by Dirk Helbing
We have seen that self-organizing systems can be very
effective and efficient, but their macro-level behavior crucially depends on
the interaction rules, interaction strength, and institutional settings. To get
things right, it's important to understand the factors that drive the dynamics
of the system.
In physics, many phenomena can be understood by means of forces,
and it takes suitable measurements to reveal them. I show that in
socio-economic systems, too, success and failure depend on hidden forces, which
are not directly accessible to our senses. But now, the data about our world increasingly
allow us to measure the forces underlying socio-economic change, empowering us
to take better decision and more effective actions in the future.
We continue FuturICT’s essays and discussion on Big Data, the ongoing Digital Revolution and the emergent Participatory Market Society written since 2008 in response to the financial and other crises. If we want to master the challenges, we must analyze the underlying problems and change the way we manage our technosocio- economic systems. Last week we discussed: HOW SOCIETY WORKS:Social order by self-organization
Societies around the world are suffering
from financial crises, crime, conflicts, wars, and revolutions. These "societal
diseases" do not occur by chance, but for a reason. The fact that they are
happening time and again proves that there are hidden regularities behind them,
but that we haven't understood them well. This is why we fail to master them. But
in the future, I argue, we will be able to understand societal diseases and
cure them. One might imagine this to work in a similar way as the x-ray
discovered by Wilhelm Conrad Röntgen (1845-1923), which has helped to cure
diseases of billions of people by revealing where something is broken in our
bodies, or wrong.
In fact, the growing amounts of data about
our world will allow us to develop entirely new measurement methods to see
hidden patterns in the activities of our global techno-socio-economic-environmental
system, very much like microscopes and telescopes in the past have enabled us
to discover and understand the micro- and macro-cosmos around us. As we have
built elementary particle accelerators to discover the forces that keep our
world together, we can now create Socioscopes to reveal the principles that
make our society succeed or fail.
Given that the loss of control over a
system often results from a lack of knowledge about the rules governing it, it
is important that we learn to measure these hidden forces, which govern our
economy and societies. This will also put us into a position to use these
forces such that we can overcome systemic instabilities and crises. To achieve
this, we don't have to save all the data in the world in one database. It's
much better to perform tailored measurements as needed for the respective
question or purpose. But how can we proceed?
Measuring the world 2.0
Some of
the greatest discoveries in human history were made by measuring the world. We
have discovered new cultures and new continents. Reaching out for the skies, we
have explored our universe and discovered black holes, dark matter, and
entirely new worlds. Now, the Internet is offering novel opportunities to
quantify our world. Performing a sentiment analysis of blogs, Facebook posts, or tweets, we can
visualize emotions of people such as happiness (see, for example, We Feel Fine, Hedonometer, and Sentistrength.
We can also get a picture of the social, economic, and political "climate,"
by determining the subjects that people publicly communicate about see the GDELT project . Mining data on the
Web, one can further determine the gross domestic product per capita,[1] violence, and crime[2], as a function of location and
time, Se Gapminder . One can even reconstruct the three-dimensional world around us from the photos
that people upload on platforms such as Flicker all the time, see Video .
Furthermore, we can create Financial Crisis
Observatories to detect the likelihood of financial bubbles and crashes See . We can build
RiskMaps and CrisisMaps, to support first aid teams in a disaster-struck area See . And we
can map innovations, the spreading of knowledge, and scientific concepts, as I
have started to do it with Amin Mazloumian, Katy Börner, Tobias Kuhn, Christian
Schulz and others (see pictures below and the Living Science app . The
spreading of culture in the world over the centuries, as studied by Maximilian
Schich together with Laszlo Barabasi, myself and others, is particularly
fascinating (Video 1 , Video 2 , Video 3 .
Inspired by Wikipedia and OpenStreetMap, we could also create an
OpenResourcesMap visualizing the resources of the world, and who uses them, to
help reduce undesirable shortages. And we could produce an OpenEcosystemsMap to
depict environmental change and who causes it.
Very soon, we will have not only maps
accumulating past activities, but also systems delivering real-time answers.
When asking a question, this will trigger tailored measurements telling us, for
example, what is the traffic situation in London's Oxford street? What's the
weather in Moscow, and how might this affect investor decisions and consumer
choices? How happy are people in London today, and how much money will they
spend shopping? Why did people in Switzerland vote against "mass
migration"? What worries people in Paris at the moment? How many people
are up between 3 and 4am on Sunday nights around Manhattan's Central Square,
and is it worth selling pizza at that time? How noisy is it to live in the
quarter of town I am considering to move to? What's the rate of flu (or Ebola)
infections in the place I want to spend my holidays in? Where are the road
bumps in my city located? And when did we have the last earthquake greater than
4 on the Richter scale within a range of 500 kilometers?
Information like this can create an
increased awareness of the world around us, and empower us to take better
decisions and more effective actions. But how will we get all this real-time
information?
Creating a Planetary Nervous System as a Citizen Web
The sensor network underlying the emerging "Internet
of Things" will enable us to perform real-time measurements of almost
everything. It will be possible to get all the information needed to establish real-time
feedbacks in complex dynamical systems. This will allow us to support a
favorable self-organization – not just of traffic lights and production, but
basically of everything that requires proper real-time adaptation.
