Content uploaded by Umit Aslan
Author content
All content in this area was uploaded by Umit Aslan on Oct 30, 2018
Content may be subject to copyright.
Constructionism 2016
PAPERS
Feb 1-5, Bangkok, Thailand
230
Restructuration in Practice: Challenging a Pop-
Culture Evolutionary Theory through Agent
Based Modeling
Ümit Aslan, umitaslan@u.northwestern.edu
Learning Sciences and Center for Connected Learning, Northwestern University
Uri Wilensky, uri@northwestern.edu
Learning Sciences, Comp. Sci., and Center for Connected Learning, Northwestern University
Abstract
An ordinary, non-scientist person’s exposure to science through pop-culture is ever growing.
However, science is still believed to have a high threshold of entry. For most people, doing science
means going through rigorous training of literature, specialization, and learning formal
mathematics. Many people feel that they have no means to figure out which idea is scientifically
accurate and which one is inaccurate. More importantly, they only have the trustworthiness of the
source as heuristic for believing or not believing in a particular scientific message. In this paper,
we argue that agent-based modeling has the potential to lower the threshold of entry to science
and to empower people against the flood of scientific messages in pop-culture. To demonstrate
our theory in practice, we conduct a thought experiment in which we extract a much debated
scientific theory on the evolution of sexes from a BBC Earth documentary and show how one can
easily recreate and explore the assumptions of such a theory using the NetLogo agent-based
modeling environment. Then, we compare traditional formal mathematics based scientific analysis
approach with our agent-based modeling approach and show how the latter affords people with
little or no training in high level formal mathematics or evolutionary biology literature to challenge
scientific ideas.
Figure 1. BBC Earth’s documentary and the agent based model we developed based on it.
Keywords
Science, literacy, agent-based modeling, restructuration.
Constructionism 2016
PAPERS
Feb 1-5, Bangkok, Thailand
231
Introduction
Research shows that TV and the Internet are by far the main scientific information resources for
just plain folks (shortly JPFs), a term coined by Lave (1988) (Brossard and Scheufele, 2013;
Horrigan, 2006). JPFs hear about topics like human evolution, depletion of fisheries or the
disappearing of honeybees on a daily basis. We begin by asking the following question: How are
the JPFs supposed to sift through all the scientific messages that they get through media and
examine them? Given that JPFs do not have the time to deeply learn the literature, is there a way
to facilitate this process while not becoming an expert in science? Unfortunately, it seems like
JPFs currently only leverage heuristics such as trust in scientists or trust in resources when
forming opinions about on scientific information, such as believing or not believing in the climate
change (Hmielowski, Feldman, Myers, Leiserowitz and Maibach, 2013). This situation shows that
JPFs are dramatically underpowered against the scientific enterprise and finding ways to lower
the threshold of entry to science is an important challenge for learning scientists. In this paper, we
reiterate a constructionist ideal, democratizing science through restructuration (see Wilensky and
Papert, 2010), and argue that agent-based modeling has the potential to empower JPFs in their
casual interactions with scientific ideas through pop-culture.
So far, a common discourse in the science education literature has been justifying the need for
formal science education based on a need for scientific literacy (e.g., Holbrook and Rannikmae,
2007; NRC, 1996). Many researchers argued that school is the place for learning scientific
practices and preparing for a life dominated by scientific messages. Falk and Dierking argue that
this “school first” paradigm is so ubiquitous that most educators or policy makers take it for granted
(2010, p. 486). Yet, research has shown that JPFs learn about science mostly out of school (Feder,
Shouse, Lewenstein and Bell, 2009). Besides, retention of even basic scientific knowledge learned
through formal schooling has always been problematic (e.g. Custers, 2010). Based on such
evidence, it becomes increasingly clear that learning about the scientific method and memorizing
scientific facts is not enough; facts are easily forgotten and JPFs neither have tools nor time to
practice the scientific method. They need better ways to deal with the scientific information thrown
at them in their daily lives. We believe that studying JPFs’ out of school interactions with science
can inform us in restructuring science education in formal settings. Therefore, we focus on JPFs
interactions with science through pop-culture.
