topics
tion of simulations for doing science in the classroom. In this that are hard to handle with the traditional practices and tools.
context, we discussed the nature of educational simulations As we said in ‘Research on the Impact of Simulations on
compared to those used in scientific research, the reliability Science Teaching and Learning’, research on the use of com-
of the simulations encountered in the classroom or the public puter simulations notes that one of their important advantages
sphere, and the possibilities afforded by simulations for pro- is that they can be incorporated into different kinds of teaching
moting contemporary educational goals. In this last regard, we approaches, which are thereby enhanced. For example, the
compared the possibilities of educational simulations with literature reviewed in Rutten et al. (2012) shows that computer
those of laboratory experiments, which have a longer didactic simulations can reinforce traditional instruction, thereby im-
tradition in classroom activities, and argued that simulations proving learning outcomes; this also indicates an effective and
offer as much or more scope for supporting reasoning and realistic way of introducing and generalizing the use of com-
inquiry activities and abilities in the classroom as traditional puter simulations in contemporary practices in schools.
experiments. Apart from the dynamical and visualized repre- In this article, we have elaborated basic elements of the
sentations it offers, computer simulating permits quick repeat methodology and epistemology of computer simulations and
runnings, and thus facilitates the construction and immediate examined their importance for understanding and doing sci-
checking of numerous modified or alternative models and ence in the classroom. In our view, the main foci of the article,
their underlying hypotheses, a possibility that makes the use the methodology and epistemology of computer simulations
of simulations a particularly suitable teaching aid for and the educational use of simulations, play an important role
supporting contemporary educational approaches, including in the development of the epistemologically and pedagogical-
model-based teaching and NOS-oriented curricula. We then ly strong frameworks and learning environments required for
provided two examples to illustrate how the use of simulations supporting the new trends in science education.
can support students’ NOS understanding and their involve-
ment in scientific-like inquiry and reasoning activities: The Acknowledgements The author thanks the anonymous reviewers
for their helpful comments and suggestions, and Dr Heinz Isliker for
first concerns the use of simulations for understanding the
useful discussions on computer simulations in scientific research.
unifying potential of the fundamental theories and explains,
in the language of the model-based view, the coexistence of
Compliance with Ethical Standards
more than one model for the same phenomenon. The second
example concerns the use of simulations for advanced deduc- Conflict of Interest The author declares that she has no conflict of
tive reasoning and specifically for the evaluation of and choice interest.
between a number of alternative models for which there is
positive, negative, or inconclusive evidence; such activities Ethical Approval This article does not contain any studies with human
participants or animals performed by any of the authors.
provide an opportunity for students to practice sounder and
more advanced reasoning and decision-taking.
Publisher’s Note Springer Nature remains neutral with regard to juris-
Finally, we examined the epistemology of computer dictional claims in published maps and institutional affiliations.
simulations in relation to the formulation or choice of
the fundamental epistemological view regarding the rela-
tionship between scientific claims and the real world in
science education. The topic was discussed in the context References
of the philosophical debate on scientific realism, with the
observation that the epistemology of models suggests a Abd-El-Khalick, F., Bell, R. L., & Lederman, N. G. (1998). The nature of
science and instructional practice: making the unnatural natural. Sci
perspectivist approach rather than the extreme versions Educ, 82(4), 417–437.
of scientific realism and constructivism. Scientific Adúriz-Bravo. (2013). A ‘semantic’ view of scientific models for science
perspectivism incorporates both the objective and the con- education. Sci & Educ, 17(2–3), 147–177.
structivist aspects of science and, in our view, can provide Adúriz-Bravo, Α., & Izquierdo-Aymerich, Μ. (2009). A research-
a suitable basis for helping students develop epistemol- informed instructional unit to teach the nature of science to pre-
service science teachers. Sci & Educ, 18(9), 1177–1192.
ogies and abilities that they will need in order to handle American Association for the Advancement of Science (AAAS). (1993).
science-related questions in their personal lives and the Benchmarks for science literacy. New York: Oxford University
socio-scientific issues of their age. Press.
J Sci Educ Technol

Andaloro, G. V., Donzelli, V., & Sperandeo-Mineo, R. M. (1991). Giere, R. N. (1988). Explaining science: a cognitive approach. Chicago:
Modelling in physics teaching: the role of computer simulation. Int University of Chicago Press.
