Three-minute Constructionist Experiences
Ken Kahn, toontalk@gmail.com
Computing Services, Oxford University
Howard
Noble, howard.noble@oucs.ox.ac.uk
Computing Services, Oxford University
Arthur Hjorth, arthur.hjorth@u.northwestern.edu
Learning Sciences, Center for Connected
Learning, Northwestern University
Fabio
Ferrentini Sampaio, ffs@nce.ufrj.br
Electronics Computing Center, Federal
University of Rio de Janeiro
Abstract
Constructionist learning experiences
typically take time: time to build and run programs – typically even more time
to learn how to program. As authors of constructionist tools and learning
designs we faced the challenge of giving visitors to the Royal Society Summer
Science Exhibition an authentic constructionist experience when most visitors
were expected to stay only a few minutes. Furthermore we needed to help the
visitors learn something about the subject matter of the exhibit – the spread
of viruses.
To meet these challenges we built and
exhibited the Epidemic Game Maker (http://m.modelling4all.org/p/en/sse.html).
The Epidemic Game Maker starts with a bare-bones model of an epidemic. Users
can enhance the model and add game play elements. Rather than exposing them to
the complexity and full power of the Behaviour Composer (Kahn & Noble 2010)
or NetLogo (Wilensky & Rand in press), we built a simplified interface
customised for building games and models of epidemics on top of the Behaviour
Composer. Interactions are limited to toggling twenty check boxes. Each check
box replays a different sequence of changes to the model that we pre-recorded.
For the adventurous user who wishes to go deeper with a few clicks the full
interface can be exposed where they can explore and alter the agents, behaviours,
and underlying code that constitute their model. The generated games can be
based upon a model significantly richer than the initial model. Game play can
be introduced by adding buttons to enact public health authority actions. All
of this is done on a single page in a web browser – an interface that visitors
are already familiar with.
Here we report on the design of the
Epidemic Game Maker and our impressions of the more than one thousand visitors
who created games during the exhibition.
Keywords
Agent-based modelling, Behaviour
Composer, NetLogo, Epidemic Game Maker
The Epidemic Game Maker
We designed the Epidemic Game Maker (EGM)
so that users with no computer programming knowledge, little computer
experience, and the bare minimum knowledge about how infections spread can
§ construct
and run epidemic models in a minute or two
§ construct
and play increasingly complex epidemic games in five to fifteen minutes
§ construct
games that help them learn about the dynamics of epidemics and the trade-offs
of different public health interventions
§ can
continue to play and share their games at home or school
§ learn
about computer modelling
§ have
fun
To address the first requirement we provide
a ready-made model that consists of a population of students who travel daily
from home to school and back. Initially one student is infected. When an
infected person is at the same location as a susceptible person then with
specified odds the infection is transmitted. People recover from infections
after a specified amount of time. This is based upon the standard
susceptible-infected-recovered model of epidemics (Scherer & McLean 2002).
The opening screen encourages them to run the model.
Figure 1 – Initial ‘home
page’ of the Epidemic Game Maker
(the check boxes are
explained below)
After clicking on the ‘Play your game’ tab,
the Java applet for the game is loaded into the user’s browser.
Figure 2 - Initial base game
screen
There are two kinds of simulation games:
(1) simulations where the player or players are among the individuals being
simulated (e.g. controlling a fish that is part of a simulated school of fish)
and (2) simulations where the player has global control (often called ‘god
games’). The EGM can be used to make pure models or models with game elements.
With minimal settings the EGM makes a simple epidemic model that lacks game
play elements. At the top of the interface there is an area for messages and a display of
the simulated time. Buttons are available to run, pause, or reset the
simulation. The world is portrayed as a circle of houses with a school in the
centre. Students are at home and one is infected (in Figure 2 the infected
student is at approximately 11 o’clock). The smiley faces act as a health
gauge. At the bottom graphs of the infected, susceptible, and recovered
populations are drawn.
The model can be run multiple times and due
to the stochastic nature of the transmission of infections each run is
different.
Figure 3 - A typical graph
showing two runs
The constructionist aspect of the Epidemic Game
Maker is supported by tick boxes that add or remove model and game enhancements
as well as fine control over parameters. The model changes supported are
§ Add
work places – Introduces adults that regularly
commute to work
§ Add
virus trails – Viruses that can infect people are
left behind by infected people
§ Add
more schools – Adds three more schools
§ Students
go to closest school – Students go to the school
closest to their home
Users can explore and discover interesting
interactions between these enhancements. Adding more schools and requiring
that students go to the local school limits the epidemic to a single school.