Although we agree with the calls for scientific literacy, we believe that we need to rethink its
framing. According to Papert, “becoming literate means thinking differently than one did
previously” (1993, p. 10). Similarly, diSessa (2001) argues that small changes in learnability can
have far greater impact on the eventual impact and theorizes that computers can be the technical
foundation of a new and enhanced literacy. Lastly, Wilensky and Papert (2010) argue that we
need to recalibrate the focus of learning from the means to the object of it. Wilensky and Papert
particularly show how agent-based representations are superior to equational representations of
science by analyzing both major representational shifts in the history, such as the effects of
switching from Roman numerals to Arabic numerals, and strengths of agent-based
restructurations of modern scientific phenomena such as kinetic molecular theory and materials
science.
Building on these theories, we claim that agent-based modeling has both the power and the
potential to play a critical role in lowering the threshold of entry to science and facilitating
democratization of scientific knowledge. In addition, we coin the term agent-based literacy as a
promising approach to rethink JPFs’ everyday interactions with scientific ideas, as well as
reframing the very construct of scientific literacy. In particular, we focus on an analysis of how
agent-based modeling could empower JPFs to challenge scientific ideas presented in pop-culture.
To illustrate our position, we run a thought experiment and demonstrate how a hypothetically
Constructionism 2016
PAPERS
Feb 1-5, Bangkok, Thailand
232
agent-based literate JPF can react to a still disputed evolutionary biology theory from a TV
documentary and easily challenge it using the NetLogo agent-based modeling environment
(Wilensky, 1999) without needing to become an expert in the domain. We also discuss how agent-
based modeling differs from the traditional scientific method in challenging the same evolutionary
theory in terms of threshold of entry for non-scientist people.
Evolution of Sexes: Challenging a Pop-Culture Theory using
Agent-Based Modeling
Let’s conduct a thought experiment and explore the learning of an imaginary JFP named Alex,
who enjoys following BBC Earth’s channel on YouTube. In April 2015, Alex would come across a
short documentary titled “Simple sex and the birth of gender” (BBC Earth, 2015), which presents
ideas on the evolution of the sexes. Following is the full excerpt of the documentary:
“The sex is a force to battle because females have a score to settle. And it all started about a billion years ago
when simple single celled animals like paramecium rocked the world. Only individuals virtually identical in shape
and form but they differ genetically. Only different genetic types can come together and reproduce. They fuse their
membranes and exchange some of their cell contents. In this form of simple sex, they part after mating as new
individuals. There can be dozens of sexes in these animals, highlighted as different colors, and there are no battles.
But that all changed long ago when one of them cheated. A quirk of evolution made one of the sexes smaller and
this turned out to be one of the most significant events in life on earth. The smaller types were more mobile and so
monopolized the matings. But the mates of the little cheats could only produce viable offspring if they grew fat to
compensate for their partner’s lack of provision. Now one sex was on a path to obesity, the other to athletic
swimming. The presence of these sex midgets drove all the other sexes to extinction, except for the huge fatty
cells. The small sex, those that cheated, became sperm and the large sex eggs. Gender as we know it was born.”
Now let’s assume Alex is not literate in agent-based modeling. She would feel this theory
compelling and probably believe in it because BBC Earth is a trustworthy news source. However,
directly accepting the BBC Earth’s theory (BET, in short) could result in Alex developing some
misconceptions about the process of evolution. Yes, BET sounds scientifically accurate initially,
but some words used by the narrator such as cheated and quirk, might have led Alex into believing
that (1) the process of evolution is one of miracles, and (2) individual agents are consciously trying
to gain selective advantage. Conversely, if we assume Alex is literate in agent-based modeling,
she would be able to test BET by developing a model in an agent-based modeling environment.
In the following section, we describe how one can replicate BET in the NetLogo agent-based
modeling environment without any exposure to the evolutionary biology literature.