J Sci Educ, 13(3), 243–254. Gilbert, J. K., & Justi, R. (2016). Modelling-based teaching in science
Annetta, L. (2012). The books: learning science through video games and education. Switzerland: Springer International Publishing.
simulations. Sci Educ, 96(3), 566–568. Gobert, J., O’Dwyer, L., Horwitz, P., Buckley, B., Levy, S. T., &
Barab, S., & Dede, C. (2007). Games and immersive participatory simu- Wilensky, U. (2011). Examining the relationship between students’
lations for science education: an emerging type of curricula. J Sci epistemologies of models and conceptual learning in three science
Educ Technol, 16(1), 1–3. domains: biology, physics, & chemistry. International Journal of
Boyd, R. Ν. (1983). On the current status of the issue of scientific realism. Science Education, 33(5), 653–684.
Erkenntnis, 19, 45–90. Gramelsberger, G. (2010). Computerexperimente. Zum Wandel der
Boyd, R. N. (1992). Constructivism, realism, and the philosophical meth- Wissenschaft im Zeitalter des Computers. Bielefeld: Transcript
od. In: J Earman (ed), Interference, explanation, and other Verlag.
frustrations (p.p. 131–198), Essays in the Philosophy of Science, Grandy, R. E. (1992). Theories of theories, a view from cognitive science.
University of California Press, Berkley. In J. Earman (Ed.), Inference, explanation, and other frustrations.
Bunge, M. (1970). Philosophy of physics, Dordrecht (Holland). Comp: Essays in the philosophy of science (pp. 216–233). Berkeley:
Reidel Publ. University of California Press.
Burian, R.M. (1980). Empirismus. In: J. Speck (ed.), Handbuch Greca, I. M., Seoane, E., & Arriazzecq, I. (2014). Epistemological issues
wissenschaftstheoretischer Begriffe, Band 1. Göttingen. concerning computer simulations in science and their implications
Bybee, R. (1997). Achieving scientific literacy: from purposes to for science education. Sci & Educ, 23(4), 897–921.
practices. Portsmouth: Heilmann. Guala, F. (2002). Models, simulations, and experiments. In L. Magnani &
Cartwright, N. (1983). How the laws of physics lie. Oxford: Oxford N. Nersessian (Eds.), Model-based reasoning: science, technolo-
University Press. gies, value (pp. 59–74). New York: Kluwer.
Clement, J. J., & Ramirez, M. A. (Eds.). (2008). Model based learning Halloun, I. A. (2004). Modelling theory in science education. Dordrecht:
and instruction in science. Dortrecht. Springer. Kluwer Academic Publishers.
Clough, M. P., & Olson, J. K. (2008). Teaching and assessing the nature Hodson, D. (2014). Nature of science in the science curriculum: origin,
of science. Science & Education (special issue), 17(2–3), 143–114. development, implications and shifting emphases. In M. Matthews
de Jong, T. (2006). Technological advances in inquiry learning. Science, (Ed.), International handbook of research in history, philosophy and
312(5773), 532–533. science teaching (pp. 911–970). Dordrecht: Springer.
de Jong, T., & van Joolingen, W. R. (1998). Scientific discovery learning Hmelo, C., & Day, R. (1999). Contextualized questioning to scaffold
with computer simulations of conceptual domains. Rev Educ Res, learning from simulations. Computer & Education, 33, 151–164.
68(2), 179–202. Hughes, R. I. G. (1999). The Ising model, computer simulation, and
de Jong, T., Linn, M. C., & Zacharia, C. Z. (2013). Physical and virtual universal physics. In M. S. Morgan & M. Morrison (Eds.), Models
laboratories in science and engineering education. Science, 340, as mediators (pp. 97–145). Cambridge: Cambridge University
305–308. Press.
Develaki, M. (2007). The model-based view of scientific theories and the Humphreys, P. (2004). Extending ourselves: computational science, em-
structuring of school science programmes. Science & Education, piricism, and scientific method. New York: Oxford University Press.
16(7), 725–749. Irzik, G., & Nola, R. (2011). A family resemblance approach to the nature
Develaki, M. (2012). Integrating scientific methods and knowledge into of science for science education. Sci & Educ, 20(7–8), 591–607.
the teaching of Newton’s theory of gravitation: an instructional se- Jimoyiannis, A. (2010). Designing and implementing an integrated tech-
quence for teachers’ and students’ nature of science education. nological pedagogical science knowledge framework for science
Science & Education, 21, 853–879. teacher’s professional development. Comput Educ, 55(3), 1259–
Develaki, M. (2016). Key aspects of scientific modeling exemplified by 1269.
school science models: some units for teaching contextualized sci- Justi, R. S., & Gilbert, J. K. (2003). Teachers’ views on the nature of
entific methodology. Interchange, 47(3), 297–327. models. Int J Sci Educ, 25(11), 1369–1386.