We can use the "Internet of
Things" also to build an intelligent information platform called the
"Planetary Nervous System" (PNS). This was proposed by the FuturICT
project . The Planetary Nervous System
would have three main functions: First, to configure sensor networks in order
to answer specific queries and provide real-time measurements; second, to
create awareness of problems and opportunities around us; and third, to measure
the hidden forces underlying socio-economic change and important intangible
factors such as trust, reputation, public security or other quantities
depending on interactions in social networks.
In fact, my team has already started to
develop such an information platform. We will build the Planetary Nervous
System as a Citizen Web together with the emerging nervousnet community committed to the further development of the
project. This approach will give the citizens control over their personal data,
in accordance with their right of informational self-determination, and create new
opportunities for everyone. Nervousnet
will be engaged in protecting privacy while offering everyone the possibility
to contribute to the measurement of our environment. In some sense, the project
may be envisaged to be a real-time Wikipedia of our world. Like OpenStreetMap
it will build on data provided by many volunteers. A large share of this data will
be Open Data, and you will be able to use it to develop your own business. So,
why don't you join the nervousnet
community today?[3]
The Planetary Nervous System will soon be a
large-scale distributed information platform providing real-time social mining
services as a public good. While existing Big Data systems threaten social
cohesion as they are designed to be closed, proprietary, privacy-intrusive and
discriminatory, we will rather create an open, privacy-preserving and
participatory platform designed to be collectively built by citizens and for
citizens. This will establish a novel social mining paradigm: Users are
provided with freedom and incentives to share, collect and, at the same time,
protect data of their digital environment in real-time. In this way, social
mining turns into a knowledge extraction service for the public good. In
perspective, it will provide a public information service for everyone, and
perhaps become a foundational public institution for the emerging digital
societies of the 21st century. But it takes more than data to understand the
world and its problems...
Sociophysics: Revealing the hidden forces governing our
society
As previously discussed, analyzing and
visualizing data should be only the first step. In the chapter on the Crystal
Ball, I have pointed out that data mining alone usually doesn't deliver a good
understanding of a complex dynamical system. In order to make sense of data,
it's important to have explanatory models, which allow one to make predictions
for situations that haven't occurred before.
In the last chapter, we have discussed
interaction rules ("social mechanisms") that influence human behavior
in a similar way as the gravitational force determines the motion of the
planets – there is just more diversity and randomness. Social roles, i.e.
behaviors that social norms expect from us, are further examples illustrating
the existence of such rules. While the scientific approach of "agent-based
simulations" specifies these rules by computer codes, the research field
of Sociophysics expresses them by mathematical formulas. In this chapter, I
will discuss the powerful concept of "social forces," which allows
one to construct a link between the micro-level interactions between
individuals and the often unexpected macro-level outcomes in socio-economic
systems resulting from them.
The concept of forces is one of the main
pillars of physics. In order to discover it, one had to replace the worldview assuming
the Earth to be the center of the universe by a worldview with the planets
moving around the sun. This new interpretation of planetary motion data allowed
Isaac Newton (1642-1727) to formulate a simple and plausible model based on the
concept of gravitational forces. By now, most parts of physics are formulated
in terms of forces and the way they influence the world. The predictive power of
the respective models is striking. It has been impressively demonstrated by the
moon shot of the Apollo program and many other examples.
A further aspect that made physics so
successful is its long tradition in building instruments to measure things that
are otherwise not accessible to our senses. This reaches from the early stages
of our universe to the exploration of elementary particles up to the study of
processes in biological cells. Therefore, we should ask ourselves how to build "Socioscopes"
that can reveal the hidden forces behind the self-organization of socio-economic
systems. In this way, we will eventually learn to understand the
counter-intuitive behaviors of complex systems. I believe we will soon be able
to diagnose emergent "diseases of society" such as financial crashes,
crime, or wars, before they happen. This would enable us to avoid or mitigate
these problems in a similar way as instruments for medical diagnosis have
helped us to cure diseases. Isn't that an exciting perspective?
Social forces between pedestrians
To demonstrate the feasibility of this
approach, let me first give an example for the usefulness of force models in
the social sciences relating to pedestrian and crowd dynamics. Starting in 1990,
when I wrote my diploma thesis, I noticed that pedestrian paths around
obstacles looked similar to streamlines in fluids. So, I decided to formulate a
fluid-dynamic theory of pedestrian flows, and I derived it from an
individual-based pedestrian model, which was inspired by Newton's force model.
This "social force model" assumes
that the acceleration, deceleration and directional changes of pedestrians can
be approximated by a sum of different forces, each capturing a different desire
or interaction effect. For example, the adaptation of the actual pedestrian
velocity to the desired speed and preferred direction of motion of a pedestrian
can be modeled by a simple "driving force," describing a gradual
adaptation of the velocity within a typical time period. Moreover, the desire
to avoid collisions and to respect a certain "territory" around
others is reflected by repulsive interaction forces between pedestrians with a
strength that exponentially decays with distance. Repulsive interactions with
walls or streets can be captured by similar forces. The attraction toward
tourist sites can be described by attractive forces, and the reason for family
members to stay together as well. Finally, a random force may reflect behavioral
variability.
