SlideShare ist ein Scribd-Unternehmen logo
1 von 39
International Workshop on Complex Systems Dynamics, IIT Madras, August 2021.
Modeling Sustainability
in Social Networks
Srinath Srinivasa
Web Science Lab
IIIT Bangalore
sri@iiitb.ac.in
Web Science
Lab
Established in 2002 as Open Systems
Lab focusing on modeling and
analytics of Complex Network
Structured Data
Changed to Web Science Lab in 2015.
Current strength: 6 PhD students, 1
postdoc, 3 MS Scholars, 22
MTech/iMTech project associates
Research verticals: Digital
Capabilities, Data Driven Governance,
Social Cognition, Responsible AI
2
Artificial vs Natural Engineering
Made of parts custom built for a specific purpose
Well-defined functionality for each part
Structure designed apriori into its present shape
Imperative design
Made of generic agents capable of playing
several roles
Autonomous actions by agents based on self-
interest and utility maximization
Structure a result of evolution and local
adjustments
Declarative design,
Sense of Self
3
Machine vs “Being” Hermeneutics
Machine hermeneutics:
Models reality in terms of inanimate matter, and
interactions between them
Roots from Ancient Greece, greatly popularized
by Newtonian models of physics
Particle foundations for physical reality
Great convergence: mass (matter) = energy
Open question: Energy and Information
Being Hermeneutics:
Models reality as a “holistic” (system of) being,
characterised by sustainable state of being,
information content in different states of being,
etc.
Characteristic of Eastern “dharmic” civilizational
thought
Being: The unit of existence, modeled as a
complex entity comprising of energy and
information
“Consciousness” foundations for reality
4
Postulates of “Being”
Primary characteristic is to “be” (settle down in
stable states or configurations)
Under certain closed or boundedness conditions,
collection of beings forms a (system of) being with
its own stable states
Sense of self (Sentient beings)
Individual and collective sense of self
Primary objective: Sustainability of the sense of self
“Being” Oriented System Design
5
Image source: Google image search
“Being” Oriented System Design
Postulate of sustainability:
Any closed system of being settles down in a
“low energy” stable state. Visible in physical
systems as elasticity, inertia, ionic interactions,
etc. and in biological systems as homeostasis.
Sentience: Systems of being with a “sense of
self”. Stable states are based on sustenance of
the sense of self, rather than on just physical
low-energy configurations.
Being: A specific form of agency. We will be
using the term “being” and “agent”
interchangeably in this work.
Being and its Environment:
Consider agent a, having its sustainable state
(w.l.o.g represented as a single state), as d(a)
Any agent interacting with a bounded environment
(called vidhi, represented as v(a)) over finitely many
interaction choices, is guaranteed to have a state
of equilibrium representing the “mutual best-
response” function. (Nash’s Theorem:
https://mathworld.wolfram.com/NashsTheorem.htm
l)
Let this equilibrium state be represented as
e(a,v(a)).
6
“Being” Oriented System Design
“Manageable” Complex Systems:
7
Tractabl
e
Mangeable Intractable
Machines
Linear
Tractable / Predictable
Dynamics by design
Beings
Non-Linear
Ergodic / Bounded state
space with invariants
Intractable, but manageable
due to invariant stable
properties
Chaos
Non-Linear
Non-Ergodic
Intractable, may have no
invariant / stable states
Networks of Beings
Understanding emergence of classes
of network topologies from individual
decision-making
Or
Understanding underlying priorities
of a population by their resultant
network structure
8
Erdös-Renyi Networks
Simplest formulation of social networks
Assumes social network connections are
formed in random
Consider a set of n nodes. There can be a total
of n(n-1)/2 undirected edges among them.
Model 1: G(n,p):
− Choose a set of m=p*n edges from this set in
random and add them to the graph
Model 2: H(n,p):
− Each of the n(n-1)/2 edges is added to the
graph with a probability p
By Vonfrisch, CC BY-SA 3.0,
https://commons.wikimedia.org/w/index.php?curid=3469734
9
Erdös-Renyi Networks
Largest Connected Component (LCC):
Most powerful community in the network
With uniform probability of addition of edges,
size of LCC undergoes an inflection when the
number of edges is approximately n/2, growing
rapidly till it start saturating approximately
around 2n edges.
Diameter of LCC increases with inflection, and
starts reducing when the LCC size saturates.
“With greater connectivity, world (LCC) grows
bigger before it grows smaller.”
10
Triadic Closure
Informally: Two people who have a common
friend are likely to become friends themselves.
The more closer they are to their common
friend, the more likely is it that they become
friends themselves
Triadic closure is not a property of how people
behave-- it is a network property
If “A is acquainted with B” implies “A spends
time with B” then increasing amount of
acquaintance between A, B and A,C within a
given time period results in B and C spending
time with each other (pigeon-hole principle)
Entrenchment: Triadic closure property creates an
effect of “entrenchment” in acquaintance networks
(Image Source: [Easley and Kleinberg 2010])
Entrenched networks low in novelty, high in mutual
familiarity (and hence, trust), thus lowering
bookkeeping costs (at the expense of novelty)
11
Watts-Strogatz Model
Refinement over the Erdos-Renyi random graph model
to accommodate triadic closure
1) Consider N nodes to be on a ring lattice labeled [0,
N-1]
2) Construct a (deterministic) graph having Nk/2 edges
by connecting each node to k/2 neighbours each on
its right and left.
3) For every node, choose one of the edges created in
step 2, and break it with a probability ß (0 ≤ ß ≤ 1)
4) Rewire the broken edges and connect them
randomly to any node in the graph
12
Watts-Strogatz Model
When ß = 0, the graph is a deterministic graph
with maximum possible triadic closure with k
edges
Creates a “resilient” network structure with a
Hamiltonian circuit: diameter no greater than
n/2, connectivity no lesser than 2, deterministic
routing with local knowledge.
Characteristic feature of entrenched
communities in human societies: high trust,
high familiarity, high resilience, low novelty.
When ß = 1, the Watts-Strogatz model is
equivalent to an Erdos-Renyi model G(n,p)
where:
When ß = 0, the graph has a deterministic
regular or near-regular structure. With ß = 1, the
degree distribution is known to be Poisson.
Degree distributions in real-world social
networks are known to be have a “hub and
spoke” (power-law, log-normal, etc.) “scale-
free” property, giving it short diameters (also
known as “small world” networks).
13
Barabasi-Albert Model
Generates a “hub-and-spoke” topology with a
power-law degree distribution:
“Scale-free” and “small-world” properties
Resilient against random failures, but
susceptible to “targeted attacks” (unlike WS
networks with clustering links)
De I, Keiono, CC BY-SA 2.5,
https://commons.wikimedia.org/w/index.php?curid=2459900
14
Barabasi-Albert Model
Preferential Attachment
Generative model for scale-free graphs:
1. Start with a small set of “seed” nodes
connected randomly
2. For every subsequent incoming node:
a. with probability γ connect to any existing
node at random
b. with probability 1-γ, connect to node k with
probability π(k), where:
where α > 0
Scale-free networks are known to be ubiquitous
in nature and emergent human networks: Blood
circulatory network, Global aviation network,
Internet topology, etc.
B-A networks are known to be resilient against
“random failures” -- i.e. failure of any k nodes
chosen at random will w.h.p. not partition the
network.
But they are not resilient against “targeted
attacks”-- failure of a small set of key hubs can
easily partition the network.
R. Cohen, K. Erez, D. Ben-Avraham, S. Havlin (2000). "Resilience of
the Internet to random breakdowns". Phys. Rev. Lett. 85: 4626. 15
Topology Breeding
Human social networks exhibit properties of
both entrenchment (WS network) and scale-free
resilience (BA network), showing resilience
against both random failures and targeted
attacks.
“Topology Breeding” an attempt to generate
networks with properties of both BA and WS
networks.
Given a society of n agents (beings):
Each agent has some “sustainability needs” which may
be potentially met by other agent in the network.
Each connection incurs a cost and brings some value
The way connections are made across the network may
give the network some “robustness” or resilience
against failure of agents and edges
Communication network has three optimization criteria:
Efficiency
Robustness
Cost 16
Patil, Sanket, Srinath Srinivasa, Saikat Mukherjee, Aditya Ramana Rachakonda, and
Venkat Venkatasubramanian. "Breeding diameter-optimal topologies for distributed
indexes." Complex Systems 18, no. 2 (2009): 175.
Patil, Sanket, Srinath Srinivasa, and Venkat Venkatasubramanian. "Classes of
optimal network topologies under multiple efficiency and robustness constraints." In
2009 ieee international conference on systems, man and cybernetics, pp. 4940-4945.
IEEE, 2009.
Topology Breeding
Use of genetic algorithms to find optimal topologies
under different constraints over efficiency, robustness
and cost
Infrastructure cost is bounded by giving each node
exactly k edges to make connections with other nodes
so as to minimize distance to all nodes, and maximize
connectivity.
Topologies generated from individual runs are
combined using a cross-over function to overcome local
minima. Topologies with lower fit functions are
discarded. Fit calculated by a parameter α that trades between
efficiency and robustness
17
Topology Breeding
Star topology
Emergent topology when α = 1 (100% importance to
efficiency and 0% importance to robustness)
Star has the smallest degree of separation for a
network of n nodes and (k=1) edge per node.
Ring topology
Emergent topology when α = 0 (100% importance
to robustness and 0% importance to efficiency)
Circle is has highest resiliency (connectivity = 2)
against targeted attacks under the cost
constraints (k=1) 18
Emergent topology when α is set to some value
between 0 and 1 (and cost factor k = 1) were a
family of topologies combining the circle and
star.
Displayed properties of “hub and spoke” with a
small diameter and a scale-free degree
distribution, and connectivity of at least 2 for a
large subset of the graph.
Degree distribution in the hub and spoke
resembles a power-law
Topology Breeding
19
Topology Breeding
With α = 1 (maximum emphasis on efficiency or
diameter reduction, and minimum emphasis on
resilience or connectivity), maximum
permissible degree (p) and number of edges (e)