Develaki, M. (2017). Using computer simulations for promoting model- Khishfe, R., & Abd-El-Khalick, F. (2002). Influence of explicit and re-
based reasoning. Epistemological and educational dimensions. flective versus implicit inquiry-oriented instruction on sixth graders’
Science & Education, 26, 1001–10027. views of nature of science. Journal of Research in Science Teaching,
Devitt, M. (1991). Realism and truth (2nd ed.). Oxford (UK) & 39(7), 551–578.
Cambridge (USA): Blackwell. Knuuttila, T. (2011). Modelling and representing: an artefactual approach
Dorn, W. S. (1975). Simulations versus models: which one and when. J to model-based representation. Studies in History and Philosophy of
Res Sci Teach, 12(4), 371–377. Science Part A, 42(2), 262–271.
Dowling, D. (1999). Experimenting on theories. Sci Context, 12(2), 261– Knuuttila, T., & Loettgers, A. (2013). Synthetic modeling and the mech-
273. anistic account: material recombination and beyond. Philos Sci,
Duhem, P. (1978). Ziel and Struktur der physicalischen Theorien. 80(5), 874–885.
Hamburg: Meiner. Koponen, I. T. (2007). Models and modelling in physics education: a
Galison, P. (1997). Image and logic: a material culture of microphysics. critical re-analysis of philosophical underpinnings and suggestions
Chicago: University of Chicago Press. for revisions. Sci & Educ, 16(7-8), 751–773.
Giere, R. (2009). Is computer simulation changing the face of experimen- Kuhn, T. S. (1989). Die Struktur wissenschaftlicher Revolutionen.
tation? Philosophical Studies, 143(1), 59–62. Suhrkamp-Taschenbuch, Frankfurt am Main (10. Aufl.).
Giere, R. N. (2006). Scientific perspectivism. Chicago: The University of Lakatos, Ι. (1974). Falsifikation und die Methodologie wissenschaftlicher
Chicago Press. Forschungsprogramme. In I. Lakatos and Musgrave, A. (Eds.),
Giere, R. N. (2001). A new framework for teaching scientific reasoning. Kritik und Erkenntnisfortschritt (pp. 89–189). Vieweg,
Argumentation, 15(1), 21–33. Braunschweig.
Giere, R. N. (1999). Science without laws. Chicago & London: Lederman, N. G. (2006). Syntax of nature of science within inquiry and
University of Chicago Press. science instruction. In L. B. Flick & N. G. Lederman (Eds.),
 J Sci Educ Technol

Scientific inquiry and nature of science (pp. 301–317). Dordrecht, Roth, W.-M. R, Woszczyna, C., & and Smith, G. (1996). Affordances and
The Netherlands: Springer. constraints of computers in science education. Journal of Research
Lehnard, J. (2007). Computer simulation: the cooperation between in Science Teaching, 33, 995–1017.
experimenting and modeling. Philosophy of Science., 74(2), 176– Rutten, N., van Joolingen, W. R., & van der Veen, J. T. (2012). The
194. learning effects of computer simulations in science education.
Lehnard, J. (2006). Surprised by a nanowire: simulation, control, and Computers & Education, 58, 136–153.
understanding. Philosophy of Science., 73(5), 605–616. Scalise, K., Timms, M., Moorjani, A., Clark, L., Holtermann, K., & Irvin,
Linn, M. C. (2003). Technology and science education: starting points, P. S. (2011). Student learning in science simulations: design features
research programs, and trends. International Journal of Science that promote learning gains. Journal of Research in Science
Education, 25(6), 727–758. Teaching, 48(9), 1050–1078.
Losee, J. (1990). A historical introduction to the philosophy of science. Schwarz, C. V., & White, B. Y. (2005). Meta-modeling knowledge: de-
Oxford: University Press. veloping students’ understanding of scientific modeling. Cognition
Louca, L. (2004). Case studies of fifth-grade student modeling in science and Instruction, 23(2), 165–205.
through programming: comparison of modeling practices and Sherrin, B., diSessa, A., & Hammer. (1993). Dynaturtle revised: learning
conversations. In Unpublished doctoral dissertation. MD: physic through collaborative design of a computer model.
University of Maryland, College Park. Interactive Learning Environments, 3(2), 91–118.
Luca, L. T., & Zacharia, Z. C. (2008). The use of computer-based pro-
Smetana, L. K., & Bell, R. L. (2012). Computer simulations to support
gramming environments as computer modelling tools in early sci-
science instruction and learning: a critical review of the literature.
ence education: the cases of textual and graphical program lan-
International Journal of Science Education, 34(9), 1337–1370.
guages. International Journal of Science Education, 30(3), 287–
323. Stöckler, M. (1995). Theoretische Modelle im Lichte der
Lunetta, V. N., & Hofstein, A. (1981). Simulations in science education. Wissenschafttheorie. Praxis der Naturwissenschaften—Physik, 1,
Science Education, 65(3), 243–252. 16–22.