It is exciting that computer simulations of
this model match many empirically observed phenomena surprisingly well despite
its simplicity. For example, it is possible to understand the emergence of
river-like flow patterns through a standing crowd of people, the wave-like
progression of individuals waiting in queues, or the lower densities on dance
floors as compared to the people standing around.
Self-organization of lanes of uniform walking direction
There are also various self-organization
phenomena that lead to fascinating collective patterns of motion. For example,
when pedestrians are entering a corridor on two sides, we observe the formation
of lanes of uniform walking direction, Video.
That is, the opposite flows are automatically coordinated in a way that
produces an efficient separation of counter-flows. One might see the Invisible
Hand at work, here. But we can actually explain how social order is created and
how a collectively desirable outcome results from local interactions. Whenever
an encounter between two pedestrians occurs, the repulsive interaction force
between them pushes the pedestrians a bit to the side. Importantly, these
interactions are more frequent between opposite directions of motion, due to
the higher relative velocity. This is the main reason that causes opposite
directions of motion to separate from each other. A preference of pedestrians
to walk on, say, the right-hand side is not needed to explain the phenomenon.
From the point of view of complexity science, lane formation is a
"symmetry breaking" phenomenon that occurs when a mixture of
different directions of motion is unstable.
Walking through a "wall" of people without
stopping
Surprisingly, the very same force model
also reproduces a number of other interesting findings in pedestrian crowds,
such as oscillatory changes of the flow direction at bottlenecks. This results
from an alternating pressure relief in the crowd and has inspired the
self-organized traffic light control discussed in a previous chapter. Another
example of self-organization is the amazing phenomenon of "stripe
formation," which allows pedestrians to cross a pedestrian flow without
having to stop (see illustrations below). It's almost as if one could walk
through a wall! Using the social force model, it's possible to understand how
this is possible. The formation of stripes – which occur for similar reasons as
the lanes discussed before – allows pedestrians to move forward with the
stripes and sideward within the stripes. Taken together, this enables the
continuous collective motion through a crossing flow, Video.
Measuring forces
In physics, forces are experimentally
determined by measuring the trajectories of particles, especially changes in
their speeds and directions of motion. It would be natural to do this for
pedestrians, too. At the time when we developed the social force model for
pedestrians, I could not imagine it would ever be possible to measure social
forces experimentally. But a few years later, we actually managed to do this. Around
2006, with the advent of powerful video cameras and video processing, my former
PhD student Anders Johansson was able to extract pedestrian trajectories. We
furthermore adapted the parameters of the social force model to optimally
reproduce the trajectories with computer simulations of the model. In 2006/07,
such tracking methods became also important for the analysis and avoidance of
crowd disasters.
Then, in 2008, Mehdi Moussaid and Guy
Theraulaz set up a pedestrian experiment in Toulouse, France, under well-controlled
lab conditions. This finally empowered us to do data-driven modeling. While
before we had to make assumptions on the functional form of pedestrian
interactions, it now became possible to determine the functional dependencies
directly from the wealth of tracking data generated by our pedestrian
experiment. After fitting the social force model to individual pedestrian data,
it was finally used to simulate flows of many pedestrians. To our excitement,
the computer simulations yielded a surprisingly accurate prediction of the
pedestrian flows observed in a wide pedestrian walkway.
So, pedestrian modeling can be considered a
great success of Sociophysics. Over time, pedestrian studies had evolved from a
social to a natural science, bringing theoretical, computational, experimental
and data-driven approaches together. This even led to practical and partly
surprising lessons for the better design of pedestrian facilities and the planning
of safer crowd events such as the annual pilgrimage in Mecca.
Most pedestrian facilities are inefficiently engineered
Back in 1994/95, when comparing different
designs of pedestrian facilities, Peter Molnar and I surprisingly found that
obstacles in the way, if properly placed, can make pedestrian counter-flows
more efficient (see figure below). In fact, all the conventional design
elements of pedestrian facilities – corridors, bottlenecks, and intersections –
turn out to be not well designed. They can be substantially improved! In many
cases, "less is more," i.e. providing less space for pedestrians
produces better flows. This surprising discovery can be best understood for
bottlenecks such as doors. Here, a funnel-shaped design can reduce disturbances
in the pedestrian flow, which result when the directions of motion are not well
aligned (e.g. when some people approach the door from the front and others from
the side).
In busy bi-directional pedestrian flows, the
efficiency of motion can be improved by a series of pillars in the middle. It
turns out that these pillars help to stabilize the interface between the
opposite flows, thereby reducing disturbances. The effectiveness of the design
becomes particularly obvious in subway tunnels, where pedestrians move both
ways and pillars exist for static reasons.