Modeling BBC Earth’s Theory in NetLogo
In order to explore how Alex might go about exploring BET, we went through the process of
creating a model that can illustrate how an agent-based literate JFP would approach such a
question. Due to the fact that we wanted to think through the lens of a JPF, we did not collaborate
with any evolutionary biologist in this process. We developed a simple model of BET and
conducted a simple, non-rigorous analysis of the model output. In this section, we begin by
explaining our modeling decisions and then describe how the model works.
First step in modeling BET is extracting the assumptions from the documentary. There are six
main assumptions of BET: (1) smaller cells move faster, (2) a minimum total size is required for
two organisms to reproduce, (3) the evolution of sexes happened among simple single celled
organisms, (4) there were more than two mating types initially, (5) mating type is a single gene/trait
and it is represented by different colors in the documentary, (6) mating type of an organism and
Constructionism 2016
PAPERS
Feb 1-5, Bangkok, Thailand
233
its size are somehow connected. On the other hand, BET gives no explanation on the mechanisms
of evolution. Therefore, we have to fill in the blanks with our best guesses. For instance, there is
no mention of life cycles of these organisms and the population size so we need to figure out how
many agents needed in the system and how would they die. Because we have limited computing
power, we decided to have a range of 0-2500 agents in the model. We also decided to have a
carrying capacity for the system so we killed agents randomly if the number of agents exceeds a
certain number. However, we did not put any natural death mechanism to the system at all. Next,
BET assumes that two organisms need a certain amount of material so that their offspring survives
but it does not mention what happens if this criterion is not met. Thus, we decided not to allow two
organisms to initiate mating if their total size does not exceed a threshold. Third, BET talks about
fusion, but it does not say anything about how these organisms are reproducing. In our model, two
mating agents hatch new baby agents and then die immediately.
In addition, we wanted to create a model in which the user can change parameters and test
different assumptions freely. For some assumptions, we placed an on/off switch (see Figure 2).
For example, we wanted to see if speed is really a selective pressure in such a system, so we
created a “smaller-is-faster?” switch. When this switch is turned on, smaller agents are faster than
bigger agents. When this switch is turned off, all agents move with the same speed. Moreover, we
did not have any idea on how big or small organisms would be. We also did not know what was
the range of size in this organisms, so we had to go with arbitrary values for organism size and
size mutations. Still, we wanted to have some sort of tinkering space so we created sliders for
each variable such as minimum-size and minimum-total-size-for-mating.
Figure 2. The evolution of sex model based on BBC Earth’s theory.
After some tests, we decided to set the variables as shown in Figure 2 for our demonstrative
experiment. In terms of defining agent behavior, each agent in our model represents a single-
celled organism and follow a very simple set of commands. In each time tick, each agent changes
its direction randomly and moves forward. Then, it checks if there is another agent on the same
place with a different mating type (color) and enough size to meet the minimum size requirement
for reproducing. If it can find such a partner, it initiates mating with its partner. In the mating
process, the mating type and the average size of prospective babies are picked randomly from
the parents’ trait pool. Next, if two-babies-per-mating? switch is turned off, multiple babies are
hatched based on total size of parents and the size of prospective babies. Otherwise, parents
hatch just 2 baby agents after each mating. Finally, both parents die.
Results
We ran the model from a state where all agents start with the same size and with 4 different setting
combinations: (1) smaller is faster and more than two babies can be produced, (2) speed is
constant and more than two babies can be produced, (3) smaller is faster and just two babies are
produced, and (4) speed is constant and just two babies are produced. At each run, we waited for
Constructionism 2016
PAPERS
Feb 1-5, Bangkok, Thailand
234
the population to come to an equilibrium point and took screenshots of the graphs. Due to space
limitations, we will not be able to dive deeper into the results of the model. However, we share
typical final states of all four combinations in Figure 3:
Figure 3. Typical results of four parameter different combinations.
