#115 Social media - the main means stigmergic today
Author: Danielle Passos
In a world with a population of over 7 billion people, increasingly complex and interconnected, where practitioner half the people has access to the internet and the great majority of these is present in social media and in their networks, it is possible to observe how, in little time, the internet and the technological developments have enabled the increasing interaction between these people - what comes sowing the formation and maintenance of various complex systems - mainly by large accession the digital social networks, which allows them to a wide sharing of information and communication at global level, breaking down barriers and expanding the local to the global level. Faced with this scenario, this work aims to show how the social media showback contributing to the great spread of stigmergic events, allowing these no longer limited to small groups, geographically close, but to assume a global dimension, involving people from different countries and having as main and greater means of spreading social networks.
#117 The attribution problem in complex networks: untangling the roles of talent and experience in growing networks via the joint estimation of fitness and preferential attachment
Authors: Thong Pham, Paul Sheridan and Hidetoshi Shimodaira
How a complex network grows, i.e. how a node attracts new edges, has long been one of the main research focuses of network scientists. From a citation network between scientific papers in a specific field to a friendship network between users of a social network to a metabolic network between genes, the mechanisms for the attraction of new edges are mainly modeled as a combination of talent and experience. For experience, we focus on preferential attachment/anti-preferential attachment phenomena which represent the experience advantage/disadvantage: the number of new edges a node will get in the future is simply an increasing/decreasing function of its current number of edges, regardless of how talented it is. The attribution problem then asks the following question: exactly how much of the attractiveness of a node can be attributed to the node’s intrinsic talent, and how much this should be attributed to the node’s experience?
Here we provide the first method to answer this attribution problem. We first give a concrete mathematical model for the combination of fitness and preferential attachment by adopting a popular formation in which the probability that a node v_i with degree k gets a new node is modeled as product of A_k and η_i, where A_k is the attachment function of degree k and η_i is the node’s fitness. On average, an increasing A_k signals the preferential attachment, while a decreasing A_k presents the anti-preferential attachment. Next we estimate A_k and η_i jointly from observed network data. Using a statistical method provided by Pham et al., we successfully measure A_k and η_i for a wide range of complex networks. Finally, based on these measurements, we compare the amount of contribution of fitness and preferential attachment at each time-step in the growth process of these networks.
#119 Emergence of opinion leaders in reference networks
Authors: Mariko Ito, Hisashi Ohtsuki and Akira Sasaki
An opinion of an individual is often influenced by those of other members of the society. We describe this situation by a network model with nodes representing individuals and edges representing information flow. In such networks, individuals who have a large number of reference links are greatly influential to others, so we call them opinion leaders. Here we ask how opinion leaders emerge in an evolving network where each agent rewires one’s reference links adaptively, and what role they will play in the evolved network.
To answer these questions, we constructed a network model where, in each iteration round, an individual makes a decision on a given problem with the majority vote among his or her neighbors’ opinions, and rewires his or her links according to the performance of the linked individuals. The rewiring rule is as follows; each individual monitors his or her neighbors’ performance and breaks the link if the neighbor’s performance becomes worse than a threshold. We also assume that individuals vary in their ability to solve problems by themselves.
Our analysis shows the following results. (1) The threshold for breaking a link strongly affects whether or not opinion leaders emerge and, if emerged, the time until the emergence. (2) Even some agents with a low ability can actually become opinion leaders by referring to others. (3) When opinion leaders emerge in the model, the average performance of agents becomes higher but shows more temporal fluctuation than when each individual makes a decision independently. Some of these results are explained analytically by considering a random walk of the monitoring value of neighbor’s performance and its first passage time to the threshold.
#122 Understanding, modelling and managing social/economic complexity
Authors: Gert Jan Hofstede, Wander Jager and Tatiana Filatova
Objectives:
Complexity is not restricted to the natural sciences, but it is intrinsic to human societies. Developing resilient policies requires interdisciplinary methods of tackling complexity that do justice to natural and technical systems as well as social ones.
Description:
Most societal challenges arise from combined social and economic complexity superimposed on the management of systems that also have natural and technical components. Unintended and often undesired outcomes of the interdependent behaviours of multiple individuals and social groups pose challenges. Some examples are the current refugee crisis, turbulences in financial markets, rapid changes in health care and transportation due to artificial intelligence, and the adaptation of society to climate change (e.g., energy and safety).