Finally, an obstacle in the middle of a
pedestrian intersection may also improve the flow. When Peter Molnar and I
discovered this, it took us a long time to understand. But eventually we
noticed that, at intersections, many different collective patterns of motion can
emerge, for example, clockwise or counter-clockwise rotary flows, or
oscillatory patterns of the crossing flows. The problem is that the different
collective patterns of motion destroy each other after a short time, such that
none of them is stable. Putting a column in the center can increase the
likelihood of rotary flows and thereby increase the overall efficiency. But a
further improvement can be reached by replacing an intersection of four flow
directions by four intersections of two flow directions each, as it may be
reached by suitably located railings. In this way, a rotary flow pattern is
supported, and disturbances can be drastically reduced.
Crowd disasters
Unfortunately, pedestrian flows don't
always self-organize in an efficient way. Sometimes, terrible crowd disasters
happen, and dozens or hundreds of people may die, even though everyone has
peaceful intentions and does not behave in a ruthless or otherwise improper way.
How is this possible?
When I got interested in the problem in
1999, crowd disasters were often treated as God-given or natural disasters that
are beyond human control. But the root cause for the breakdown of social order
in pedestrian crowds has something in common with the occurrence of traffic
jams. If the density gets too high, pedestrian flows turn unstable. The resulting
crowd dynamics can be uncontrollable for individual people, and even for
hundreds of security forces. But I will show below that crowd disasters can nevertheless be
avoided, when their reasons are understood and when proper preparations are
undertaken.
Crowd disasters have happened since at
least Roman times. That’s why building codes were developed for stadiums, as
exemplified by the Coliseum in Rome. The Coliseum had 76 numbered entrances and
could accommodate between 50,000 and 73,000 visitors, who would exit through
the same gate through which they had entered. With these rules and its generous
provision of exits, the Coliseum could be evacuated within 5 minutes. Modern
stadiums, which generally have a smaller number of exits, can rarely match this
figure.
Despite the frequent and tragic occurrences
of crowd disasters in the past, they still continue to happen. In other words,
they are not properly understood. Media reports often suggest that crowd
disasters occur when a crowd panics, causing a stampede in which people are
crushed or trampled. The implication is that crowd disasters would be the
result of unreasonable or aggressive behavior, with some individuals pushing
others relentlessly as they try to escape. But why would people panic?
Referring to crowd disasters, Keith Still once said to me: "People don't
die because they panic, they panic because they die."
In fact, in my studies with Illes Farkas,
Tamas Vicsek, Anders Johansson, Wenjian Yu and others, we revealed that many
crowd disasters have physical rather than psychological causes. They may occur
even if everybody behaves reasonably and tries not to harm anyone else.
Therefore, the view that crowd disasters are mostly a result of panic is
outdated. Alternatively, one might suspect that people are crushed when the inflow into
a spatially constrained area exceeds the outflow for an extended period of time.
Certainly, the density can become life-threatening under such conditions, as
more and more people accumulate in too little space. But when in 2006 another
crowd disaster happened during the Hajj – the annual Muslim pilgrimage around
Mecca – it occurred on a large plaza.
Being
experts in crowd dynamics, Anders Johansson and I were asked to evaluate videos
showing the accident area. In the beginning, when we played the videos, we saw
nothing informative. Due to the high pedestrian density, people just moved very
slowly, some centimeters per second. However, when I asked Anders to play the
videos 10 or even 100 times faster, we were totally surprised!
The
accelerated videos showed some striking phenomena. First we discovered an
unexpected, sudden transition from smooth pedestrian flows to stop-and-go flows
(see the long-term photograph above and the Video.
In contrast to freeway traffic, however, these stop-and-go flows were
previously unknown and unlikely to result from delayed adaptation. We discovered
that they were caused by a competition of too many pedestrians for too few gaps
in the crowd, i.e. by a coordination problem. The stop-and-go movement set in.
When the overall flow suddenly dropped as a critical pedestrian density was
crossed. As a consequence of the drop, the outflow from the area was
drastically reduced, while the inflow stayed the same. So, the density
increased quickly, but this was not the final cause of the tragedy!
To our further surprise, some minutes later
we witnessed another unexpected transition – from stop-and-go flows to a
phenomenon we call "crowd turbulence" (see picture above and the
Video.
In this situation, people were pushed around in random ways. Eventually, Anders
Johansson and I discovered in 2006/07 that it was not the density, but the
density times the variability of speeds, the so-called "crowd
pressure," which determined the onset and location of the crowd disaster.
It turned out that, when the density
crosses a critical threshold, any body movements – even unintentional ones –
can create forces acting on another pedestrian body. These forces can add up
from one body to the next, such that the resulting force quickly changes the
strength and direction. Therefore, people find themselves pushed around in unpredictable
and uncontrollable ways.
It is just a matter of time until someone
loses the balance, stumbles, and falls to the ground. This produces a
"hole" in the crowd, such that the forces acting on the surrounding
people get unbalanced, as the counter-force from the front is missing.
Therefore, the surrounding people tend to fall on top of previously fallen
persons or are forced to step on them. The situation ends with many people
piled up on top of each other, such that those on the ground have difficulties
to breathe and die of suffocation. Similar observations were made in other
crowd disasters, for example, the Love Parade disaster in Duisburg, Germany.