Matthews, M. R. (2012). Changing the focus: from nature of science to Suppe, F. (1977). The structure of scientific theories (2nd ed.). Chicago:
features of science. In M. S. Khine (Ed.), Advances in nature of University of Illinois Press.
science research (pp. 3–26). Dordrecht, The Netherlands: Springer. Suppes, P. (1997). Perception, models, and data: some comments.
Matthews, M. R. (1994). Science teaching. New York: Routledge. Behavior Research Methods, Instruments, & Computers., 29(1),
McComas, W. F. (2008). Seeking historical examples to illustrate key 109–112.
aspects of the nature of science. Science & Education, 17(2–3), Tala, S. (2013). Knowledge building expertise: nanomodellers’ education
249–263. as an example. Science & Education, 20, 1323–1346.
Mellar, H., Bliss, J., Boohan, R., Ogborn, J. & Tompsett, C. (Eds) (1994). Tala, S. (2011). Enculturation into technoscience: analysis of the views of
Learning with artificial worlds: computer-based modelling in the novices and experts on modelling and learning in nanophysics.
curriculum. London: the Falmer Press. Science & Education, 20, 733–760.
Morrison, M. S., & Morgan, M. (1999). Introduction. In M. S. Morgan & Tala, S., & Vesterinen, V. Μ. (2015). Nature of science contextualized:
M. Morrison (Eds.), Models as mediators (pp. 1–9). Cambridge studying nature of science with scientists. Science & Education, 24,
University Press. 435–457.
Morgan, M. (2003). Experiments without material intervention: model van Fraasen, B. C. (1980). The scientific image. Oxford University Press.
experiments, virtual experiments and virtually experiments. In H. Webb, M. E. (2005). Affordances of ICT in science learning: implications
Radder (Ed.), The philosophy of scientific experimentation (pp. for an integrated pedagogy. International Journal of Science
216–235). Pittsburg, PA: University of Pittsburgh Press. Education, 27(6), 705–735.
Morgan, M. S. (1998). Learning from models. In M. S. Morgan & M. Winsberg, E. B. (2010). Science in the age of computer simulation. The
Morrison (Eds.), Models as mediators (pp. 326–346). Cambridge University of Chicago Press, Chicago and London.
University Press.
Wong, S. L., & Hodson, D. (2009). From the horse’s mouth: what scien-
Morrison, M. (2009). Models, measurement and computer simulation:
tists say about scientific investigation and scientific knowledge.
the changing face of experimentation. Philosophical Studies, 143,
Science Education, 93, 109–130.
33–57.
National Research Council (NRC). (1996). National Science Education Wu, H.-K. (2010). Modeling a complex system: using novice-expert
Standards. Washington, DC: National academy Press. analysis for developing an effective technology-enhanced learning
NGSS Lead States. (2013). Next Generation Science Standards: for environment. International Journal of Science Education, 32(2),
states, by states. Washington: The National Academies Press. 195–219.
Norton, S., & Suppe, F. (2001). Why atmospheric modeling is good Zacharia, Z. C. (2005). The impact of interactive computer simulations on
science. In C. Miller & P. Edwards (Eds.), Changing the atmo- the nature and quality of postgraduate science teachers’ explanations
sphere: expert knowledge and environmental governance (pp. 88– in physics. International Journal of Science Education, 27(14),
133). Cambridge, MA: MIT Press. 1741–1767.
Oh, P. S., & Oh, S. J. (2011). What teachers of science need to know about Zacharia, Z. C. (2007). Comparing and combining real and virtual exper-
models. International Journal of Science Education, 33, 1109– imentation: an effort to enhance students’ conceptual understanding
1130. of electric circuits. Journal of Computer Assisted Learning, 232(2),
Osborne, J., Collins, S., Ratcliffe, M., Millar, R., & Duschl, R. (2003). 120–132.
What ‘ideas-about-science’ should be taught in school science? A Zacharia, Z. C., Olympiou, G., & Papaevripidou, M. (2008). Effects of
Delphi Study of the Expert Community. Journal of Research in experimenting with physical and virtual manipulatives on students’
Science Teaching, 40(7), 692–720. conceptual understanding in heat and temperature. Journal of
Parker, W. (2009). Does matter really matter? Computer simulations, Research in Science Teaching, 45(9), 1021–1035.
experiments, and materiality. Synthese, 169, 483–496. Zeidler, D. L., Sadler, T. D., Simmons, M. L., & Howes, E. V. (2005).
Popper, K. R. (1959). The logic of scientific discovery. London: Beyond STS: a research-based framework for socioscientific issues
Hutchinson. education. Science Education, 89(3), 357–377.