Introducing adults then provides a means for an infection to jump from school
to school (by a student infecting a parent at home who infects a co-worker from
another neighbourhood who infects one of their children). This pattern is
similar to how a disease can spread between countries via air travellers.
Users can also add buttons to introduce
game elements that enable players to try to stop the epidemic. The supported
enhancements are
§ School
closing – When schools are closed students stay
home. Keeping schools closed has societal costs and they will be automatically
re-opened if funds run out.
§ Voluntary
quarantine ad campaign – Every button click reaches
a specified proportion of the population to stay home once they are aware that
they are infected.
§ Hand
washing ad campaign - Every button click reaches a
specified proportion of the population causing them to wash their hands
frequently thereby reducing the odds of acquiring the infection from viruses
left behind.
§ Catch
It, Bin It, Kill It ad campaign - Every button
click reaches a specified proportion of the population to use tissues to reduce
the trail of viruses they leave behind.
Each run of an ad reduces the remaining
budget. The latter two only make sense if ‘Add virus trails’ has been
added to the model. All of these enhancements can spark good discussions. Does
school closing really cause children to stay home? What if the model was enhanced
with shopping centres where the students might go when school is closed? A ‘Hand
washing ad campaign’ encourages people to act in their own best
self-interest to reduce their odds becoming ill while the ‘Catch It, Bin It,
Kill It ad campaign’ encourages people to stop harming others.
The effects of all of these enhancements
depend upon the values of model parameters. Users can choose to add sliders to
explore the consequences of different values. The sliders they can add are
§ Infection
odds
§ Encounter
rate
§ Infection
duration
§ Symptoms
delay
§ Virus
duration
§ Trail
reduction factor
§ Infection
from trails odds
§ Hand
washing odds factor
§ School
closing cost
§ Reach
of hand washing ad
§ Reach
of stay home ad
§ Reach
of Catch It, Bin It, Kill It ad
To enable visitors to continue to play and
enhance their game the EGM generates a four-digit serial number for each game.
We gave each visitor a nicely printed card where they could write down their
serial number. The card included the project URL (modelling4all.org) where they
can enter their serial number to restore all of their settings.
Figure 4 – A game with all
enhancements and four sliders
The Royal Society Summer Science
Exhibition
The Royal Society holds the United
Kingdom’s most prestigious science exhibition every summer. We participated in
the 2010 event (http://seefurtherfestival.org/)
which was expanded to celebrate the 350th anniversary of the Royal
Society. Fifty thousand people visited the exhibition over ten days. This
included almost two thousand registered students and 240 teachers. 1750 VIPs
(including the Queen of England) came to a closed showing. The EGM was part of
the exhibit on ‘Emerging infections: viruses that come in from the wild‘ (http://seefurtherfestival.org/exhibition/view/emerging-infections-viruses-come-wild).
The main aim of the Royal Society event was
to provide the general public with an opportunity to speak to real scientists
about their field of research. We were asked to focus primarily on 12-16 year
olds but to expect people of all ages. The EGM allowed us to discuss both the
factors that determine the spread of a virus and how computer models might
contribute to our understanding.
Experiences of a thousand visitors ranging from young
children to fellows of the Royal Society
During the exhibit over 1100 games were
created using the EGM by approximately a thousand visitors (many games were made
by small groups). In the two weeks following the exhibition 200 games were made
by on-line visitors. It was noticeable that people varied in terms of how
willing they were to experiment with the model as opposed to discuss epidemics
with members of the Oxford team hosting the stand. As the exhibition continued
we arrived at ways to structure conversation so that:
§ Those
who were quick to centre their attention on experimenting with the EGM could be
encouraged to generate ideas for improving the model
§ Those
who preferred to discuss epidemics orally would turn to the EGM to test the
assumptions they held in their head.
In both cases the art of conversation was
to arrive at a point with each visitor gained a healthy level of scepticism
about computer modelling i.e. they should neither treat the model as a black
box for predicting the future, nor just as simply a toy that did not match the common
sense model they held in their head.
In this way we tried to give each visitor
the intuition that:
§ Computer
models of epidemics are useful because they augment our ability to think about systems
that are too complex to think about in our heads alone.
§ All
models of epidemics are built by making very important assumptions about how
people will behave in a given situation, and many assumptions they might think
are important could have been left out for the sake of simplicity.