Countering crowd disasters
Can we use the above knowledge to avoid
crowd disasters in the future? Yes, indeed! Some years back, together with
several colleagues, I became involved in a project aiming to improve pedestrian
flows during the annual Muslim pilgrimage to Mecca. We were asked to find a
better way of organizing the crowd movements around the New Jamarat Bridge, a
focal point of the pilgrim route See more. On and around the
previous Jamarat Bridge, thousands of pilgrims had died in tragic crowd
disasters over the years. How could one avoid them?
This
was a challenge that required us to take into account not just technical
matters such as crowd densities, but also dozens of religious, political,
historical, cultural, financial, and ethical factors. Our use of crowd modeling
led us to propose measures including the counting and monitoring of crowds
through newly developed video analysis tools,[4]
the implementation of time schedules for pilgrim groups, re-routing strategies
for crowded situations, contingency plans for possible incidents, an awareness
program to inform pilgrims in advance about the procedures during the Hajj, and
an improved information system that had to guide millions of pilgrims speaking
about 200 different languages. After implementing these proposals, the 2007
Hajj (in 1427H) was indeed safely performed.[5]
Like for the traffic assistant system discussed before, the main underlying
success principle was to gather real-time information and respond to it
adaptively.
Since
then, the principle of providing real-time feedback has widely spread. An interesting example for this
is crowd sensing. Paul Lukowicz, a member of the FuturICT project, and a number
of further scientists such as Martin Wirz and Ulf Blanke, recently developed an
app for safer mass events. In a
number of festivals in London, Vienna and Zurich, they used voluntarily
provided GPS traces of visitors to determine the crowd densities and pedestrian
flows, i.e. averaged quantities determined from the GPS data of many people.
These data were then provided to the visitors of the mass events, to give them
a better idea of over-crowded areas that they should better avoid.
Forces describing opinion formation and other behaviors
Of course, one might ask whether the
concept of social forces can be also used to understand different social
phenomena and to overcome other kinds of problems, too, such as crime or
conflict. I am convinced of this! The success of force models in describing
pedestrian flows is related to the fact that pedestrians are moving continuously
in space. Therefore, the dynamics of a pedestrian can be represented by an
equation of motion, which says that the change of its spatial position with
time is given by the velocity. Complementary to this, the change of velocity
with time, i.e. the acceleration, is modeled by a sum of forces. But can we
understand opinion formation processes or other behavioral changes by social
forces as well? Surprisingly, the answer is "yes," if we have more or
less gradual changes on a continuous opinion scale or in a continuous
behavioral space. Otherwise, one must use generalized models, which exist as
well.
After formulating the social force concept
for pedestrians in Göttingen, Germany, in 1990 I joined the team of Professor
Wolfgang Weidlich at the University of Stuttgart. He was probably the only
physicists at this time working on modeling socio-economic processes and
systems. In some sense, Professor Weidlich might be seen as grandfather of Sociophysics.
My plan at this time was to learn, how opinion formation and decision making could be
modeled. Since my work on pedestrians, I had the idea that both, the individual
and collective behavior of people could be understood through social forces, and
I managed to formulate a corresponding theory (see Information Box 1).
Interestingly, social force models can be
formulated for migration processes, too, when people are assumed to relocate
within a certain (not too large) radius. In one of my models formulated in
2009, I assume success-driven migration, where individuals try to avoid locations,
in which they expect bad outcomes, but seek locations that appear to be
favorable. Bad neighborhoods (those, where people were uncooperative) turn out
to have a repulsive effect, while good neighborhoods (where people were
cooperative) have an attractive effect. It is even possible to calculate the
direction and strength of the repulsion or attraction effect, i.e. the force
describing the average direction and speed of motion in a certain location.
A great advantage of using the concept of
"social forces" is that it can help us to get a better imagination of
complex processes underlying social change. Movements towards some subject or
object are reflected by attractive forces, movements away by repulsive forces.
It is also important to recognize that such forces may not be attributed to individuals,
but rather to groups of individuals, companies, or institutions. In other
words, social forces may be a collective effect. Group dynamics, or "group
think" as a result of some emergent group identity, is probably a good
example for this. There, the interaction of individuals creates a collective
"group" perspective, which in turn changes the behavior of
individuals. In fact, the theory of social milieus knows that the behavior of an
individual is largely influenced by its environment. This can now increasingly
be quantified and put into mathematical formulas with predictive power. But
what is more powerful: physical or social forces?
Culture: More persistent than steel
It has often been claimed that war is the
mother of civilization, and whoever has better weapons will rule the world. However,
I don’t buy this. Even though war may have spread civilization, I believe the
underlying mechanism is migration and the exchange of goods and ideas. Today,
with the Internet, civilization can spread in ways that does not have to cost
human lives.
But what is the basis of civilizations?
It’s culture! Culture is largely a collection of rule sets, such as procedures
and social conventions, norms, values and roles. These determine the success
and failure of societies and guide their evolution over sometimes thousands of
years. Just take religious values, which can determine the behaviors of people
over thousands of years. It is therefore not exaggerated to say that culture is
more persistent than steel. And culture is more relevant for success than
weapons. In other words, social forces can be stronger than physical ones. A
good example is the ancient Greek culture, which managed to spread to their
Roman occupants, since it was more advanced.