were varied, with n=20 nodes.
Result is a class of tree/star structured
topologies until e=20, and then topologies with
a core ring (connectivity ≥ 2) with spokes
connecting to the core.
Degree distributions approximated by a power
law.
20
Topology Breeding
With α = 0 (minimum emphasis on efficiency or
diameter reduction, and maximum emphasis on
resilience or connectivity), maximum
permissible degree (p) and number of edges (e)
were varied, with n=20 nodes.
Result is uniformly a class of circular skip lists
(CSL) (connectivity ≥ 2) with no spokes-- only
chords. Degree distributions still approximated
by a power law (unlike WS networks).
CSL has properties of both WS and BA networks, and can be
resilient to both random failures and targeted attacks while
optimising on efficiency. Seems to appear in real world banking
networks. 21
Lux, Thomas. "Emergence of a core-periphery
structure in a simple dynamic model of the
interbank market." Journal of Economic Dynamics
and Control 52 (2015): A11-A23
Sense of Self
Basic unit of sustainability.
Agency is modeled as an optimization process
of utility maximization, driven by “self-interest”
While much research has focused on strategies
for utility maximization, relatively little interest
has gone into (computationally) modeling the
“sense of self” that drives self interest.
Classical model
From the theory of games and rational choice, by
von-Neumann and Morgenstern.
Self-interest (and the idea of “Self” itself) modeled as
a preference relation across pairs of choices: Strong
preference (>), Weak preference (≥), Indifference (||)
Valuation modeling: If A > B > C, and choice I returns
B with probability 1, while choice II returns A with
probability p, and B with probability 1-p.
The choices are said to be indifferent when for some
value of p, E(II) = E(I), or p v(A) + (1-p) u (C) = u(B)
22
Sense of Self
Rational Fools:
Critique of classical model by Amartya Sen.
Argues that it is too simplistic to reduce “sense
of self”
Human sense of self contains at least the
following extra elements:
● Rational Empathy
● Sense of fairness
● Basic level of trust
If humans were strict rational maximizers, above
kinds of interactions would be more
commonplace.
23
Agents pursuing “Rational empathy” (Pareto
improvements, rather than rational
maximization), can agree to cooperate in a one-
shot PD game.
Sense of Self
Risk Aversion:
Kahnemann and Tversky in their work on
“prospect theory” show that the human sense of
self does not treat the prospect of gains and
losses symmetrically.
Fund I, requires investment of Rs. 5000, and has a
guaranteed return of Rs. 7500 at the end of its
term. Fund II, requires investment of Rs. 5000, and
returns either nothing or Rs. 15,000 with equal
probability.
Both funds have same expected utility in classical
model, but humans shown to prefer Fund I over
Fund II.
Utility of prospects of gains saturate with more
expected gains (diminishing value of returns),
while utility of prospects of losses, grows with a
high negative slope.
24
Sense of Self
Elastic sense of identity:
Human sense of self is not a monolithic entity.
Humans often “identify” with external objects and
concepts, by making it a part of their sense of self.
Given agent a, sense of self S(a) given by:
Sa = (I, da, γa), where
I is the “identity set” comprising of objects (including
‘a’ itself) to which, the sense of self is attached, da:
{a} xI → R represents “semantic distance” to each
object in I, and 0 ≤ γa ≤ 1 represents the rate at
which identification attenuates.
Agent a identifies with object at distance d with
an attenuation of γa
d
25
Sense of Self
Computational Transcendence:
Modeling an elastic sense of identity, where an
agent “identifies” with its neighbour to different
extents.
Modeled as a utility function that combines
payoffs from oneself and neighbour, with
attenuation factor.
When both players “transcend” their sense of
self to include their neighbour to an extent of
(⅓) or more, it makes rational sense to
corporate, rather than defect. 26
Sense of Self
Regional and Global Identities:
27
HH HT TH TT
H 6 6 6 6 1 6 1 6 6 10 10 2
T 6 6 1 2 10 10 10 2 10 0 0 0
Identity and Sustainability
Regional and Global Identities:
Social identities within communities in a
population often conflict with interests of the
population as a whole, leading to the “salad
bowl” problem of diversity
Ex: Regional vs national language conflicts,
Religion versus nation loyalty conflicts, etc
Homophily: Tendency of an agent to prefer
other agents of the same regional identity,
when establishing new connections
Insularity: Tendency of an agent to distrust by
default, agents belonging to another regional
identity, and trust by default, agents belonging
to the same regional identity.
The extent of homophily and insularity within a
population of regional identities may vary.
28
Jayati Deshmukh, Srinath Srinivasa, Sridhar
Mandyam. What keeps a vibrant population
together? Complex Systems journal. (to appear)
Identity and Sustainability
Network evolution:
Agents in a network adopt either one of several
“regional” identities, or a “global” identity
Agents play a game of Iterated Prisoners’
Dilemma (IPD) with their neighbours using the
game matrix shown
Global Agents: Global agents establish new
connections with a probability in proportion to
the degree of the target agent (preferential
attachment)
Network Formation: With a probability hp
(homophily probability), regional agents
connect to other regional agents of the same
identity, and with a probability (1-hp) regional
agents connect rationally, using preferential
attachment.
29
Identity and Sustainability
Network Dynamics: With probability ip a
regional agent is flagged as “insular” and “non-
insular” with a probability (1-ip).
Non-insular regional agents, and global agents,
adopt a cooperative TIT-FOR-TAT (TFT)
strategy, that begins with offering cooperation,
and reflecting the previous move from the other
player, in subsequent iterations.
Insular agents adopt a Distrustful TFT (DTFT)
that begins with non-cooperation for agents not
belonging to same identity.
Network Evolution: A given network plays IPD
for an “epoch” τ, after which, links giving a net
negative payoff are discarded and new ones
established according to Network Formation
heuristics.
Network reaches an equilibrium (Pareto
optimality) when all links have positive payoffs.
30
Identity and ...
Simulation runs were conducted for four
extreme configurations: Low (20%)/High (80%)
Insularity/Modularity and results calibrated
when network reached equilibrium.
Case 1: No global agents
With no “global” agents, regional agents have
no problem forming a melting pot at LILH, but
break apart into separate clusters at HIHH.
At LIHH or HILH, they form a “salad bowl” of
segregated clusters, but still connected.
31
Identity and ...
Case 2: Small number (20%) of global agents
Global agents form the glue that keep the
network connected, in all four cases.
With low insularity, network is less segregated
and has a smaller diameter.
Despite playing a key role in keeping the
network together, global agents neither have
high payoffs, nor high bargaining power based
on Dominance of Neighbours (DON) metric!
32
Identity and ...
Average payoffs for global vs regional agents in
all four configurations for Case - 2.
Scatter plot of bargaining power (DON) and payoff
for global and regional agents for Case - 2.
33
Identity and ...
Case - 3: High number (80%) of global agents
Global agents overwhelm the network and
create a densely connected melting pot for LILH
and LIHH configurations. When insularity is high
(HILH and HIHH), insular regional agents are
alienated, to the extent that they find it lucrative
to break apart from the network altogether
(HIHH).
34
Identity and Sustainability
Average payoffs for global and
regional agents in HIHH
configuration, for variation in the
percentage of global agents.
Global agents’ payoffs lesser than
that of regional agents, till their
population reaches ~75% at which
stage, they alienate insular regional
agents anyway!
Identity needs to be stronger than
rationality to be a global agent!
35
Identity and Sustainability
The dynamics of identity:
Our sense of self is elastic, and can be attached
to external concepts and ideas
Identifying with an external entity,
characteristically different from rationally
associating with it
Cooperative behaviour can emerge without
long-term iteration and evolution, with elastic
identity
Regional and global identities:
Global identities are critical to keep a diverse
population of regional identities united
No rational incentive for global identity-- global
agents neither get most wealthy, nor most
powerful
Very large proportion global agents in the
demographics, can alienate regional identities
36
Conclusions
“Being-oriented computing” as a potential
modeling paradigm for analyzing and building
manageable complex systems
Elements of being: Resilience (sustainability),
Identity (sense of self)
Current work: Introducing an elastic sense of
self in RL agents for designing responsible AI
Acknowledgments:
Sanket Patil
Aditya Ramana Rachakonda
Prof. Venkatasubramanian (formerly Purdue)
Jayati Deshmukh
Prof. Sridhar Mandyam
37
Relevant Publications
Sanket Patil, Srinath Srinivasa, and Venkat
Venkatasubramanian. Classes of Optimal Network
Topologies under Multiple Efficiency and Robustness
Constraints. Proc. of the IEEE Int’l Conference on Systems,
Man and Cybernetics (SMC 2009), San Antonio, Texas, USA,
October 2009, pp. 4940 – 4945
Sanket Patil, Srinath Srinivasa. Theoretical Notes on Regular
Graphs as applied to Optimal Network Design. Proceedings
of the International Conference on Distributed Computing
and Internet Technology (ICDCIT 2010), Bhubaneswar, India,
February 2010.
Patil, Sanket, Srinath Srinivasa, Saikat Mukherjee, Aditya
Ramana Rachakonda, and Venkat Venkatasubramanian.
"Breeding diameter-optimal topologies for distributed
indexes." Complex Systems 18, no. 2 (2009): 175.
Patil, Sanket, Srinath Srinivasa, and Venkat
Venkatasubramanian. "Classes of optimal network
topologies under multiple efficiency and robustness
constraints." In 2009 ieee international conference on
systems, man and cybernetics, pp. 4940-4945. IEEE, 2009.
Jayati Deshmukh, Srinath Srinivasa. Evolution of Cooperation
with Entrenchment Effects. Proceedings of the International
Conference on Autonomous Agents and Multi-Agent
Systems (AAMAS 2015), Istanbul, Turkey, ACM Press, May
2015.
Jayati Deshmukh, Srinath Srinivasa. Cooperation and the
Globalization-Localization Dilemmas. Complex Systems
journal. (to appear)
Jayati Deshmukh, Srinath Srinivasa, Sridhar Mandyam. What
keeps a vibrant population together? Complex Systems
journal. (to appear) 38
Thank You!
39

