Social Decision Making with Semantic Networks and Grammar-based Particle-Swarms Marko A. Rodriguez Los Alamos National Laboratory http://cdms.lanl.gov
May 11, 2015
Social Decision Making with Semantic Networks and
Grammar-based Particle-Swarms
Marko A. Rodriguez
Los Alamos National Laboratory
http://cdms.lanl.gov
http://www.tagcrowd.com
Outline
• General Vote System Model• Proposed Semantic Network Ontology
– Tagging of individuals according to domains of trust and problems (issues) according domains
• Grammar-based Particle Swarms– Rank solutions (options) by traversing the semantic network in a
constrained manner.
• Dynamically Distributed Democracy• Complete System Model
Direct Democracy
Majority Wins
General Vote System Model
General Vote System ModelSocial networks to support fluctuating levels of participation
Semantic Network Defined
• Heterogeneous set of artifacts (nodes) and a heterogeneous set of relationships (edges).
• An ontology abstractly defines the types of artifacts and set of possible relationships.
• Requires “semantically-aware” graph algorithms for analysis.
Network Description
• Social Network - Individuals connected to one another by domains of trust.
• Decision Network - Individuals connected to the problems (issues) they raise/categorize and solutions (options) they propose.
Humans
Decisions
Social Network Description
• Humans are related according to the domains in which they trust one another.
• These domains can be top-down prescribed (taxonomy) or bottom-up defined (folksonomy).
• Domains are related to one another by their subjective similarity or can be automatically related by various text analysis algorithms.
Social Network Ontology
h_0 believes that h_2 will make a “good” decision.
NOT USED - “warm up example”
Social Network Ontology
h_0 believes that h_2 will make a “good” decisionin the domain of economics, but not in the domainof politics.
NOT USED - “warm up example”
d_1 = economicsd_0 = politics
Social Network Ontology
h_0 believes that h_2 will make a “good” decisionin the domain of d_1 (economics), but not in thedomain of d_0 (politics).
NOT USED - “warm up example”
Social Network Ontology
h_0 believes that h_2 will make a “good” decisionin the domain d_1 (economics) and furthermore,that d_0 (politics) is similar to d_1.
Decision Network Description
• Humans raise problems (issues).
• Humans categorize problems in particular domains.
• Humans propose solutions to problems (options).
• Humans vote on solutions.
Decision Network Ontology
h_1 created problem p_0. h_0 proposed s_0 asa potential solution to p_0. h_2 categorized p_0 as inthe domain d_0 and has voted on proposed solution s_2.
Grammar-Based Particles
• The purpose of the particle swarm is to calculate a stationary probability distribution in a subset of the full decision making network.– eigenvector centrality, ?PageRank?, discrete
form of constrained spreading activation.
• The propagation of the particle is constrained by its grammar.
Grammar-Based Particles
• Each particle has an abstract model of its allowed node and edge traversals (e.g. only take votedOn edges, or only go to Human nodes).
• This can be represented as a finite state machine internal to the particle (aka. a grammar)
• Each collective decision making algorithm is represented by a different grammar.– Direct Democracy and Dynamically Distributed
Democracy (DDD).• (Representative Democracy, Dictatorship, Proxy Vote)
Grammar-Based ParticlesParticle
Direct Democracy
Grammar-Based ParticlesParticle
Dynamically Distributed Democracy
Grammar-Based ParticlesDynamically Distributed Democracy
Rodriguez, M.A., Steinbock, D.J., “Societal-Scale Decision Making with Social Networks”, NACSOS, 2004.
Complete System Model
Conclusion
http://cdms.lanl.gov/
http://www.soe.ucsc.edu/~okram/
http://en.wikipedia.org/wiki/Dynamically_Distributed_Democracy