1. Mind, computers and communities
How psychology may help computer science,
and permit global simulations of societies
Franco Bagnoli
Laboratorio Fisica dei Sistemi Complessi - Dip. Energetica
e CSDC, Universit` di Firenze
a
www.complexworld.net
Franco Bagnoli
Mind, computers and communities
2. Psychohistory
In the year 1951 Asimov published the first volume of the
Foundation cycle.
The protagonist of the whole opera (7 volumes, 500 years) is Hari
Seldon, the inventor of the psychohistory.
It is a “science” that permits to combine history, sociology, and
statistics to make mathematical predictions about the future
behavior of very large groups of people, such as the Galactic Empire.
This forecasting possibility allows Seldon and his later followers to
change the “future”, and avoid/mitigate crisis.
Franco Bagnoli
Mind, computers and communities
3. Is psychohistory possible?
Although Asimov’s novels deals with telepathy and other
non-scientific issues, the idea of global simulations and predictions is
taken seriously.
The European flagship project FuturICT, that might be funded with
a billion euros in 10 years, states:
FuturICT will build a sophisticated simulation, visualization
and participation platform, called the living earth platform
(planetary-scale data collection and simulations). This
platform will power crisis observatories, to detect and
mitigate crises, and participatory platforms, to support the
decision-making of policy-makers, business people and
citizens, and to facilitate a better social, economic and
political participation.
Essentially, a global social empowerment program.
Franco Bagnoli
Mind, computers and communities
4. Psychohistory again
It is interesting to consider the premises of psychohistory (mathematical
sociology).
Psycho-history was the quintessence of sociology, it was the science
of human behavior reduced to mathematical equations...
The laws of history are as absolute as the laws of physics, and if the
probabilities of error are greater, it is only because history does not
deal with as many humans as physics does atoms, so that individual
variations count for more.
Franco Bagnoli
Mind, computers and communities
5. Physics of history
Psychohistory is similar to gas theory:
The population whose behavior was modeled should be sufficiently
large (at least 1010 individuals).
The population should remain in ignorance of the results of the
application of psychohistorical analyses.
It assumes that Human Beings are the only sentient intelligences in
the Galaxy (no mutants, robots, etc.).
Franco Bagnoli
Mind, computers and communities
6. Is it only imagination?
In the ’50, Asimov considered only “equations”, with an “electronic”
device (the Prime Radiant) that helped in visualization and annotation
(say, a computer screen..). Asimov probably considered differential
equations, say, epidemic modeling. However, after the use of computers,
researchers discovered some drawbacks:
Differential equations represent a sort of “mean field”, or moment
expansions. They are not very appropriate for small numbers (as
stated also by Asimov), or very non-linear dynamics. Spatial effects
are even more difficult to be captured by partial differential
equations.
One could resort to agent-based simulations (the current trend): one
agent for each person (an “avatar”). Kind of Matrix.
But even in this case there are conceptual and practical difficulties.
Franco Bagnoli
Mind, computers and communities
7. Difficulties
Noise and chaos: we are not able to simulate an individual with
sufficient detail to make the model deterministic, so we have to
insert stochastic components. In any case, the resulting dynamics
would probably be chaotic.
Detailed individual models require a lot of parameters, and therefore
a lot of measurements (some of which are quite hard to be
performed).
Auto-organization: people behavior is extremely sensitive to
collective effects. Let think to markets: the price of an item is not
due to its value, but to the value attributed by other participants.
Such effects are hard to include, ans in any case, if the simulations
can affect the behavior of the system, it should be included in a
self-consistent way (this the the reason for which Asimov supposed
that the foundation had to remain secret).
Little knowledge of the individual behavior (to be investigated in the
following).
Franco Bagnoli
Mind, computers and communities
8. Good news
There are also some arguments in favor of a success
Universality: many of the particular details of the individual should
not affect the dynamics of a large population. However, microscopic
“constraints” may well reflect in the large population dynamics (for
instance fermionic Pauli exclusion principle determines the small
electron contribution to the specific heat of metals).
Global measurement: electronic sensors and devices, electronic
transactions, internet, etc. allow to perform detailed measurements
on individual behavior, an essential element to validate any model.
Delegation to software. Many of our acts are actually delegated to
software, often embedded in cars, browsers, phones, etc. This is
another point of integration between psychology and computer
science (to be explored below), but makes predictions easier.
Franco Bagnoli
Mind, computers and communities
9. Individual behavior
The starting point of any analysis should be the individual (i.e.,
psychology).
Evolutionary constraints give the main framework, through genetics,
womb gestation and education. The knowledge of these factors is
still limited.
The brain is a complex multi-level object, not easy to be modeled at
a global level. Most of modeling is done statistically using linear
models (for instance, factorial analysis).
Learning makes individual “state machines”, so that repeating
measurements is almost impossible. Learning is generally absent in
statistical modeling.
Franco Bagnoli
Mind, computers and communities
10. Evolutionary constraints
Our brain was not selected to cope with algebra, path integrals or TV
recording programs. Some recent investigation suggests:
Probability reasoning, in terms of natural frequencies (one case over
ten, not 0.1).