We believe if our plain folk Alex developed such a model, or just even ran it, she would see the
emergence of a dichotomous state with numerous extremely small cells and a few very big cells
in first two settings. Therefore, she would realize that BET has some explanatory value if the
system is setup in this particular way. However, she would also see that there is neither cheating
nor a quirk in this process. It all happens through stochasticity and selective pressures in the
system. When she took a closer look at the difference between graphs, she would get the chance
to test the assumptions of BET, too. Interestingly, there is no substantial difference between
smaller cells being faster or all cells being the same speed when we compare the outcome of
setting 1 to setting 2. Plus, there is no evolution of small-big dichotomy in settings 3 and 4 at all. It
seems like producing more agents in a mating is more important than the speed of individual
agents. Particularly, no significant difference between setting 1 and setting 2 shows that the
difference is not actually in speeds of the cells. Small agents dominate matings simply because
there are many more of them in the system. When this advantage is taken out in settings 3 and 4,
big-small dichotomy does not evolve and the model does not end up with a population with two
final mating types.
Discussion
Although our model was able to produce a dichotomous state, these findings should not be
considered as the proof of how sexes evolved. Instead, our agent-based exploration should be
interpreted as a modest exploration of BET and disproof of its assumption about the selective role
of agent speed in the hypothetical environment described by it. Likewise, one can rightfully argue
that our model is scientifically incorrect because we developed it based on BBC Earth’s narration,
not on a scientific literature review. However, it still allowed our imaginary JFP Alex to start from
somewhere. If Alex had no tools like NetLogo, she would have to first start reading the literature
on evolutionary biology. Consider the following excerpt from a paper written by Bulmer and Parker
(2002) on this topic:
“In the ancestral unicellular state the gametic and zygotic survival functions, g(m) and f(S), are likely to be
similar in shape and location, leading to isogamy. The development of multicellularity may leave g(m) relatively
unchanged, but will push f(S) to the right as the need to provision the zygote increases, eventually leading to
anisogamy. This is most clearly seen when the survival functions are sigmoidal, exemplified by the inverse
exponential Vance function in equation (2.2). The situation is more complicated in the less likely case when
the survival functions are concave, exemplified by the complementary exponential function in equation (2.12).”
Constructionism 2016
PAPERS
Feb 1-5, Bangkok, Thailand
235
Unfortunately, the literature on the evolution of sexes is very challenging for outsiders. In the first
place, Alex has to know that this particular phenomenon is called anisogamy, “the occurrence
within a population of two gamete types of different size” (Bell, 1978, p. 73), in order to be able to
even start a literature review. Then, it is important to realize that there is still no consensus on how
anisogamy actually evolved (Blute, 2013). Beyond the literature barriers, Alex also has to know at
least the basics of evolutionary game theory (Smith, 1982) to be able to conduct further
mathematical analysis in order to challenge BET. Following is another excerpt from Bulmer and
Parker’s paper (2002, p. 2382):
“We assume that fusions occur between gametes of ‘+’ and ‘–’ individuals (mating types). All + individuals
produce n1 gametes of size m1, and – individuals produce n2 gametes of size m2, with ni = M/mi, where M is
the fixed budget for reproduction. Zygotes are of size S = m1 + m2, and the survival to adulthood of a zygote
of size S is f(S ). The chance that a gamete of size m will survive to mate is g(m).
Under these assumptions, the reproductive fitness of + individuals is
,
and the reproductive fitness of – individuals is
.”
This is actually the start of Bulmer and Parker’s mathematical analysis and they only talk about
how they are going to calculate fitness functions of individuals. Still, even if Alex understands some
parts of their paper, it is unfortunately impossible for her to move on from there if she does not
have a good background in formal mathematics. Now consider the following code from our agent-
based model (gray lines starting with a semicolon are our comments):
to maybe-mate
; I check if there are any cells here, who is of a different mating type big enough to mate with me
let potential-partners other cells-here with [
mating-type != [mating-type] of myself
and (size + [size] of myself > min-total-size-for-mating) ]
; if I bump into any potential partners
if any? potential-partners [
; I pick one of them randomly and initiate the mating with it
let my-partner one-of potential-partners
; First, we pool our traits so that our offspring inherits the mating type from either parent randomly
let mating-type-of-babies one-of list ((mating-type) ([mating-type] of my-partner))
; Our offspring inherits their average size based on the mating type
let average-size-of-babies size
if mating-type-of-babies != my-mating-type [ set average-size-of-babies [size] of partner ]
; based on the total size of me and my partner, the count of baby cells is calculated
Constructionism 2016
PAPERS
Feb 1-5, Bangkok, Thailand
236
let number-of-babies ceiling ((my-size + partner-size) / average-size-of-babies) + 1
; then we hatch baby agents one by one
hatch number-of-babies [
; each baby inherits the same mating type
set mating-type mating-type-of-babies
; and then mutates its size by adding or subtracting a small random number to the average baby size
set size average-size-of-babies + random-float size-mutation - random-float size-mutation ]
; after mating, me and my partner die!