However, while we all learned about
physical forces at school, only very few people have an explicit knowledge of the
social forces determining the behavior of socio-economic systems. This has to
change, if we want to overcome or at least mitigate socio-economic problems. As
the last chapter has shown, it's now possible to identify the interaction
mechanisms that promote social order. Information Box 2 on social capital further
illustrates that the success and failure of societies largely depends on
invisible factors. In a sense, social norms are the fabric of our society, and
social capital acts like a catalyst of socio-economic success.
Avoiding conflicts
Conflicts, wars and revolutions, too, can
be understood with a social force approach. They relate to forces that destabilize
a system and may break it into pieces. There are at least three types of
conflict situations: (1) An encounter (say, between two countries) causes
losses on both sides. This might be avoided by better advance awareness of the
likely outcomes of such an encounter. (2) The encounter is beneficial for one
side and unfavorable for the other, while it causes an overall damage. Here,
the second party needs to be protected from exploitation (e.g. by solidarity from
third parties or by establishing an efficient separation of the conflict-prone
parties). (3) The encounter is advantageous for one side and undesirable for
the other, but this time the overall outcome is positive. Then, value exchange
can make the interaction beneficial for both sides, i.e. it's possible to align
the interests and create a win-win situation. Recently, I have proposed Social
Information Technologies that aim to reduce the occurrence of conflicts between
companies or people.
Would it also be possible to measure the
forces creating conflict? I think so. We could build a ConflictMap, revealing
regional and international tensions and how they come about. In fact, when
working in my team, Thomas Chadefaux mined millions of news articles over a
period of more than 100 years and performed a sentiment analysis for words
indicating conflict. This allowed him to quantify the level of tension between
countries in the world. Moreover, he could show that the level of tension
allows one to predict the likelihood of war outbreaks in the next six or twelve
months. Such advance warning signals can provide valuable time for diplomatic
efforts to reduce political tensions before it's too late for a peaceful
solution. Our analyses also revealed how tension spreads from one country to
the next, as it happened after the war on Iraq, thereby destabilizing the
entire region. Apparently, this has produced fertile ground for the rise of the
Islamic State (IS).
Conflict in the Middle East
Another data-driven study analyzed a
problem that worries the world since many decades, namely, the conflict in the
Middle East. Why haven't we so far been able to stop this conflict? A classical
Big Data approach, even if we knew all the trajectories of all bullets shot,
couldn't really reveal the causal interdependencies underlying the conflict.
Therefore, in a study with Ravi Bhavnani, Dan Miodownik, and Maayan Mor,
Karsten Donnay, we developed instead an empirically grounded agent-based model.
The validation procedure of our model suggests that intercultural distance is
the main driving force of the conflict.
An analysis of the violent events reveals
that they are correlated with each other. There is rather a responsive
dynamics, where each side "pays back" for the previous attacks from
the other side (see Video 1 and Video 2).
For example, Palestinians retaliate violence on the Israeli side and vice
versa. What does this tell us? Basically, both sides punish each other for the
violence they were suffering from before. From a rational choice point of view,
this should stop the chain of violence, as one event triggers another, usually
bigger one, or even several ones. Therefore, the conflict is costly for both
sides, and increasingly so over time. The Israeli movie
"Gatekeepers," which interviews previous chiefs of secret service,
therefore, comes to a remarkable conclusion: "We have won every battle,
but we are losing the war." In other words: it does not pay off to be
violent – on the contrary. It seems that each party tries to send a message to
the other one: "Stop being violent to us – you will otherwise have to pay
a high price!"
So, why does such counter-violence cause
escalation rather than stopping the chain of violence? Because both sides think
they are right in what they do. In fact, they apply the right principles, but
to the wrong situation. It is very important to recognize that the reason that
makes us punish others is related to the way we use to establish social order.
We have learned that punishment is a
mechanism that can establish and stabilize social norms. Therefore, we punish
those who do not follow our norms. However, such punishment is only effective,
if the punished side accepts the punishment. Otherwise it will strike back and
pay revenge, which gives rise to an escalating conflict. It is, therefore,
important to recognize that punishment will only be effective, if people share
the same values, norms, and culture.
Therefore, in a multi-cultural society,
punishment may not be effective in creating social order. Under such
conditions, a possible way to reduce the level of conflict would be to separate
the opposing parties, i.e. to live in different areas. Another one is to
develop a culture of tolerance, understanding and respect. In fact, as we have
seen in the last chapter, there are many social mechanisms that support the
creation of social order, for example, reputation mechanisms. I am, therefore,
confident that the deeper understanding of the mechanisms and forces producing
conflicts will eventually allow us to overcome or mitigate them. Personally, I
would strongly advice to go away from a punitive culture and to engage in a differentiated,
reputation-based culture appreciating diversity. This means, almost all of us
would have to change the way we are treating others. It would require a global
change of culture. But the Internet may (help to) bring it on the way.
Flu prediction, better than Google
Not just wars, but pandemics too are a
major threat to humanity. Some of them have killed millions of people. The
Spanish flu in 1918 was a shocking example of this. Such pandemics are, in
fact, expected to happen again, as viruses mutate all the time, finding our
immune systems more or less (un)prepared. The world has also been surprised by
the recent outbreak of Ebola.