Social tasks (see Wason selection task), not abstract reasoning.
Speech-oriented interactions (mainly recreational).
Mating behavior (indeed the strongest selection factor).
Social conformism (fashion, religion, politics) vs individualism, a
with a “bonus” for small innovations (fashion innovation).
Kin, mutual and reputation cooperation, leading to group selection
(strong cooperation inside a group, racism against other groups.
Cross-over (synesthesia) among tasks. This could be the origin of
innovation.
Franco Bagnoli
Mind, computers and communities
11. Mechanisms
Human mind is not rational (unless forced).
As revealed by response times, we can classify the level of mental
involvement as reflexes, mental scheme, heuristics, meta-heuristics...
Heuristics (say: fast and frugal, anchoring, etc.) allows quick
responses with bounded resources in uncertain contexts. We can
learn a lot from them, and apply the outcome to ICT. Clearly, they
sometimes fail spectacularly... That’s so human...
The implementation of human heuristic is particularly appreciated in
devices that has to interact with humans (or be delegated..)
Franco Bagnoli
Mind, computers and communities
12. Failures
Epidemic spreading is one of the biggest failures of psychohistory..
In classical epidemic modeling, there is a threshold related to the
infectivity of the disease and the average number of contacts.
Most of human social networks are scale-free, with a diverging
connectivity.
So, for any infectivity rate, a disease should become a pandemy (like
bubonic pest in medieval times).
But actually (modern) people react to epidemics by changing the
social network so to prevent pandemies. Risk perception acts as a
main factor..
Franco Bagnoli
Mind, computers and communities
13. Delegation
People like to delegate boring tasks to machines (computers,
devices).
So we can expect more and more delegation to cars, domotic,, web
searches and mainly to portable devices (e.g. smarthphones, audio
devices). Let’s think to music for instance: people like to have a
portable radio station that select the “right” music.
“Right” of course depends on the individual, past and recent
experiences, mood, status, available clips (copyright), etc.
This is prototypic of future home use of ICT. Trading is another
crucial playground for human heuristics (it depends on other human
reactions, even if economic schools try to make it inhuman..)
Franco Bagnoli
Mind, computers and communities
14. Human social scales
As said, most of human-generated networks exhibit a certain degree
of scale independence, with long tails.
However, there are a certain number of “magical” numbers in
human activities: the size of a chatting group (around four, as most
of card games), the size of a small group (from a chatting group to
around twelve, as apostols – thirteen is already too many), the size
of a community (about 150, Dunbar’s number, the presumable size
of a good-knowledge memory).
To my knowledge, little is known about the actual dynamics of such
groups (and how they could be modeled starting from cognitive
assumptions).
Franco Bagnoli
Mind, computers and communities
15. Small group dynamics
In particular, small group dynamics is interesting:
Temporally quite complex, while community-size dynamics is
statistically much better determined.
A small group has only a short time horizon.
It is however the “drive” of a community, together with chatting
groups and couple communication.
Interactions are mainly non-informative and somewhat ritual.
Franco Bagnoli
Mind, computers and communities
16. The RECOGNITION project
It is part of the AWARENESS Fet proactive initiative. It aims at porting
the knowledge about humans (psychology) to the ICT domain and
implementing awareness at the device.
Implementation of human heuristics in ICT world, in particular the
“fast and frugal” protocol with bounded rationality.
Developing and implementing the concept of data centric approach
(data should not be separated by their elaboration structures).
Implementation of elaboration “at the device” (saY: cars,
smartphones), not relying on central servers.
Franco Bagnoli
Mind, computers and communities
17. Web measurement
In order to quantitative study human dynamics we need a lot of data.
One of the sources is the web sphere
People reveal a lot of personal data to Internet (say, facebook).
we can have quite a good timing measurements (mental activation).
Moreover, data are already in digital form.
However, a compromize is needed between non-verbal content and
environental control (think to second life vs. textual chat).
A good semantic analyzer would be useful (but quite hard to be
developed): most of effective data analysis is done by humans (see
Google page rank mechanism).
Franco Bagnoli
Mind, computers and communities
18. Implementation
It is not easy to implement such a approach. We are working on the
following topics:
Role of risk perception in epidemics.
Opinion dynamics, role of affinity, peace makers and anticonformism.
Opinion anticipation, origin of personality factors, extension to
nonlinear predictors [De Gustibus]
From attractor dynamcs to feed forward networks and the origin of
synsthesia.
Heuristics for community detection [RECOGNITION].
Personalized audio suggestions - wikiradio [RECOGNITION].
Franco Bagnoli
Mind, computers and communities
19. Conclusions
Psychology (from cognitive science to sociology) could become a
central (quantitative) discipline for ICT.
The field of human-inspired computing includes also physics of
complex systems, biology, evolutionary theory, computer science,
linguistics and is related to all humanities (they are product of the
human mind..)
Unfortunately, no school offers such a study program...
Franco Bagnoli
Mind, computers and communities