As these examples show, neglecting ill-understood social sides of complex systems leads to non-resilient policy making. It is not the lack of insight into human social behaviours, but the lack of their application in managing complex systems, that limits our capacity for understanding. Effective management of this kind of complexity must address both emergence and downward causation in social/economic systems. Insights from cognitive psychology, social psychology, economics and sociology must contribute to the development of formal agent-based simulation models. Scientifically the challenges are (1) applying the existing body of social scientific knowledge in simulations of complex societal developments, and (2) offering a formal linkage between social science and the harder sciences (e.g. ecosystems science, engineering).
The presentation presents a vision carried by scientists from six Dutch Universities.
Expected gain:
This presentation showcases some pioneering work that has been done so far in Groningen (http://www.rug.nl/research/gcscs/), Wageningen (www.cas.wur.nl), and Twente (https://www.utwente.nl/bms/cstm/) mentions a large number of collaborators at other universities, and calls for cross-disciplinary collaboration in future research. Those present can decide to join the effort.
#128 Twitter sentiment in relation to political issues
Authors: Miha Grčar, Sašo Rutar, Jasmina Smailović, Darko Cherepnalkoski and Igor Mozetič
Activities on social media are reflecting political realities surprisingly well. Our previous research demonstrated that from the Twitter data, one can reconstruct political groups and nationalities of the members of European Parliament (EP) [1], or predict election results [2,3].
We analyze Twitter volume and sentiment in relation to three highly relevant political themes: public opinion in UK about Brexit, activities of the EU Parliament, and reaction of the European citizens to the current mass migration crisis. The sentiment in tweets is analyzed by constructing language- and topic-specific classification models from manually annotated tweets [4].
In this study, we focus on monitoring the public mood in UK regarding the Brexit referendum. We continuously collected relevant geo-coded tweets. A portion of the collected tweets (around 35,000) was manually labeled as "for Brexit", "against Brexit", or "no stance". The labeled tweets were used to train a supervised machine learning model which automatically labels the remaining Brexit tweets we acquired. We monitor the Twitter volume and expressed sentiment about Brexit in real-time (see http://kanarcheck.org/), and also evaluate the forecasting power of Twitter activities w.r.t. the actual referendum results.
We expect that the presented results will convince the audience of the great potential of Twitter sentiment analysis in addressing some highly controversial political issues.
[1] Cherepnalkoski, Mozetič. Retweet networks of the European Parliament: Evaluation of the community structure, Applied Network Science 1:2, 2016.
[2] Smailović, Kranjc, Grčar, Žnidaršič, Mozetič. Monitoring the Twitter sentiment during the Bulgarian elections, Proc. IEEE Intl. Conf. on Data Science and Advanced Analytics, Paris, 2015.
[3] Eom, Puliga, Smailović, Mozetič, Caldarelli. Twitter-based analysis of the dynamics of collective attention to political parties, PLoS ONE 10(7):e0131184, 2015.
[4] Mozetič, Grčar, Smailović. Multilingual Twitter sentiment classification: The role of human annotators, PLoS ONE 11(5):e0155036, 2016.
Author: Danielle Passos
In a world with a population of over 7 billion people, increasingly complex and interconnected, where practitioner half the people has access to the internet and the great majority of these is present in social media and in their networks, it is possible to observe how, in little time, the internet and the technological developments have enabled the increasing interaction between these people - what comes sowing the formation and maintenance of various complex systems - mainly by large accession the digital social networks, which allows them to a wide sharing of information and communication at global level, breaking down barriers and expanding the local to the global level. Faced with this scenario, this work aims to show how the social media showback contributing to the great spread of stigmergic events, allowing these no longer limited to small groups, geographically close, but to assume a global dimension, involving people from different countries and having as main and greater means of spreading social networks.
#117 The attribution problem in complex networks: untangling the roles of talent and experience in growing networks via the joint estimation of fitness and preferential attachment
Authors: Thong Pham, Paul Sheridan and Hidetoshi Shimodaira
How a complex network grows, i.e. how a node attracts new edges, has long been one of the main research focuses of network scientists. From a citation network between scientific papers in a specific field to a friendship network between users of a social network to a metabolic network between genes, the mechanisms for the attraction of new edges are mainly modeled as a combination of talent and experience. For experience, we focus on preferential attachment/anti-preferential attachment phenomena which represent the experience advantage/disadvantage: the number of new edges a node will get in the future is simply an increasing/decreasing function of its current number of edges, regardless of how talented it is. The attribution problem then asks the following question: exactly how much of the attractiveness of a node can be attributed to the node’s intrinsic talent, and how much this should be attributed to the node’s experience?