Weitere ähnliche Inhalte

Ähnlich wie Modeling sustainability in social networks

Wanted: a larger, different kind of box
Wanted: a larger, different kind of boxWanted: a larger, different kind of box
Wanted: a larger, different kind of boxLina Martinsson Achi
 
Socialnetworkanalysis (Tin180 Com)
Socialnetworkanalysis (Tin180 Com)Socialnetworkanalysis (Tin180 Com)
Socialnetworkanalysis (Tin180 Com)Tin180 VietNam
 
COMMUNICATIONS OF THE ACM November 2004Vol. 47, No. 11 15.docx
COMMUNICATIONS OF THE ACM November  2004Vol. 47, No. 11 15.docxCOMMUNICATIONS OF THE ACM November  2004Vol. 47, No. 11 15.docx
COMMUNICATIONS OF THE ACM November 2004Vol. 47, No. 11 15.docxmonicafrancis71118
 
ICPSR - Complex Systems Models in the Social Sciences - Lecture 3 - Professor...
ICPSR - Complex Systems Models in the Social Sciences - Lecture 3 - Professor...ICPSR - Complex Systems Models in the Social Sciences - Lecture 3 - Professor...
ICPSR - Complex Systems Models in the Social Sciences - Lecture 3 - Professor...Daniel Katz
 
It’s a “small world” after all
It’s a “small world” after allIt’s a “small world” after all
It’s a “small world” after allquanmengli
 
Financial Networks and Cartography
Financial Networks and CartographyFinancial Networks and Cartography
Financial Networks and CartographyKimmo Soramaki
 
AI alignment from the Active Inference perspective 2023.pdf
AI alignment from the Active Inference perspective 2023.pdfAI alignment from the Active Inference perspective 2023.pdf
AI alignment from the Active Inference perspective 2023.pdfRoman Leventov
 
Building Stateful Clustered Microservices with Java, Actors, and Kubernetes
Building Stateful Clustered Microservices with Java, Actors, and KubernetesBuilding Stateful Clustered Microservices with Java, Actors, and Kubernetes
Building Stateful Clustered Microservices with Java, Actors, and KubernetesHugh McKee
 
Searching for patterns in crowdsourced information
Searching for patterns in crowdsourced informationSearching for patterns in crowdsourced information
Searching for patterns in crowdsourced informationSilvia Puglisi
 
Community detection in social networks[1]
Community detection in social networks[1]Community detection in social networks[1]
Community detection in social networks[1]sdnumaygmailcom
 
Akka and Kubernetes: A Symbiotic Love Story
Akka and Kubernetes: A Symbiotic Love StoryAkka and Kubernetes: A Symbiotic Love Story
Akka and Kubernetes: A Symbiotic Love StoryLightbend
 
Topology ppt
Topology pptTopology ppt
Topology pptboocse11
 
Lloyd Swarmfest 2010 Presentation
Lloyd   Swarmfest 2010 PresentationLloyd   Swarmfest 2010 Presentation
Lloyd Swarmfest 2010 Presentationkalloyd
 
Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...
Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...
Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...gerogepatton
 
Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...
Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...
Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...gerogepatton
 