ask my-partner [die]
die ]
end
NetLogo code of our agent based model is much more readable and easier to understand
compared to mathematical analysis of the evolution of sexes. The primitives of NetLogo such as
one-of, hatch and die, require little to no explanation. There is no scientific jargon in our code at
all. More importantly, our agent-based model focuses on a concrete process whereas Bulmer and
Parker’s study focuses on an abstract mathematical analysis. In other words, Alex does not need
to be trained to acquire a specific way of thinking.
As a side note, it is worth mentioning that we are not trying to contribute to evolutionary biology
literature, at all, while Bulmer and Parker’s analysis is built on a very rigorous scientific analysis.
Our model, in that respect, is not an effort to do science and it is not comparable to Bulmer and
Parker’s work. Our goal is to lower the threshold whereas Bulmer and Parker are trying to raise
the ceiling. We must also mention that our agent-based code is not a direct equivalent of Bulmer
and Parker’s equations. The two approaches are distinctly different from each other by nature. In
agent based modeling, our focus is on defining the behavior of individual agents and investigating
emergent phenomena (Wilensky, 2001), while evolutionary game theory focuses on figuring out
strategies that maximize the fitness using equational models (McNamara and Weissing, 2010).
Nevertheless, for our imaginary JFP Alex, who did not study biology formally, challenging BET
would have been extremely hard and frustrating if she was agent-based illiterate. She would have
no joy in exploring such a domain at all and give up quickly after realizing that she needs much
more expert knowledge in order to answer her questions on the theory. However, as an agent
based literate person, she can potentially enter the world of evolutionary biology, discover that
BET has some flaws, and begin to explore more about the evolution of sexes. It is true that she
cannot produce any new scientific theories based on this model, but she can still feel satisfied
because this model has potential to empower her against a scientific idea that she is exposed
through pop-culture. If she was not literate in agent based modeling, she might have just chosen
to believe in the BBC’s narrator because it was published by a credible source or she might have
rejected it based on her intuitions or superstitions.
Conclusion and Future Directions
In this paper, we proposed agent-based literacy as a way to rethink JPFs’ interaction with scientific
information in daily life. We argued that people are continuously exposed to a flood of scientific
messages through pop-culture but they do not have reliable means to evaluate these ideas other
than heuristics about the trustworthiness of resources. The traditional method of challenging these
pop-media theories would be diving into scientific literature, which requires understanding heavy
Constructionism 2016
PAPERS
Feb 1-5, Bangkok, Thailand
237
jargon and formal mathematical analysis. We argued that agent-based modeling literacy would
support a richer participation in scientific practices and conducted a thought experiment based on
a pop-culture evolutionary biology theory that we extracted from a documentary. We showed that
an agent-based literate person has the power to challenge such a theory and demonstrated how
one would develop an agent-based model of this particular theory. Then we compared agent-
based modeling with evolutionary game theory analysis.
Based on our current analysis, and also based on decades of constructionist research, we strongly
argue that a major restructuration in science is much needed (e.g. Papert, 1993; diSessa, 2001;
Wilensky and Papert, 2010; Wilkerson-Jerde and Wilensky, 2010). As part of this restructuration,
we need to reconsider the way we conceptualize scientific literacy and teach science in schools.
