Few scholastic disciplines have within them an explicit ideal beyond the production of knowledge. With computer science and engineering, the implicit ideal is to ensure better living through circuitry. Personally, my motivation is driven by the sense that social algorithms will lead to a greater human experience.
Collective Decision Making Systems: From the Ideal State to Human Eudaimonia
1. Collective Decision Making Systems:
From the Ideal State to Human Eudaimonia
Marko A. Rodriguez
T-5, Center for Nonlinear Studies
Los Alamos National Laboratory
http://markorodriguez.com
February 13, 2009
2. Collaborators
• Jennifer H. Watkins Collective Decision Making Systems
Los Alamos National Laboratory
International and Applied Technology
Los Alamos National Laboratory
http://public.lanl.gov/jhw
• Alberto Pepe
Center for Embedded Networked
Sensing
University of California at Los Angeles
http://albertopepe.com
http://cdms.lanl.gov
External Advisory Committee Presentation – Los Alamos, New Mexico – February 13, 2009
3. Why do we do the things we do?
• Few scholastic disciplines have within them an explicit ideal beyond the
production of knowledge.
• With computer science and engineering, the implicit ideal is to ensure
better living through circuitry.
• Personally, my motivation is driven by the sense that social algorithms
will lead to a greater human experience.
External Advisory Committee Presentation – Los Alamos, New Mexico – February 13, 2009
4. The Eighteenth Century’s Age of Enlightenment
• Citizen’s began to question the traditional forms of aristocratic and monarchic
governance and began to envision an ideal state.
• The United States was the social experiment to achieve this ideal state.
• Unfortunately, the ideals of these thinkers could not reach their purest forms due to the
limitations of the technology at the time. Moreover, as a people we should value the
ideals, not the implementation of government.1
Marquis de Condorcet Thomas Paine Adam Smith
1
Marko A. Rodriguez and Jennifer H. Watkins, “Revisiting the Age of Enlightenment from a Collective Decision Making
Systems Perspective”, LA-UR-09-00324, February 2009.
External Advisory Committee Presentation – Los Alamos, New Mexico – February 13, 2009
5. Marquis de Condorcet (French: 1743 – 1794)
Social Choice Theory
1.0
0.8
0.6
p
0.4
0.2
0.0
0 10 20 30 40 50 60 70 80 90 100
n
• Suppose a group of n decision makers under a two option, majority rule vote, where each decision
maker has probability p of choosing the optimal option.
• If p > 0.5 and as n → ∞, then the probability of yielding the optimal decision approaches 1.0.
• if p < 0.5 and as n → ∞, then the probability of yielding the optimal decision approaches 0.0.
• Condorcet’s Jury Theorem is considered the first non-ethical justification for democratic governance2 .
2
Marquis de Condorcet, “Essai sur l’Application de l’Analyse aux Probabilit´s des Decisions prises ` la Pluralit´ des Voix”,
e a e
1785.
External Advisory Committee Presentation – Los Alamos, New Mexico – February 13, 2009
6. Thomas Paine (English: 1737 – 1809)
Representation – Ensuring a Large Population
• Ardent supporter of the American Revolution. His passion was driven
primarily by his ideal of self-governance.3
• When a population is small, “some convenient tree will afford them a
State house.”
• As the population increases in size, representatives must “act in the
same manner as the whole body would act were they present.”
3
Thomas Paine, “Common Sense”, 1776.
External Advisory Committee Presentation – Los Alamos, New Mexico – February 13, 2009
7. Dynamically Distributed Democracy
• Imagine a government architecture where there is no a priori established
power structure: no president, no representatives, no senators.
• Imagine an Internet-based, fraud-proof, direct democratic, decision
making system.
• Problem: there are too many decisions and not enough time in a
citizen’s day to participate in all decision making processes.