Some nice examples that we feel demonstrate
a positive outcome from using the EGM:
§ Many
visitors were very quick to point out that in the middle of London it is very
important to model how people move between home, work and school i.e. whether
they walked, cycled, took the bus, tube or boat.
§ Particularly
teenage girls were quick to point out that any public health intervention that
focussed on improved personal hygiene (washing hands) would be most effective
if targeted at boys.
§ Most
teenagers who engaged with the model were adamant that a stay-at-home policy
would never work – teenagers would instead congregate at someone’s house or a
park. (They correctly pointed out that this would likely negate any gains from
closing the school).
§ Some
young children (6-10 year-olds) built a series of games and demonstrated a
clear understanding the model and the interventions. Younger children enjoyed
playing with the game – a few played for more than ten minutes.
§ Many
people questioned the assumptions built into the model pertaining to the cost
of closing schools and workplaces to the economy.
§ A
few learners asked to see the underlying code. One “improved” the model by
adding one of the generic Behaviour Composer micro-behaviours “wander randomly”
to a school. This caused the school to move around on the screen, and all its
pupils to chase around after the school. While this is not directly related to
learning about epidemiology, we felt it was an interesting example of a learner
opening up the black box and seeing the model as something malleable that he
could do with whatever he pleased.
Of course we would have loved to have spent
more time with people to support them in building their own assumptions and
ideas into an enhanced version of the EGM. In the few brief minutes we had with
visitors we hope they were left with at least a latent desire to do so one day.
Theoretical underpinnings
We wanted the EGM to work
as an object to think with for the learners when engaging with new
knowledge of epidemics. It was important to us that the EGM would facilitate
thinking and conversation about epidemics that learners could relate
meaningfully to their own lived lives.
Epidemiological
modelling is largely taught using variations of the SIR
(Susceptible-Infected-Recovered) model, using sets of deterministic
differential equations that predict the relationship between changes in the
three categories of people. Because of the use of differential equations, the
conventional SIR model requires a high level of knowledge of calculus, and is
typically not taught until the undergraduate level. However, using Agent Based
Modelling as a restructuration (Wilensky & Papert, 2010), we were able to
meaningfully engage young learners in discussing and thinking about this
complex issue.
(Wilensky & Resnick,
2006) argue that Agent Based Modelling can facilitate an ‘embodied modelling
approach’ by asking learners to think like the agents they are modelling. By
doing so they are able to break down the behaviours of agents into bits of code
and generate computational theories about the relationships between them. We
took this idea as a starting point for the design of the EGM. By focusing on the
lives of our learners, the main question that we hoped to engage our learners
in was, “Given this situation or policy or intervention, is this what you would
do?” As we demonstrated above in the examples of learning with the EGM, those
were exactly the questions that people asked of the EGM in order to both make
sense of the model, and to offer informed critiques of it. For instance, in the
example of the stay-at-home policy not being effective, the main critique
offered by learners was that they would in fact not stay at home, but go and
hang out with their friends. This critique was drawn from learners’ own lives,
and illustrated to us that learners were able to engage with this complex issue
in a deep and critical manner.
Finally, we wanted to
engage learners in thinking about the relationships between their own and their
family’s behaviours; the properties of the virus; the policies enacted; and the
aggregate outcome of the epidemic model. (Wilensky & Resnick, 1999) argue
for an ‘emergent view’ of levels in complex systems in which aggregate outcomes
appear at ‘levels’ that emerge out of the complex interactions between agents.
By allowing learners to add or remove complexities (number of schools,
workplaces, virus trails, etc.) our hope was that learners would experience
these emergent levels and see that what happens at the aggregate level in the
model is not a hard coded, black box phenomenon, but simply the result of the
complexities that learners added to the model.
Building other ‘Game Makers’ in the
Behaviour Composer
The Epidemic Game Maker was implemented by
building it on top of the Behaviour Composer. This approach provides a smooth
integration but requires a deep understanding of the implementation of the
Behaviour Composer. Subsequently we have enhanced the Behaviour Composer so
that an ‘end user’ could build something like the EGM without touching the
source code of the Behaviour Composer.
To create a different game maker (or a
model maker) one starts by creating the base game or model in the Behaviour
Composer. From the ‘share’ tab one can obtain a URL that will load the model.