GRAPH ALGORITHM TO FIND CORE PERIPHERY STRUCTURES USING MUTUAL K-NEAREST NEIG...
GRAPH ALGORITHM TO FIND CORE PERIPHERY STRUCTURES USING MUTUAL K-NEAREST NEIG...GRAPH ALGORITHM TO FIND CORE PERIPHERY STRUCTURES USING MUTUAL K-NEAREST NEIG...
GRAPH ALGORITHM TO FIND CORE PERIPHERY STRUCTURES USING MUTUAL K-NEAREST NEIG...ijaia
 
Defining Business Network
Defining Business NetworkDefining Business Network
Defining Business NetworkWaqas Tariq
 

Ähnlich wie Modeling sustainability in social networks (20)

Wanted: a larger, different kind of box
Wanted: a larger, different kind of boxWanted: a larger, different kind of box
Wanted: a larger, different kind of box
 
Socialnetworkanalysis (Tin180 Com)
Socialnetworkanalysis (Tin180 Com)Socialnetworkanalysis (Tin180 Com)
Socialnetworkanalysis (Tin180 Com)
 
COMMUNICATIONS OF THE ACM November 2004Vol. 47, No. 11 15.docx
COMMUNICATIONS OF THE ACM November  2004Vol. 47, No. 11 15.docxCOMMUNICATIONS OF THE ACM November  2004Vol. 47, No. 11 15.docx
COMMUNICATIONS OF THE ACM November 2004Vol. 47, No. 11 15.docx
 
ICPSR - Complex Systems Models in the Social Sciences - Lecture 3 - Professor...
ICPSR - Complex Systems Models in the Social Sciences - Lecture 3 - Professor...ICPSR - Complex Systems Models in the Social Sciences - Lecture 3 - Professor...
ICPSR - Complex Systems Models in the Social Sciences - Lecture 3 - Professor...
 
It’s a “small world” after all
It’s a “small world” after allIt’s a “small world” after all
It’s a “small world” after all
 
Financial Networks and Cartography
Financial Networks and CartographyFinancial Networks and Cartography
Financial Networks and Cartography
 
AI alignment from the Active Inference perspective 2023.pdf
AI alignment from the Active Inference perspective 2023.pdfAI alignment from the Active Inference perspective 2023.pdf
AI alignment from the Active Inference perspective 2023.pdf
 
Building Stateful Clustered Microservices with Java, Actors, and Kubernetes
Building Stateful Clustered Microservices with Java, Actors, and KubernetesBuilding Stateful Clustered Microservices with Java, Actors, and Kubernetes
Building Stateful Clustered Microservices with Java, Actors, and Kubernetes
 
Topology ppt
Topology pptTopology ppt
Topology ppt
 
Searching for patterns in crowdsourced information
Searching for patterns in crowdsourced informationSearching for patterns in crowdsourced information
Searching for patterns in crowdsourced information
 
Community detection in social networks[1]
Community detection in social networks[1]Community detection in social networks[1]
Community detection in social networks[1]
 
Akka and Kubernetes: A Symbiotic Love Story
Akka and Kubernetes: A Symbiotic Love StoryAkka and Kubernetes: A Symbiotic Love Story
Akka and Kubernetes: A Symbiotic Love Story
 
Topology ppt
Topology pptTopology ppt
Topology ppt
 
Topology ppt
Topology pptTopology ppt
Topology ppt
 
Lloyd Swarmfest 2010 Presentation
Lloyd   Swarmfest 2010 PresentationLloyd   Swarmfest 2010 Presentation
Lloyd Swarmfest 2010 Presentation
 
Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...
Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...
Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...
 
Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...
Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...
Graph Algorithm to Find Core Periphery Structures using Mutual K-nearest Neig...
 
GRAPH ALGORITHM TO FIND CORE PERIPHERY STRUCTURES USING MUTUAL K-NEAREST NEIG...
GRAPH ALGORITHM TO FIND CORE PERIPHERY STRUCTURES USING MUTUAL K-NEAREST NEIG...GRAPH ALGORITHM TO FIND CORE PERIPHERY STRUCTURES USING MUTUAL K-NEAREST NEIG...
GRAPH ALGORITHM TO FIND CORE PERIPHERY STRUCTURES USING MUTUAL K-NEAREST NEIG...
 
TopologyPPT.ppt
TopologyPPT.pptTopologyPPT.ppt
TopologyPPT.ppt
 
Defining Business Network
Defining Business NetworkDefining Business Network
Defining Business Network
 

Mehr von Srinath Srinivasa

Characterizing online social cognition
Characterizing online social cognitionCharacterizing online social cognition
Characterizing online social cognitionSrinath Srinivasa
 
Big Social Machines: Architecture and Challenges
Big Social Machines: Architecture and ChallengesBig Social Machines: Architecture and Challenges
Big Social Machines: Architecture and ChallengesSrinath Srinivasa
 
Abstraction and Expression on the Web
Abstraction and Expression on the WebAbstraction and Expression on the Web
Abstraction and Expression on the WebSrinath Srinivasa
 
The Power Law of Social Media: What CIOs Should Know
The Power Law of Social Media: What CIOs Should KnowThe Power Law of Social Media: What CIOs Should Know
The Power Law of Social Media: What CIOs Should KnowSrinath Srinivasa
 
Big Data and the Semantic Web: Challenges and Opportunities
Big Data and the Semantic Web: Challenges and OpportunitiesBig Data and the Semantic Web: Challenges and Opportunities
Big Data and the Semantic Web: Challenges and OpportunitiesSrinath Srinivasa
 
Aggregating Operational Knowledge in Community Settings
Aggregating Operational Knowledge in Community SettingsAggregating Operational Knowledge in Community Settings
Aggregating Operational Knowledge in Community SettingsSrinath Srinivasa
 
Information Networks and Semantics
Information Networks and SemanticsInformation Networks and Semantics
Information Networks and SemanticsSrinath Srinivasa
 
Semantics hidden within co-occurrence patterns
Semantics hidden within co-occurrence patternsSemantics hidden within co-occurrence patterns
Semantics hidden within co-occurrence patternsSrinath Srinivasa
 
The open problem of open-world computing
The open problem of open-world computingThe open problem of open-world computing
The open problem of open-world computingSrinath Srinivasa
 
Trends In Graph Data Management And Mining
Trends In Graph Data Management And MiningTrends In Graph Data Management And Mining
Trends In Graph Data Management And MiningSrinath Srinivasa
 
Information Networks And Their Dynamics
Information Networks And Their DynamicsInformation Networks And Their Dynamics
Information Networks And Their DynamicsSrinath Srinivasa
 

Mehr von Srinath Srinivasa (15)

AI and the sense of self
AI and the sense of selfAI and the sense of self
AI and the sense of self
 
Characterizing online social cognition
Characterizing online social cognitionCharacterizing online social cognition
Characterizing online social cognition
 
Open ended data
Open ended dataOpen ended data
Open ended data
 
The Web and the Mind
The Web and the MindThe Web and the Mind
The Web and the Mind
 
Big Social Machines: Architecture and Challenges
Big Social Machines: Architecture and ChallengesBig Social Machines: Architecture and Challenges
Big Social Machines: Architecture and Challenges
 
Abstraction and Expression on the Web
Abstraction and Expression on the WebAbstraction and Expression on the Web
Abstraction and Expression on the Web
 
Towards a "Mindful" Web
Towards a "Mindful" WebTowards a "Mindful" Web
Towards a "Mindful" Web
 
The Power Law of Social Media: What CIOs Should Know
The Power Law of Social Media: What CIOs Should KnowThe Power Law of Social Media: What CIOs Should Know
The Power Law of Social Media: What CIOs Should Know
 
Big Data and the Semantic Web: Challenges and Opportunities
Big Data and the Semantic Web: Challenges and OpportunitiesBig Data and the Semantic Web: Challenges and Opportunities
Big Data and the Semantic Web: Challenges and Opportunities
 
