DT&SC 7-10: Why do we need computer simulations?


What are computer simulations and what role
do they play in understanding society better? Big Data, on the one hand, refers to empirical
analysis of society, it’s data about what happened in the past. Something has to happen,
and only then can we record it. And then we have data and it’s always from the past. Sometimes
it’s real time, but as soon as you record it, it already must have passed. So actually,
real time is a funny word it’s actually always from the past because you have to record it.
So, it’s an empirical footprint of what has happened. And computer simulations help us
to theoretically explore to what could happen in the future, in theory. That is very important
because especially social systems are notoriously, non-stationary. Data from the past has problems
of limitations with predicting changing futures. I’ll give you an example here: What some researchers
from Google did in a very good example of the usage of big data is they used google
search tendencies to predict the spread of diseases, for example the flu or Dengue Fever.
The problem is with the spread of diseases is that data comes a couple of months after
the fact because usually this data is collected by people who go to hospitals or by pharmacies
or by surveys or by some kind of health registry and until it gets all together, several weeks
have already passed and we have no idea actually where the disease spreads right now. So, what
the Google researchers said is “let’s just look at what people Google, what they search
on the internet. And with that, let’s see if we can predict when and where the disease
spreads.” Now there are two ways of going about it. One is you go according to theory,s
o let’s look when people search for cough or headache or stomach ache. But they said
no, no. the google researchers say, we don’t want any theory we just go by machine learning
so big data doesn’t need theory and we have big data about what people searches. So they
just ran the 50 million most common search terms and they correlated it with historical
data of the spread of the flu. They ran 450 million mathematical correlation models on
search terms and flu outbreaks and then identified 45 search terms that highly correlated and
were able to predict the outbreak of the flu better than any traditional model. Now, it
turned out that all of these 45 search terms had something to do with the flu. It was terms
like stomachache and headache and whatever, but in theory it might as well been “orange
car” and as long as orange car was a good predictor of flu outbreak out of some inexplicable
reasons, they would have taken it. I mean, there was no theory involved. And that was
a very famous example, I mean people were very excited about it because it helped us
to track the outbreak of flu in real time. As you see hear we could make predictions
where official data was not available yet. Now they ran the same model a couple of years
later and what they found then several years later was is that the model did not work anymore
at all. It couldn’t make predictions anymore as good, what happened? Well, what happened
basically here is that reality changed. So the model was trained on data from a past
from years earlier and then years later the search behavior of people just changed and
the model was not able to make predictions anymore as well. so the best you can do then
is just run after the fact you have to change your predictive model all the time as reality
changes but the best you can do is get as closely as you can to the present and always
adjust your model but you can not really make predictions into a changing future. now most
of what we are interested in in social science is we want to create a better world, a different
world in the future, a world without poverty without hunger, with freedom and democracy
for everybody. Now that world does not exist in empirical data so what we do is we work
with models with theoretical models and we simulate future scenarios that never existed
in empirical reality. The result is a logic that is very similar
to what engineers have been doing for decades when they plan and design buildings that never
exist in empirical realities, if you design a new bridge or a new building. What they
were then they created models of them, and that could allow them to explore scenarios
that were unprecedented they didn’t have an empirical precedent and so in todays architecture
world and engineering world, you have a lot of these models we can explore and plan and
go down even inside the building see what it actually looks like and these are theoretical
scenarios because these buildings don’t exist in empirical reality, they exist on
the in theory in these computer simulations. In the social sciences you can imagine the
result more like, its very similar to playing SimCity, SimCity is a city building video
game thats been around now for over two decades. By the way, that up here, thats what it looked
like when I played it as a kid, and that down here thats what it looks like. Now its just,
really frustrating to me but ( haha ) thats just how it is. And recently they came out
with an application called SimCity edu. So basically they used this video game to help
high-school children to explore what what if scenarios. For example you allow high-school
students to explore the correlation between pollution alternative energies and its effect
on employment. Now if I tell you, explore the correlation between pollution and the
transistion to alternative energy and how that effects employment. You think like well,
that sounds like a PH.D thesis in economics, right? Now these computer simulations can
be used for high-school students to develop a basic intuitions about these correlations,
check it out. For example here we have our city, its a model city, its a city that does
not really exist of course, its a game. But in case it would be a real scientific exercise
we might want to model a real city, so we would take big data or any kind of data and
look for a example here or at the pollution map, so which ones are big polluters in are
city, and employment map, which companies are big employers in our city? And we can
do that for example we identify here the energy sources, thats a code power plant, obviously
a big polluter, and we simply go ahead and, change the course of history and bulldoze
it and while we are at it we take this other core power plant and, there you go you just
bulldoze it as well. Now we have a city that never existed before, city with ought these
core power plants and we install a brand new small wind power plant, right here. And now
we are told that our power in our city is dangerously low, so lets put a large solar
power plant right next to it. So we change the course of history, and we created an unprecedented
scenario where we have these alternative sources of energy and even so we don’t have empirical
data from reality we can now study what happens in our city, and a merit of new things and
unexpected things might happen. The total is more than the sum of its parts, so its
a social emergent phenomena for example, here in front of city hall we can see that some
people, they start to protest. We don’t know why, maybe because of high energy prices,
and traffic patterns might change for example because now the different companies establishing,
different energy structures in the city, and we can see well how do traffic patterns change,
do we get traffic jams and so forth. So we can now produce artificial data that have
been, in theory and compare well, how did we do? How was the pollution scenario? Not
very different? How was the employment scenario? Well, maybe we could do a little bit better,
the basic idea is that we have theoretical tools to explore what might happen in theory,
even so we don’t have empirical data, just aiming at making the world a better place.




Leave a Reply

Your email address will not be published. Required fields are marked *