One then authors and hosts an ordinary HTML web page. The model URL needs to
be enhanced to include ‘&tab=…’ where the URL to the web page is added. To
add check boxes the page should contain the following text for each check box:
Begin
Replay Session Events Check Box:
ID: a
name for the check box
Label: the
label associated with the check box
Session
ID: a session ID described below
Do
message: a message displayed when checked
Undo
message: a message displayed when unchecked
Title: the
title displayed when the mouse hovers over the check box
End
Replay Session Events Check Box
Figure 5 – Mark up code
needed to generate a session replay check box
When this page is loaded into the Behaviour
Composer all this text is removed and replaced by a check box. The session ID
of a check box is obtained by loading the base model into the Behaviour
Composer and then making changes to the model. The system provides a session ID
that can be used to recreate the changes performed. By copying and pasting the
ID into the check box form the check box when ticked will replay all the
changes and unticking it will remove the changes. The rest of the web page can
provide background and instructions.
Conclusions
Although our experience with the EGM lasted
for just some minutes with each participant we think that the game, together
with the Behaviour Composer environment, open many possibilities to teachers
and students to explore models and simulations in educational settings
integrating different disciplines such as biology, mathematics and social
science.
The Logo family of languages, including
NetLogo and Scratch (Resnick et. al., 2009), aspire to have a low threshold so
users can begin to build without a large upfront investment to acquire the
prerequisite knowledge. However, in the context of an exhibit the threshold
needs be a few seconds of instruction. For turtle programming there have been
near-zero threshold implementations called ‘Instant Logo’ or ‘single-key Logo’
since the mid-1970s (Goldenberg, 1974; Solomon & Papert, 1975). In the same
period Radia Perlman pioneered special hardware interfaces for pre-school
children to program turtles (Morgado et. al., 2006).
Agent-based modelling is significantly more
complex than turtle programming. NetLogo, for example, extends turtle
programming with agents, agent sets, links, patches, and much more. The author
of a game maker such the EGM will need much of this richness. The challenge is how
to (temporarily) hide this complexity from users. The approach presented here
is to provide a simple or partial model and to provide a user interface that
automates the addition of several major model enhancements. Unlike ‘Instant
Logo’ we are not providing the user with generic primitives upon which to build
but instead high-level domain-specific components. We are attempting to support
‘middle-up’ programming rather than ‘bottom up’. The low level code (i.e.,
NetLogo) is available and accessible. Users are encouraged, but not required,
to deal with it.
Program and source code availability
The EGM is freely available from http://m.modelling4all.org/p/en/sse.html.
The Behaviour Composer is available from http://m.modelling4all.org.
The source code is available at http://code.google.com/p/modelling4all/source/checkout.
References
Goldenberg, P. (1974) FASTR – A Simple
Turtle Runner. Logo Working Paper No. 30, MIT AI Lab, Massachusetts Institute
of Technology, Cambridge, MA, USA.
Morgado, L., Cruz, M., & Kahn, K.
(2006). Radia Perlman – A pioneer of young children
computer programming. Current Developments in Technology-Assisted Education:
1903–1908.
Kahn, K. & Noble, H., (2010). The
Modelling4All Project — A web-based modelling tool embedded in Web 2.0. J.
Clayson & I. Kallas (Eds.), Proceedings of the Constructionism 2010
Conference. Paris, France.
Resnick, M., Maloney, J.,
Monroy-Hernandez, A., Rusk, N., Eastmond, E., Brennan, K., Millner, A.,
Rosenbaum, E., Silver, J., Silverman, B., Kafai, Y. (2009). Scratch:
Programming for All. Communications of the ACM, November 2009.
Scherer A. & McLean A., (2002) Mathematical
models of vaccination, British Medical Bulletin 2002;62 187-199.
Solomon, C. & Papert, S., (1975)
Teach: A Step Toward More Interactive Programming. Logo Working Paper No. 43,
MIT AI Lab, Massachusetts Institute of Technology, Cambridge, MA, USA.
Wilensky, U & Rand,W. (in press). An
introduction to agent-based modeling: Modeling natural, social and engineered
complex systems with NetLogo. Cambridge, MA: MIT Press.
Wilensky, U. & S. Papert, S.
(2010). Restructurations: Reformulations of knowledge disciplines through new
representational forms. J. Clayson & I. Kallas (Eds.), Proceedings of
the Constructionism 2010 Conference. Paris, France
Wilensky, U. & Reisman, K. (2006) Thinking
like a wolf, a sheep, or a firefly: Learning biology through constructing and
testing computational theories—an embodied modeling approach. Cognition and
Instruction 24, no. 2 (2006): 171–209.
Wilensky, U. & Resnick, M. (1999). Thinking
in levels: A dynamic systems approach to making sense of the world. Journal
of Science Education and Technology 8, no. 1 (1999): 3–19.