External Advisory Committee Presentation – Los Alamos, New Mexico – February 13, 2009
8. citizen’s color denotes their “political tendency”, where full red is 0, full blue
citizen in this population, where xi is the is 1, and purple is 0.5. The layout algorithm chosen is the Fruchterman-
en i and, for the purpose of simulation, is Reingold layout.
a uniform distribution. Assume that every
tion of n citizens uses some social network-
create links to those individuals that they 1. Let y ∈ Rn denote the total amount of vote power that has
+
Dynamically
ir tendency the best. In practice, these links Distributed Democracy of the algorithm. Finally,
flowed to each citizen over the course
ose friend, a relative, or some public figure a ∈ {0, 1}n x ∈ [0, 1]: votercitizen i is participating (ai = 1)
• denotes whether tendency.
endencies resonate with the individual. In in the current decision : making process or not (ai = 0). The
• A ∈ Rn×n the social network.
resentatives are any citizens, not political +
values of a are biased by an unfair coin that has probability k
ve in public office. Let A ∈ [0, 1]n×n denote • a ∈ {0, 1}n: k-percent participation.
of making the citizen an active participant and 1−k of making
presenting the network, where the weight of • y ∈ Rn : received vote power.
the citizen inactive. The iterative algorithm is presented below,
+
urpose of simulation, is denoted
xj = 0.5 where ◦ denotes Rn : propagated vote power.
• π ∈ entry-wise multiplication and ≈ 1.
+
1 − |xi − xj | if link exists
= π←0
0 otherwise. i≤n
while i=1 yi < do
y ← y + (π ◦ a)
inked citizens arei = 1.0
x identical in their political
π ← π ◦ (1 − a)
strength of the link is 1.0. If their tendencies
π ← Aπ
posing, then their trust (and the strength of end
Note that a preferential attachment network
is used to generate a degree distribution that
xk = 0.0
ical social networks “in the wild” (i.e. scale- In words,Collective tendency: y ·as vote power “sinks” in
• active citizens serve x
Moreover, an assortativity parameter is used that once•they receive of y ·power, from themselves or from
The round vote x is the collective vote.
ctions in the network towards citizens with a neighbor in the network, they do not pass it on. Inactive
. The assumption here is that given a system citizens serve as vote power “sources” in that they propagate
s more likely for citizens to create links to their vote Presentation – Losthe network links to their neighbors
External Advisory Committee power over Alamos, New Mexico – February 13, 2009
dividuals than to those whose opinions are iteratively until all (or ) vote power has reached active
e resultant link matrix A is then normalized citizens. At this point, the tendency in the active population
ic in order to generate a probability distribu- is defined as δ tend = x · y. Figure 4 plots the error incurred
9. Dynamically Distributed Democracy
0.50 0.60 0.70 0.80 0.90 1.00
0.20
dynamically distributed democracy dynamically distributed democracy
proportion of correct decisions
direct democracy direct democracy
0.15
i≤n
error
1
0.10
error = xi − (x · y)
n i=1
0.05
0.00
100 90 80 70 60 50 40 30 20 10 0 100 90 80 70 60 50 40 30 20 10 0
percentage of active citizens (n) percentage of active citizens (n)
• A parameter k ∈ [0, 1] denotes the percentage of citizens that are actively
participating.
• Any subset of the whole can be made to behave as the whole. In other words, “act in
the same manner as the whole body would act were they present.”
External Advisory Committee Presentation – Los Alamos, New Mexico – February 13, 2009
10. Adam Smith (Scottish: 1723 – 1790)
Self Interested Actors - Ensuring an Enlightened Majority
• When a citizen pursues “his own interest he frequently promotes
that of the society more effectually than when he really intends to
promote it”.4
• Market mechanisms are not only useful for determining commodity prices
as they can be generally applied to information aggregation.
4
Adam Smith, “An Inquiry in the Nature and Causes of the Wealth of Nations”, 1776.
External Advisory Committee Presentation – Los Alamos, New Mexico – February 13, 2009
11. Decision Markets
• A decision market functions because it guarantees a return on
investment for quality information.
• A decision market is a tool for attracting a population of
knowledgeable citizens much like a commodity market is a tool for
attracting knowledgeable speculators.
External Advisory Committee Presentation – Los Alamos, New Mexico – February 13, 2009
12. Decision Markets
[1,1,1]
e
[0.7,0.6,0.7] • e ∈ {1}d: the environment.