Aggregating Operational Knowledge in Community Settings
Aggregating Operational Knowledge in Community SettingsAggregating Operational Knowledge in Community Settings
Aggregating Operational Knowledge in Community Settings
 
Information Networks and Semantics
Information Networks and SemanticsInformation Networks and Semantics
Information Networks and Semantics
 
Semantics hidden within co-occurrence patterns
Semantics hidden within co-occurrence patternsSemantics hidden within co-occurrence patterns
Semantics hidden within co-occurrence patterns
 
The open problem of open-world computing
The open problem of open-world computingThe open problem of open-world computing
The open problem of open-world computing
 
Trends In Graph Data Management And Mining
Trends In Graph Data Management And MiningTrends In Graph Data Management And Mining
Trends In Graph Data Management And Mining
 
Information Networks And Their Dynamics
Information Networks And Their DynamicsInformation Networks And Their Dynamics
Information Networks And Their Dynamics
 

Kürzlich hochgeladen

basic entomology with insect anatomy and taxonomy
basic entomology with insect anatomy and taxonomybasic entomology with insect anatomy and taxonomy
basic entomology with insect anatomy and taxonomyDrAnita Sharma
 
preservation, maintanence and improvement of industrial organism.pptx
preservation, maintanence and improvement of industrial organism.pptxpreservation, maintanence and improvement of industrial organism.pptx
preservation, maintanence and improvement of industrial organism.pptxnoordubaliya2003
 
Pests of soyabean_Binomics_IdentificationDr.UPR.pdf
Pests of soyabean_Binomics_IdentificationDr.UPR.pdfPests of soyabean_Binomics_IdentificationDr.UPR.pdf
Pests of soyabean_Binomics_IdentificationDr.UPR.pdfPirithiRaju
 
Speech, hearing, noise, intelligibility.pptx
Speech, hearing, noise, intelligibility.pptxSpeech, hearing, noise, intelligibility.pptx
Speech, hearing, noise, intelligibility.pptxpriyankatabhane
 
Bioteknologi kelas 10 kumer smapsa .pptx
Bioteknologi kelas 10 kumer smapsa .pptxBioteknologi kelas 10 kumer smapsa .pptx
Bioteknologi kelas 10 kumer smapsa .pptx023NiWayanAnggiSriWa
 
Topic 9- General Principles of International Law.pptx
Topic 9- General Principles of International Law.pptxTopic 9- General Principles of International Law.pptx
Topic 9- General Principles of International Law.pptxJorenAcuavera1
 
LIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptx
LIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptxLIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptx
LIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptxmalonesandreagweneth
 
《Queensland毕业文凭-昆士兰大学毕业证成绩单》
《Queensland毕业文凭-昆士兰大学毕业证成绩单》《Queensland毕业文凭-昆士兰大学毕业证成绩单》
《Queensland毕业文凭-昆士兰大学毕业证成绩单》rnrncn29
 
ALL ABOUT MIXTURES IN GRADE 7 CLASS PPTX
ALL ABOUT MIXTURES IN GRADE 7 CLASS PPTXALL ABOUT MIXTURES IN GRADE 7 CLASS PPTX
ALL ABOUT MIXTURES IN GRADE 7 CLASS PPTXDole Philippines School
 
THE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptx
THE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptxTHE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptx
THE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptxNandakishor Bhaurao Deshmukh
 
STOPPED FLOW METHOD & APPLICATION MURUGAVENI B.pptx
STOPPED FLOW METHOD & APPLICATION MURUGAVENI B.pptxSTOPPED FLOW METHOD & APPLICATION MURUGAVENI B.pptx
STOPPED FLOW METHOD & APPLICATION MURUGAVENI B.pptxMurugaveni B
 
ECG Graph Monitoring with AD8232 ECG Sensor & Arduino.pptx
ECG Graph Monitoring with AD8232 ECG Sensor & Arduino.pptxECG Graph Monitoring with AD8232 ECG Sensor & Arduino.pptx
ECG Graph Monitoring with AD8232 ECG Sensor & Arduino.pptxmaryFF1
 
Pests of Bengal gram_Identification_Dr.UPR.pdf
Pests of Bengal gram_Identification_Dr.UPR.pdfPests of Bengal gram_Identification_Dr.UPR.pdf
Pests of Bengal gram_Identification_Dr.UPR.pdfPirithiRaju
 
User Guide: Capricorn FLX™ Weather Station
User Guide: Capricorn FLX™ Weather StationUser Guide: Capricorn FLX™ Weather Station
User Guide: Capricorn FLX™ Weather StationColumbia Weather Systems
 
Pests of jatropha_Bionomics_identification_Dr.UPR.pdf
Pests of jatropha_Bionomics_identification_Dr.UPR.pdfPests of jatropha_Bionomics_identification_Dr.UPR.pdf
Pests of jatropha_Bionomics_identification_Dr.UPR.pdfPirithiRaju
 
Best Call Girls In Sector 29 Gurgaon❤️8860477959 EscorTs Service In 24/7 Delh...
Best Call Girls In Sector 29 Gurgaon❤️8860477959 EscorTs Service In 24/7 Delh...Best Call Girls In Sector 29 Gurgaon❤️8860477959 EscorTs Service In 24/7 Delh...
Best Call Girls In Sector 29 Gurgaon❤️8860477959 EscorTs Service In 24/7 Delh...lizamodels9
 
Call Girls in Majnu Ka Tilla Delhi 🔝9711014705🔝 Genuine
Call Girls in Majnu Ka Tilla Delhi 🔝9711014705🔝 GenuineCall Girls in Majnu Ka Tilla Delhi 🔝9711014705🔝 Genuine
Call Girls in Majnu Ka Tilla Delhi 🔝9711014705🔝 Genuinethapagita
 
Fertilization: Sperm and the egg—collectively called the gametes—fuse togethe...
Fertilization: Sperm and the egg—collectively called the gametes—fuse togethe...Fertilization: Sperm and the egg—collectively called the gametes—fuse togethe...
Fertilization: Sperm and the egg—collectively called the gametes—fuse togethe...D. B. S. College Kanpur
 
Dubai Calls Girl Lisa O525547819 Lexi Call Girls In Dubai
Dubai Calls Girl Lisa O525547819 Lexi Call Girls In DubaiDubai Calls Girl Lisa O525547819 Lexi Call Girls In Dubai
Dubai Calls Girl Lisa O525547819 Lexi Call Girls In Dubaikojalkojal131
 
Neurodevelopmental disorders according to the dsm 5 tr
Neurodevelopmental disorders according to the dsm 5 trNeurodevelopmental disorders according to the dsm 5 tr
Neurodevelopmental disorders according to the dsm 5 trssuser06f238
 

Kürzlich hochgeladen (20)

basic entomology with insect anatomy and taxonomy
basic entomology with insect anatomy and taxonomybasic entomology with insect anatomy and taxonomy
basic entomology with insect anatomy and taxonomy
 
preservation, maintanence and improvement of industrial organism.pptx
preservation, maintanence and improvement of industrial organism.pptxpreservation, maintanence and improvement of industrial organism.pptx
preservation, maintanence and improvement of industrial organism.pptx
 
Pests of soyabean_Binomics_IdentificationDr.UPR.pdf
Pests of soyabean_Binomics_IdentificationDr.UPR.pdfPests of soyabean_Binomics_IdentificationDr.UPR.pdf
Pests of soyabean_Binomics_IdentificationDr.UPR.pdf
 
Speech, hearing, noise, intelligibility.pptx
Speech, hearing, noise, intelligibility.pptxSpeech, hearing, noise, intelligibility.pptx
Speech, hearing, noise, intelligibility.pptx
 
Bioteknologi kelas 10 kumer smapsa .pptx
Bioteknologi kelas 10 kumer smapsa .pptxBioteknologi kelas 10 kumer smapsa .pptx
Bioteknologi kelas 10 kumer smapsa .pptx
 