In our opinion, the advent of a computer-based literacy is far more critical than the current
operationalization of scientific literacy. We believe that JPFs need more than just heuristics, simple
facts, or basic knowledge of the scientific method. We need to lower the threshold of entry to
science so that they have better tools and objects to challenge scientific messages thrown at them
by entertainers, politicians, or policy makers. In this respect, agent-based modeling is a promising
restructuration of science. It is much easier to learn and use compared to the scientific method
and complex formal mathematical analyses. It is also much more accessible for JPFs who has
neither time to conduct scientific experiments nor access to scientific facilities.
However, we are cautious that we do not propose agent-based modeling as a total replacement
for current scientific practices. It would, of course, be absurd to reach in such a conclusion. In
addition, we are we are aware of the flaws of our agent-based exploration of the evolution of sexes.
In the future studies, we hope to collaborate with evolutionary biologists and develop a better
model of anisogamy with correct scientific assumptions and mechanisms. We are also aware of
the fact that just one thought experiment has little or no generalizability on the effects of agent-
based literacy on lowering the threshold of entry to science. Therefore, our main challenge in
future studies is to study the challenges of becoming agent-based literate for JPFs, the potential
ways to facilitate this kind of scientific explorations for them, and learning gains of such
experiences.
References
BBC Earth. (2015, April 8). Simple sex and the birth of gender - Battle of the Sexes in the Animal
World [Video file]. Retrieved from https://www.youtube.com/watch?v=dWa9NmNvChU
Bell, G. (1978). The evolution of anisogamy. Journal of Theoretical Biology,73(2), 247-270.
Blute, M. (2013). The evolution of anisogamy: more questions than answers.Biological Theory,
7(1), 3-9.
Brossard, D., & Scheufele, D. A. (2013). Science, new media, and the public.Science, 339(6115),
40-41.
Bulmer, M. G., & Parker, G. A. (2002). The evolution of anisogamy: a game-theoretic
approach. Proceedings of the Royal Society of London B: Biological Sciences, 269(1507), 2381-
2388.
Custers, E. J. (2010). Long-term retention of basic science knowledge: a review study. Advances
in Health Sciences Education, 15(1), 109-128.
diSessa, A. A. (2001). Changing minds: Computers, learning, and literacy. Mit Press.
Falk, J., & Dierking, L. (2010). School is not where most Americans learn most of their
science. American Scientist, 98(6), 486.
Feder, M. A., Shouse, A. W., Lewenstein, B., & Bell, P. (Eds.). (2009). Learning Science in
Informal Environments:: People, Places, and Pursuits. National Academies Press.
Constructionism 2016
PAPERS
Feb 1-5, Bangkok, Thailand
238
Hmielowski, J. D., Feldman, L., Myers, T. A., Leiserowitz, A., & Maibach, E. (2013). An attack on
science? Media use, trust in scientists, and perceptions of global warming. Public Understanding
of Science, 0963662513480091.
Holbrook, J., & Rannikmae, M. (2007). The nature of science education for enhancing scientific
literacy. International Journal of Science Education, 29(11), 1347-1362.
Horrigan, J. B. (2006). The Internet as a resource for news and information science. Pew Internet
& American Life Project.
Lave, J. (1988). Cognition in practice: Mind, mathematics and culture in everyday life. Cambridge
University Press.
McNamara, J. M., & Weissing, F. J. (2010). Evolutionary game theory. Social behaviour: Genes,
ecology and evolution, 109-33.
National Research Council (Ed.). (1996). National science education standards. National
Academy Press.
Papert, S. (1993). The children's machine: Rethinking school in the age of the computer. Basic
Books.
Smith, J. M. (1982). Evolution and the Theory of Games. Cambridge university press.
Wilensky, U. (1999). {NetLogo}.
Wilensky, U. (2001). Modeling nature’s emergent patterns with multi-agent languages. In
Proceedings of EuroLogo (pp. 1-6).
Wilensky, U., & Papert, S. (2010). Restructurations: Reformulations of knowledge disciplines
through new representational forms. Constructionism.
Wilkerson-Jerde, M. H., & Wilensky, U. (2010). Restructuring change, interpreting changes: The
deltatick modeling and analysis toolkit. Proceedings of constructionism.