To contain epidemic spreading, the World
Health Organization (WHO) is continuously monitoring emerging diseases. It takes
about 2 weeks to collect the data from all the hospitals of the world, such
that one typically gets an overview of the actual situation with a two weeks
delay. Then, Google Flu Trends invented an approach called
"nowcasting," which was celebrated as major success of Big Data
analytics. It was claimed that it was possible to estimate the number of
infections in real-time, based on the search queries of Google users. The underlying idea was that queries such as "I
have a headache" or "I don't feel well" or "I have a
fever" and so on might be indicative of having the flu. Recently, however,
the Google Flu approach was found to
be unreliable, partly because of Google
steadily changes its search algorithms and also because advertisements may bias
people's behaviors. Fortunately, there is a model-based approach using much
less data, which considers the mechanism of disease spreading. It looks at
infection data in a way that is augmented by a model based on air travel data.
How did Dirk Brockmann and I discover this approach in 2012/13?
Independently of each other, we had been
interested in modeling epidemic spreading processes already for a couple of
years. In 2002, in the wake of the September 11 attacks in 2001, there were
fears of bioterror using anthrax or other deadly germs, which threatened the
USA and the rest of the world. At this time, I was proposing to Otto Schily,
the then German Minister of Internal Affairs, to build a self-calibrating
epidemic simulator to predict the spreading of pandemics. While infection and
recovery rates are often not well known after a disease outbreak, the idea was
that a self-adaptive calibration model would produce increasingly accurate
predictions, as more data of infected people would become available. However, I
received a letter that such an approach wouldn't be feasible. I did not really
believe this, but it delayed progress for an entire decade, because I did not
have any funding for such a study.
Dirk Brockmann, however, started to
investigate the spatio-temporal spread of diseases by means of computer
simulations, and also by analyzing the paths of dollar bills in his famous
"Where is George?" study . However,
when visualizing spatio-temporal spreading patterns of epidemics, the patterns
looked frustratingly chaotic and unpredictable. The relationship between the
arrival time of a new disease as a function of the distance from its origin
location was so scattered that one could not make much sense of it. But it
became increasingly clear that this problem resulted from the high volumes of
air passenger travels. So, Dirk had the idea to define an effective distance
based on the travel volumes between all airports in the world and to study the
spreading dynamics as a function of this alternative distance measure.
Our collaboration finally emerged in 2011,
when Germany was shocked to see the spreading of the deadly, food-borne EHEC
epidemic. I got in touch with Dirk, because I thought we could combine his
epidemic spreading model with a model of food supply chains and, thereby, help
to identify the origin location of the disease, which was unknown at that time.
Unfortunately, we could not get hold of proper supply chain data at that time.
But our discussion triggered a number of important ideas. Particularly, our
attention moved from predicting the spreading of diseases to detecting their
origin locations.
In fact, looking at the empirical infection
cases in an effective distance representation from the perspective of all
airports in the world, we found that the most circular spreading pattern
identified the most likely origin of the disease. But what is more important:
once the origin location is known, combined with the circular spreading pattern
in the effective distance representation, it becomes possible to predict the
order, in which cities will be hit by a pandemic wave. See Video. Furthermore,
it turns out that this technique can be successfully applied even if the
epidemic spreading parameters such as the infectiousness and recovery rate are
not well-known, which is typical the case after the outbreak of a new disease. The
only data important for our analysis are the air travel volumes between all
airports, which are needed to specify the effective distance.
Shortly later, Ebola broke out and, using
our previously developed method, Dirk made early predictions of Ebola imports
into other countries, which became the basis of international preparations to
contain its spreading. See . I would also like to mention the team of Alessandro Vespignani and Vittoria
Colizza, both partners of the FuturICT initiative, who have built a
sophisticated simulator to predict diseases and test the effectiveness of
political measures. See
INFORMATION BOX 1: Social Fields and Social Forces
When I worked on the social force model, I soon discovered the book by Kurt Levin (1890-1947) on the social field concept. I like his idea a lot, even though a behavioral and theoretical foundation of the concept was missing. So I decided to develop such a foundation in my PhD thesis in 1992. This resulted in the derivation of Boltzmann-like and Boltzmann-Fokker-Planck equations from behavioral assumptions.
These equations containa quantity determining the systematic motion in the behavioral space, which can be interpreted as "social force" and often expressed as the slope of a "social field." Such a social field can be imagined like a mountain chain in behavioral space, where the steepest slope in a location determines the social force that a person with the corresponding behavior would feel. This social force describes the expected size and direction of the behavioral change. Valleys of the social field correspond to social norms. If complying with the norm, the social force is zero. But when deviating from a social norm, one will feel a social force, as it happens in reality.
Note, however, that the above discussed "mountain chain" and, with it, the corresponding social field is variable. It changes depending on the behavioral changes of others. Therefore, the social field influences the behavior of a person, but at the same time, it is potentially modified by that person's behavior and the behaviors of others. In other words, social norms may change over time as a result of social interactions.