Topic 9- General Principles of International Law.pptx
Topic 9- General Principles of International Law.pptxTopic 9- General Principles of International Law.pptx
Topic 9- General Principles of International Law.pptx
 
LIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptx
LIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptxLIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptx
LIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptx
 
《Queensland毕业文凭-昆士兰大学毕业证成绩单》
《Queensland毕业文凭-昆士兰大学毕业证成绩单》《Queensland毕业文凭-昆士兰大学毕业证成绩单》
《Queensland毕业文凭-昆士兰大学毕业证成绩单》
 
ALL ABOUT MIXTURES IN GRADE 7 CLASS PPTX
ALL ABOUT MIXTURES IN GRADE 7 CLASS PPTXALL ABOUT MIXTURES IN GRADE 7 CLASS PPTX
ALL ABOUT MIXTURES IN GRADE 7 CLASS PPTX
 
THE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptx
THE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptxTHE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptx
THE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptx
 
STOPPED FLOW METHOD & APPLICATION MURUGAVENI B.pptx
STOPPED FLOW METHOD & APPLICATION MURUGAVENI B.pptxSTOPPED FLOW METHOD & APPLICATION MURUGAVENI B.pptx
STOPPED FLOW METHOD & APPLICATION MURUGAVENI B.pptx
 
ECG Graph Monitoring with AD8232 ECG Sensor & Arduino.pptx
ECG Graph Monitoring with AD8232 ECG Sensor & Arduino.pptxECG Graph Monitoring with AD8232 ECG Sensor & Arduino.pptx
ECG Graph Monitoring with AD8232 ECG Sensor & Arduino.pptx
 
Pests of Bengal gram_Identification_Dr.UPR.pdf
Pests of Bengal gram_Identification_Dr.UPR.pdfPests of Bengal gram_Identification_Dr.UPR.pdf
Pests of Bengal gram_Identification_Dr.UPR.pdf
 
User Guide: Capricorn FLX™ Weather Station
User Guide: Capricorn FLX™ Weather StationUser Guide: Capricorn FLX™ Weather Station
User Guide: Capricorn FLX™ Weather Station
 
Pests of jatropha_Bionomics_identification_Dr.UPR.pdf
Pests of jatropha_Bionomics_identification_Dr.UPR.pdfPests of jatropha_Bionomics_identification_Dr.UPR.pdf
Pests of jatropha_Bionomics_identification_Dr.UPR.pdf
 
Best Call Girls In Sector 29 Gurgaon❤️8860477959 EscorTs Service In 24/7 Delh...
Best Call Girls In Sector 29 Gurgaon❤️8860477959 EscorTs Service In 24/7 Delh...Best Call Girls In Sector 29 Gurgaon❤️8860477959 EscorTs Service In 24/7 Delh...
Best Call Girls In Sector 29 Gurgaon❤️8860477959 EscorTs Service In 24/7 Delh...
 
Call Girls in Majnu Ka Tilla Delhi 🔝9711014705🔝 Genuine
Call Girls in Majnu Ka Tilla Delhi 🔝9711014705🔝 GenuineCall Girls in Majnu Ka Tilla Delhi 🔝9711014705🔝 Genuine
Call Girls in Majnu Ka Tilla Delhi 🔝9711014705🔝 Genuine
 
Fertilization: Sperm and the egg—collectively called the gametes—fuse togethe...
Fertilization: Sperm and the egg—collectively called the gametes—fuse togethe...Fertilization: Sperm and the egg—collectively called the gametes—fuse togethe...
Fertilization: Sperm and the egg—collectively called the gametes—fuse togethe...
 
Dubai Calls Girl Lisa O525547819 Lexi Call Girls In Dubai
Dubai Calls Girl Lisa O525547819 Lexi Call Girls In DubaiDubai Calls Girl Lisa O525547819 Lexi Call Girls In Dubai
Dubai Calls Girl Lisa O525547819 Lexi Call Girls In Dubai
 
Neurodevelopmental disorders according to the dsm 5 tr
Neurodevelopmental disorders according to the dsm 5 trNeurodevelopmental disorders according to the dsm 5 tr
Neurodevelopmental disorders according to the dsm 5 tr
 