• m ∈ [0, 1]d: the market.
[0.5,0.5,0.4] [0.7,0.6,0]
• red path: incentive-free market.
• blue path: incentive market.
q
1
Pi≤d
• Market accuracy: √
d i=1 (mi − ei)2.
[0,0,0.4] [0.5,0,0.4] • Rounding each dimension of m yields the
m collective decision.
[0,0,0] [0.7,0,0]
External Advisory Committee Presentation – Los Alamos, New Mexico – February 13, 2009
13. Decision Markets
1.0
1.0
incentive market
proportion of correct decisions
i≤d incentive-free market
1
0.8
0.8
error = √ (mi − ei )2
d i=1
0.6
0.6
error
0.4
0.4
0.2
0.2
incentive market
incentive-free market
0.0
0.0
1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0
average citizen knowledge (p) average citizen knowledge (p)
• Incentives in decision making ensure a thoughtful contribution of
knowledge.
• Moreover, it ensures participation from those who have knowledge of the
domain.
External Advisory Committee Presentation – Los Alamos, New Mexico – February 13, 2009
14. Artistotle (Greek: 384 – 322 B.C.)
Eudaimonia – Ensuring a Virtuous Citizenry
• Being virtuous is repeatedly choosing correctly.
• Habitual correct behavior leads to the ultimate, objective goal of life: eudaimonia –
complete engagement in the world, doing what you do because nothing else matters.5,6
• Can systems aid citizens in choosing correctly – in all aspects of life?
Aristotle
5
Aristotle, “Nicomachean Ethics”, 350 B.C.
6
Mihaly Csikszentmihalyi, “Flow: The Psychology of Optimal Experience”, Harper Perennial, 1990.
External Advisory Committee Presentation – Los Alamos, New Mexico – February 13, 2009
15. Recommendation Systems
But if the development of character is a the moral objective, it is obvious that
[...] the choices of vocation and avocations to pursue, of friends to cultivate, of
books to read are moral for they clearly influence such development.7
• Web services are continuing to build richer models of humans, resources,
and the relationships between them.
• There exists an increasing reliance on such services to aid in decision
making: correct books (Amazon.com), correct movies (NetFlix.com),
correct music (Pandora), correct occupation (Monster.com), correct
friends (PointsCommuns.com), correct life partner (Match.com), etc.
7
David L. Norton, “Democracy and Moral Development: A Politics of Virtue”, University of California Press, 1991.
External Advisory Committee Presentation – Los Alamos, New Mexico – February 13, 2009
16. Grammar-Based Random Walkers
• Algorithms can search multi-relational data structures in a way that
biases towards the requirements of the problem domain?8
What venue should I submit this article to?
Who is the best person to peer-review this article?
Who should I talk to at this conference and what should I talk to them about?
Conference attending Person
sponsors attending attending editorOf
?
?
Person Person Journal
affiliation read authored containedIn
Institution Document cites Document
8
Marko A. Rodriguez, “Grammar-Based Random Walkers”, Knowledge-Based Systems, 21(7), pp. 727-739, 2008.
External Advisory Committee Presentation – Los Alamos, New Mexico – February 13, 2009
17. Conclusion
• Thomas Jefferson stated that the purpose of a government is to ensure
“life, liberty, and the pursuit of happiness.”
• Are we as a society still too myopic to take on the bigger task of ensuring
life, liberty, and the guarantee of eudaimonia?
• This is the only reason why we should do the things we do.9
• Collective decision making systems offer solutions to this age old
problem.10
9
Marko A. Rodriguez and Alberto Pepe, “Faith in the Algorithm, Part 1: Beyond the Turing Test”, 2008.
10
Jennifer H. Watkins and Marko A. Rodriguez, “A Survey of Collective Decision Making Systems”, in Studies in
Computational Intelligence: Evolution of the Web in Artificial Intelligence Environments, Springer-Verlag, pp. 245–279, 2008.
External Advisory Committee Presentation – Los Alamos, New Mexico – February 13, 2009