INFORMATION BOX 2: Social Capital
Most of us have probably learned that money makes the world go round and all that matters is to have enough of it. Money is certainly a powerful invention, but there is more that contributes to economic development. This includes human capital (like education), but also social capital.
So, what is social capital? I define it as everything that results from social network interactions and can potentially be turned into a benefit. Examples are cooperativeness, public safety, and culture of punctuality, reputation, trust, respect, and power. While our own actions influence our social capital, we can't fully control it, which is in contrast to money. In many cases, we can't buy social capital (or only to a limited extent), but social capital creates value added. Interestingly, by doing certain things, we are not automatically entitled to get a certain amount of social capital. As we know from reputation and respect, these things are given to us by others. They depend on interaction effects.
Note that the amount of social capital also determines the resilience of a system, and its risk of failure. Social capital influences both, the probability and size of damage. This became clear to me in a seminar of ETH Zurich’s Risk Center. The Risk Center brings together experts in probability theory with experts in complexity and network theory. We discussed that large disasters have an over-proportional impact on public opinion. That's why plane crashes and terror attacks matter a lot to people, while they seem to feel less threatened by everyday risks such as car accidents or deaths of smokers. Therefore, it is often believed that "size matters," i.e. large disasters make people respond in an irrational, perhaps even panic way.
However, having studied the phenomenon of panic for some time, I rather concluded it was more likely that "people respond to the fact that there has been more damage than just the physical one – namely damage to the social capital." For example, a large-scale disaster often damages the trust into the risk management of companies or public authorities, in particular when it was caused by unprofessional behavior or corruption. While people care about such things, no insurance company is covering this damage to social capital. Hence, we must protect social capital in a similar way as we protect economic capital or our environment. Social capital can be damaged and exploited, but this should be prevented. In order to get there, we must learn to measure social capital and to quantify its value. Quantifying the value of our environment also helped eventually to protect it better.
Trust and power
To stress the importance of social capital, it is important to recall that the financial crises resulted from a loss of trust: banks did not trust other banks anymore and did not want to lend their money; customers did not trust their banks anymore and emptied their bank accounts; banks did not want to give loans to companies anymore; people did not want to invest in financial derivatives anymore, etc. In the end, the resulting financial meltdown amounted to an estimated 20,000 billion US dollars. So, trust has a pretty high value, and when it gets lost, the economic losses are tremendous. To give another example: the recent loss of trust into US cloud storage companies after the NSA scandal was estimated to cause an economic loss of up to one third of the previous business volume.
Trust is also the basis of power and legitimacy. When I studied in Göttingen in Germany, one day a deadly car accident caused by a mistake of the police triggered a large public outcry and massive demonstrations. This was the first instance, when I noticed that public institutions can easily lose their public support, in other words: their social capital. This happens, if trust gets lost over something that the authorities should not have done according to the moral beliefs of the public. I made the same observation in Zurich, Switzerland, when there were many complaints about the work of the migration office. During this time, the windows of the migration office were broken time and again, but when the office director was replaced, the problem disappeared.
Interestingly, soft factors such as credibility and trust are the basis of power. For example, the loss of control during the England riots in 2011 occurred, after the London police shot a person without giving a sufficient justification to the public. A similar thing happened 2014 in the US city of Ferguson, Missouri. In fact, riots in many other countries, too, were triggered by events where public authorities hadn't done a proper job in the eyes of the people. The Arab spring, for example, started in Tunisia, after Mohamed Bouazizi burnt himself, because of police corruption and bad treatment.
In other words: legitimacy and power result from doing the right thing in the eyes of the people. When people don't offer their idealistic or practical support anymore, authority and power are gone. While one may buy weapons and, with this, destructive power, constructive power depends on the trust and support of the people – otherwise they won't provide support, and this basically means that one hasn’t got any power. Brutality does not create respect. It might create fear, but this can replace legitimacy only up to a certain degree. As the situation gets increasingly unacceptable, more and more people will lose their fear and start to resist the previously respected authorities actively or passively, or even become ready to sacrifice their lives. The problem of such extremism or even terrorism is well-known from freedom fighters, who have previously led normal family lives.
But even passiveness of its citizens can make a country fail within just a few years or decades. This could, for example, be seen in the former German Democratic Republic. I, therefore, believe that trust is the only sustainable basis of power and social order. It must, therefore, concern us all that, in many countries, politics and management are currently the professions with the lowest levels of reputation. In contrast, "social" professions that create public goods – firefighters, scientists, doctors, nurses and teachers – earn the highest levels of reputation. As we will see later, this has some important implications for the future.
[3] You
can get in touch with us at mailto:nervousnet@ethz.ch. For
further information see http://futurict.blogspot.ch/2014/09/creating-making-planetary-nervous.html
[4] For
example, one may play back videos of
security cameras in an accelerated fashion, say, every 60 seconds, which allows
the brain to notice advance warning signals of critical crowd conditions such
as stop-and-go waves. Video post-processing can overlay colors representing the
local density, or arrows representing the average flow in a certain location –
important information that the naked eye cannot see.
[5] and,
in fact, in the following years as well. However, as I moved to another
university, where I focused on new tasks, I haven't been involved in the
changes that have been made since 2007.