Modeling sustainability in social networks

  • 1. International Workshop on Complex Systems Dynamics, IIT Madras, August 2021. Modeling Sustainability in Social Networks Srinath Srinivasa Web Science Lab IIIT Bangalore sri@iiitb.ac.in
  • 2. Web Science Lab Established in 2002 as Open Systems Lab focusing on modeling and analytics of Complex Network Structured Data Changed to Web Science Lab in 2015. Current strength: 6 PhD students, 1 postdoc, 3 MS Scholars, 22 MTech/iMTech project associates Research verticals: Digital Capabilities, Data Driven Governance, Social Cognition, Responsible AI 2
  • 3. Artificial vs Natural Engineering Made of parts custom built for a specific purpose Well-defined functionality for each part Structure designed apriori into its present shape Imperative design Made of generic agents capable of playing several roles Autonomous actions by agents based on self- interest and utility maximization Structure a result of evolution and local adjustments Declarative design, Sense of Self 3
  • 4. Machine vs “Being” Hermeneutics Machine hermeneutics: Models reality in terms of inanimate matter, and interactions between them Roots from Ancient Greece, greatly popularized by Newtonian models of physics Particle foundations for physical reality Great convergence: mass (matter) = energy Open question: Energy and Information Being Hermeneutics: Models reality as a “holistic” (system of) being, characterised by sustainable state of being, information content in different states of being, etc. Characteristic of Eastern “dharmic” civilizational thought Being: The unit of existence, modeled as a complex entity comprising of energy and information “Consciousness” foundations for reality 4
  • 5. Postulates of “Being” Primary characteristic is to “be” (settle down in stable states or configurations) Under certain closed or boundedness conditions, collection of beings forms a (system of) being with its own stable states Sense of self (Sentient beings) Individual and collective sense of self Primary objective: Sustainability of the sense of self “Being” Oriented System Design 5 Image source: Google image search
  • 6. “Being” Oriented System Design Postulate of sustainability: Any closed system of being settles down in a “low energy” stable state. Visible in physical systems as elasticity, inertia, ionic interactions, etc. and in biological systems as homeostasis. Sentience: Systems of being with a “sense of self”. Stable states are based on sustenance of the sense of self, rather than on just physical low-energy configurations. Being: A specific form of agency. We will be using the term “being” and “agent” interchangeably in this work. Being and its Environment: Consider agent a, having its sustainable state (w.l.o.g represented as a single state), as d(a) Any agent interacting with a bounded environment (called vidhi, represented as v(a)) over finitely many interaction choices, is guaranteed to have a state of equilibrium representing the “mutual best- response” function. (Nash’s Theorem: https://mathworld.wolfram.com/NashsTheorem.htm l) Let this equilibrium state be represented as e(a,v(a)). 6
  • 7. “Being” Oriented System Design “Manageable” Complex Systems: 7 Tractabl e Mangeable Intractable Machines Linear Tractable / Predictable Dynamics by design Beings Non-Linear Ergodic / Bounded state space with invariants Intractable, but manageable due to invariant stable properties Chaos Non-Linear Non-Ergodic Intractable, may have no invariant / stable states
  • 8. Networks of Beings Understanding emergence of classes of network topologies from individual decision-making Or Understanding underlying priorities of a population by their resultant network structure 8
  • 9. Erdös-Renyi Networks Simplest formulation of social networks Assumes social network connections are formed in random Consider a set of n nodes. There can be a total of n(n-1)/2 undirected edges among them. Model 1: G(n,p): − Choose a set of m=p*n edges from this set in random and add them to the graph Model 2: H(n,p): − Each of the n(n-1)/2 edges is added to the graph with a probability p By Vonfrisch, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=3469734 9
  • 10. Erdös-Renyi Networks Largest Connected Component (LCC): Most powerful community in the network With uniform probability of addition of edges, size of LCC undergoes an inflection when the number of edges is approximately n/2, growing rapidly till it start saturating approximately around 2n edges. Diameter of LCC increases with inflection, and starts reducing when the LCC size saturates. “With greater connectivity, world (LCC) grows bigger before it grows smaller.” 10
  • 11. Triadic Closure Informally: Two people who have a common friend are likely to become friends themselves. The more closer they are to their common friend, the more likely is it that they become friends themselves Triadic closure is not a property of how people behave-- it is a network property If “A is acquainted with B” implies “A spends time with B” then increasing amount of acquaintance between A, B and A,C within a given time period results in B and C spending time with each other (pigeon-hole principle) Entrenchment: Triadic closure property creates an effect of “entrenchment” in acquaintance networks (Image Source: [Easley and Kleinberg 2010]) Entrenched networks low in novelty, high in mutual familiarity (and hence, trust), thus lowering bookkeeping costs (at the expense of novelty) 11
  • 12. Watts-Strogatz Model Refinement over the Erdos-Renyi random graph model to accommodate triadic closure 1) Consider N nodes to be on a ring lattice labeled [0, N-1] 2) Construct a (deterministic) graph having Nk/2 edges by connecting each node to k/2 neighbours each on its right and left. 3) For every node, choose one of the edges created in step 2, and break it with a probability ß (0 ≤ ß ≤ 1) 4) Rewire the broken edges and connect them randomly to any node in the graph 12
  • 13. Watts-Strogatz Model When ß = 0, the graph is a deterministic graph with maximum possible triadic closure with k edges Creates a “resilient” network structure with a Hamiltonian circuit: diameter no greater than n/2, connectivity no lesser than 2, deterministic routing with local knowledge. Characteristic feature of entrenched communities in human societies: high trust, high familiarity, high resilience, low novelty. When ß = 1, the Watts-Strogatz model is equivalent to an Erdos-Renyi model G(n,p) where: When ß = 0, the graph has a deterministic regular or near-regular structure. With ß = 1, the degree distribution is known to be Poisson. Degree distributions in real-world social networks are known to be have a “hub and spoke” (power-law, log-normal, etc.) “scale- free” property, giving it short diameters (also known as “small world” networks). 13
  • 14. Barabasi-Albert Model Generates a “hub-and-spoke” topology with a power-law degree distribution: “Scale-free” and “small-world” properties Resilient against random failures, but susceptible to “targeted attacks” (unlike WS networks with clustering links) De I, Keiono, CC BY-SA 2.5, https://commons.wikimedia.org/w/index.php?curid=2459900 14
  • 15. Barabasi-Albert Model Preferential Attachment Generative model for scale-free graphs: 1. Start with a small set of “seed” nodes connected randomly 2. For every subsequent incoming node: a. with probability γ connect to any existing node at random b. with probability 1-γ, connect to node k with probability π(k), where: where α > 0 Scale-free networks are known to be ubiquitous in nature and emergent human networks: Blood circulatory network, Global aviation network, Internet topology, etc. B-A networks are known to be resilient against “random failures” -- i.e. failure of any k nodes chosen at random will w.h.p. not partition the network. But they are not resilient against “targeted attacks”-- failure of a small set of key hubs can easily partition the network. R. Cohen, K. Erez, D. Ben-Avraham, S. Havlin (2000). "Resilience of the Internet to random breakdowns". Phys. Rev. Lett. 85: 4626. 15
  • 16. Topology Breeding Human social networks exhibit properties of both entrenchment (WS network) and scale-free resilience (BA network), showing resilience against both random failures and targeted attacks. “Topology Breeding” an attempt to generate networks with properties of both BA and WS networks. Given a society of n agents (beings): Each agent has some “sustainability needs” which may be potentially met by other agent in the network. Each connection incurs a cost and brings some value The way connections are made across the network may give the network some “robustness” or resilience against failure of agents and edges Communication network has three optimization criteria: Efficiency Robustness Cost 16 Patil, Sanket, Srinath Srinivasa, Saikat Mukherjee, Aditya Ramana Rachakonda, and Venkat Venkatasubramanian. "Breeding diameter-optimal topologies for distributed indexes." Complex Systems 18, no. 2 (2009): 175. Patil, Sanket, Srinath Srinivasa, and Venkat Venkatasubramanian. "Classes of optimal network topologies under multiple efficiency and robustness constraints." In 2009 ieee international conference on systems, man and cybernetics, pp. 4940-4945. IEEE, 2009.
  • 17. Topology Breeding Use of genetic algorithms to find optimal topologies under different constraints over efficiency, robustness and cost Infrastructure cost is bounded by giving each node exactly k edges to make connections with other nodes so as to minimize distance to all nodes, and maximize connectivity. Topologies generated from individual runs are combined using a cross-over function to overcome local minima. Topologies with lower fit functions are discarded. Fit calculated by a parameter α that trades between efficiency and robustness 17
  • 18. Topology Breeding Star topology Emergent topology when α = 1 (100% importance to efficiency and 0% importance to robustness) Star has the smallest degree of separation for a network of n nodes and (k=1) edge per node. Ring topology Emergent topology when α = 0 (100% importance to robustness and 0% importance to efficiency) Circle is has highest resiliency (connectivity = 2) against targeted attacks under the cost constraints (k=1) 18
  • 19. Emergent topology when α is set to some value between 0 and 1 (and cost factor k = 1) were a family of topologies combining the circle and star. Displayed properties of “hub and spoke” with a small diameter and a scale-free degree distribution, and connectivity of at least 2 for a large subset of the graph. Degree distribution in the hub and spoke resembles a power-law Topology Breeding 19
  • 20. Topology Breeding With α = 1 (maximum emphasis on efficiency or diameter reduction, and minimum emphasis on resilience or connectivity), maximum permissible degree (p) and number of edges (e) were varied, with n=20 nodes. Result is a class of tree/star structured topologies until e=20, and then topologies with a core ring (connectivity ≥ 2) with spokes connecting to the core. Degree distributions approximated by a power law. 20
  • 21. Topology Breeding With α = 0 (minimum emphasis on efficiency or diameter reduction, and maximum emphasis on resilience or connectivity), maximum permissible degree (p) and number of edges (e) were varied, with n=20 nodes. Result is uniformly a class of circular skip lists (CSL) (connectivity ≥ 2) with no spokes-- only chords. Degree distributions still approximated by a power law (unlike WS networks). CSL has properties of both WS and BA networks, and can be resilient to both random failures and targeted attacks while optimising on efficiency. Seems to appear in real world banking networks. 21 Lux, Thomas. "Emergence of a core-periphery structure in a simple dynamic model of the interbank market." Journal of Economic Dynamics and Control 52 (2015): A11-A23
  • 22. Sense of Self Basic unit of sustainability. Agency is modeled as an optimization process of utility maximization, driven by “self-interest” While much research has focused on strategies for utility maximization, relatively little interest has gone into (computationally) modeling the “sense of self” that drives self interest. Classical model From the theory of games and rational choice, by von-Neumann and Morgenstern. Self-interest (and the idea of “Self” itself) modeled as a preference relation across pairs of choices: Strong preference (>), Weak preference (≥), Indifference (||) Valuation modeling: If A > B > C, and choice I returns B with probability 1, while choice II returns A with probability p, and B with probability 1-p. The choices are said to be indifferent when for some value of p, E(II) = E(I), or p v(A) + (1-p) u (C) = u(B) 22
  • 23. Sense of Self Rational Fools: Critique of classical model by Amartya Sen. Argues that it is too simplistic to reduce “sense of self” Human sense of self contains at least the following extra elements: ● Rational Empathy ● Sense of fairness ● Basic level of trust If humans were strict rational maximizers, above kinds of interactions would be more commonplace. 23 Agents pursuing “Rational empathy” (Pareto improvements, rather than rational maximization), can agree to cooperate in a one- shot PD game.
  • 24. Sense of Self Risk Aversion: Kahnemann and Tversky in their work on “prospect theory” show that the human sense of self does not treat the prospect of gains and losses symmetrically. Fund I, requires investment of Rs. 5000, and has a guaranteed return of Rs. 7500 at the end of its term. Fund II, requires investment of Rs. 5000, and returns either nothing or Rs. 15,000 with equal probability. Both funds have same expected utility in classical model, but humans shown to prefer Fund I over Fund II. Utility of prospects of gains saturate with more expected gains (diminishing value of returns), while utility of prospects of losses, grows with a high negative slope. 24
  • 25. Sense of Self Elastic sense of identity: Human sense of self is not a monolithic entity. Humans often “identify” with external objects and concepts, by making it a part of their sense of self. Given agent a, sense of self S(a) given by: Sa = (I, da, γa), where I is the “identity set” comprising of objects (including ‘a’ itself) to which, the sense of self is attached, da: {a} xI → R represents “semantic distance” to each object in I, and 0 ≤ γa ≤ 1 represents the rate at which identification attenuates. Agent a identifies with object at distance d with an attenuation of γa d 25
  • 26. Sense of Self Computational Transcendence: Modeling an elastic sense of identity, where an agent “identifies” with its neighbour to different extents. Modeled as a utility function that combines payoffs from oneself and neighbour, with attenuation factor. When both players “transcend” their sense of self to include their neighbour to an extent of (⅓) or more, it makes rational sense to corporate, rather than defect. 26
  • 27. Sense of Self Regional and Global Identities: 27 HH HT TH TT H 6 6 6 6 1 6 1 6 6 10 10 2 T 6 6 1 2 10 10 10 2 10 0 0 0
  • 28. Identity and Sustainability Regional and Global Identities: Social identities within communities in a population often conflict with interests of the population as a whole, leading to the “salad bowl” problem of diversity Ex: Regional vs national language conflicts, Religion versus nation loyalty conflicts, etc Homophily: Tendency of an agent to prefer other agents of the same regional identity, when establishing new connections Insularity: Tendency of an agent to distrust by default, agents belonging to another regional identity, and trust by default, agents belonging to the same regional identity. The extent of homophily and insularity within a population of regional identities may vary. 28 Jayati Deshmukh, Srinath Srinivasa, Sridhar Mandyam. What keeps a vibrant population together? Complex Systems journal. (to appear)
  • 29. Identity and Sustainability Network evolution: Agents in a network adopt either one of several “regional” identities, or a “global” identity Agents play a game of Iterated Prisoners’ Dilemma (IPD) with their neighbours using the game matrix shown Global Agents: Global agents establish new connections with a probability in proportion to the degree of the target agent (preferential attachment) Network Formation: With a probability hp (homophily probability), regional agents connect to other regional agents of the same identity, and with a probability (1-hp) regional agents connect rationally, using preferential attachment. 29
  • 30. Identity and Sustainability Network Dynamics: With probability ip a regional agent is flagged as “insular” and “non- insular” with a probability (1-ip). Non-insular regional agents, and global agents, adopt a cooperative TIT-FOR-TAT (TFT) strategy, that begins with offering cooperation, and reflecting the previous move from the other player, in subsequent iterations. Insular agents adopt a Distrustful TFT (DTFT) that begins with non-cooperation for agents not belonging to same identity. Network Evolution: A given network plays IPD for an “epoch” τ, after which, links giving a net negative payoff are discarded and new ones established according to Network Formation heuristics. Network reaches an equilibrium (Pareto optimality) when all links have positive payoffs. 30
  • 31. Identity and ... Simulation runs were conducted for four extreme configurations: Low (20%)/High (80%) Insularity/Modularity and results calibrated when network reached equilibrium. Case 1: No global agents With no “global” agents, regional agents have no problem forming a melting pot at LILH, but break apart into separate clusters at HIHH. At LIHH or HILH, they form a “salad bowl” of segregated clusters, but still connected. 31
  • 32. Identity and ... Case 2: Small number (20%) of global agents Global agents form the glue that keep the network connected, in all four cases. With low insularity, network is less segregated and has a smaller diameter. Despite playing a key role in keeping the network together, global agents neither have high payoffs, nor high bargaining power based on Dominance of Neighbours (DON) metric! 32
  • 33. Identity and ... Average payoffs for global vs regional agents in all four configurations for Case - 2. Scatter plot of bargaining power (DON) and payoff for global and regional agents for Case - 2. 33
  • 34. Identity and ... Case - 3: High number (80%) of global agents Global agents overwhelm the network and create a densely connected melting pot for LILH and LIHH configurations. When insularity is high (HILH and HIHH), insular regional agents are alienated, to the extent that they find it lucrative to break apart from the network altogether (HIHH). 34
  • 35. Identity and Sustainability Average payoffs for global and regional agents in HIHH configuration, for variation in the percentage of global agents. Global agents’ payoffs lesser than that of regional agents, till their population reaches ~75% at which stage, they alienate insular regional agents anyway! Identity needs to be stronger than rationality to be a global agent! 35
  • 36. Identity and Sustainability The dynamics of identity: Our sense of self is elastic, and can be attached to external concepts and ideas Identifying with an external entity, characteristically different from rationally associating with it Cooperative behaviour can emerge without long-term iteration and evolution, with elastic identity Regional and global identities: Global identities are critical to keep a diverse population of regional identities united No rational incentive for global identity-- global agents neither get most wealthy, nor most powerful Very large proportion global agents in the demographics, can alienate regional identities 36
  • 37. Conclusions “Being-oriented computing” as a potential modeling paradigm for analyzing and building manageable complex systems Elements of being: Resilience (sustainability), Identity (sense of self) Current work: Introducing an elastic sense of self in RL agents for designing responsible AI Acknowledgments: Sanket Patil Aditya Ramana Rachakonda Prof. Venkatasubramanian (formerly Purdue) Jayati Deshmukh Prof. Sridhar Mandyam 37
  • 38. Relevant Publications Sanket Patil, Srinath Srinivasa, and Venkat Venkatasubramanian. Classes of Optimal Network Topologies under Multiple Efficiency and Robustness Constraints. Proc. of the IEEE Int’l Conference on Systems, Man and Cybernetics (SMC 2009), San Antonio, Texas, USA, October 2009, pp. 4940 – 4945 Sanket Patil, Srinath Srinivasa. Theoretical Notes on Regular Graphs as applied to Optimal Network Design. Proceedings of the International Conference on Distributed Computing and Internet Technology (ICDCIT 2010), Bhubaneswar, India, February 2010. Patil, Sanket, Srinath Srinivasa, Saikat Mukherjee, Aditya Ramana Rachakonda, and Venkat Venkatasubramanian. "Breeding diameter-optimal topologies for distributed indexes." Complex Systems 18, no. 2 (2009): 175. Patil, Sanket, Srinath Srinivasa, and Venkat Venkatasubramanian. "Classes of optimal network topologies under multiple efficiency and robustness constraints." In 2009 ieee international conference on systems, man and cybernetics, pp. 4940-4945. IEEE, 2009. Jayati Deshmukh, Srinath Srinivasa. Evolution of Cooperation with Entrenchment Effects. Proceedings of the International Conference on Autonomous Agents and Multi-Agent Systems (AAMAS 2015), Istanbul, Turkey, ACM Press, May 2015. Jayati Deshmukh, Srinath Srinivasa. Cooperation and the Globalization-Localization Dilemmas. Complex Systems journal. (to appear) Jayati Deshmukh, Srinath Srinivasa, Sridhar Mandyam. What keeps a vibrant population together? Complex Systems journal. (to appear) 38