Here we provide the first method to answer this attribution problem. We first give a concrete mathematical model for the combination of fitness and preferential attachment by adopting a popular formation in which the probability that a node v_i with degree k gets a new node is modeled as product of A_k and η_i, where A_k is the attachment function of degree k and η_i is the node’s fitness. On average, an increasing A_k signals the preferential attachment, while a decreasing A_k presents the anti-preferential attachment. Next we estimate A_k and η_i jointly from observed network data. Using a statistical method provided by Pham et al., we successfully measure A_k and η_i for a wide range of complex networks. Finally, based on these measurements, we compare the amount of contribution of fitness and preferential attachment at each time-step in the growth process of these networks.
#119 Emergence of opinion leaders in reference networks
Authors: Mariko Ito, Hisashi Ohtsuki and Akira Sasaki
An opinion of an individual is often influenced by those of other members of the society. We describe this situation by a network model with nodes representing individuals and edges representing information flow. In such networks, individuals who have a large number of reference links are greatly influential to others, so we call them opinion leaders. Here we ask how opinion leaders emerge in an evolving network where each agent rewires one’s reference links adaptively, and what role they will play in the evolved network.
To answer these questions, we constructed a network model where, in each iteration round, an individual makes a decision on a given problem with the majority vote among his or her neighbors’ opinions, and rewires his or her links according to the performance of the linked individuals. The rewiring rule is as follows; each individual monitors his or her neighbors’ performance and breaks the link if the neighbor’s performance becomes worse than a threshold. We also assume that individuals vary in their ability to solve problems by themselves.
Our analysis shows the following results. (1) The threshold for breaking a link strongly affects whether or not opinion leaders emerge and, if emerged, the time until the emergence. (2) Even some agents with a low ability can actually become opinion leaders by referring to others. (3) When opinion leaders emerge in the model, the average performance of agents becomes higher but shows more temporal fluctuation than when each individual makes a decision independently. Some of these results are explained analytically by considering a random walk of the monitoring value of neighbor’s performance and its first passage time to the threshold.
#122 Understanding, modelling and managing social/economic complexity
Authors: Gert Jan Hofstede, Wander Jager and Tatiana Filatova
Objectives:
Complexity is not restricted to the natural sciences, but it is intrinsic to human societies. Developing resilient policies requires interdisciplinary methods of tackling complexity that do justice to natural and technical systems as well as social ones.
Description:
Most societal challenges arise from combined social and economic complexity superimposed on the management of systems that also have natural and technical components. Unintended and often undesired outcomes of the interdependent behaviours of multiple individuals and social groups pose challenges. Some examples are the current refugee crisis, turbulences in financial markets, rapid changes in health care and transportation due to artificial intelligence, and the adaptation of society to climate change (e.g., energy and safety).
As these examples show, neglecting ill-understood social sides of complex systems leads to non-resilient policy making. It is not the lack of insight into human social behaviours, but the lack of their application in managing complex systems, that limits our capacity for understanding. Effective management of this kind of complexity must address both emergence and downward causation in social/economic systems. Insights from cognitive psychology, social psychology, economics and sociology must contribute to the development of formal agent-based simulation models. Scientifically the challenges are (1) applying the existing body of social scientific knowledge in simulations of complex societal developments, and (2) offering a formal linkage between social science and the harder sciences (e.g. ecosystems science, engineering).
The presentation presents a vision carried by scientists from six Dutch Universities.
Expected gain:
This presentation showcases some pioneering work that has been done so far in Groningen (http://www.rug.nl/research/gcscs/), Wageningen (www.cas.wur.nl), and Twente (https://www.utwente.nl/bms/cstm/) mentions a large number of collaborators at other universities, and calls for cross-disciplinary collaboration in future research. Those present can decide to join the effort.
#128 Twitter sentiment in relation to political issues
Authors: Miha Grčar, Sašo Rutar, Jasmina Smailović, Darko Cherepnalkoski and Igor Mozetič
Activities on social media are reflecting political realities surprisingly well. Our previous research demonstrated that from the Twitter data, one can reconstruct political groups and nationalities of the members of European Parliament (EP) [1], or predict election results [2,3].
We analyze Twitter volume and sentiment in relation to three highly relevant political themes: public opinion in UK about Brexit, activities of the EU Parliament, and reaction of the European citizens to the current mass migration crisis. The sentiment in tweets is analyzed by constructing language- and topic-specific classification models from manually annotated tweets [4].
In this study, we focus on monitoring the public mood in UK regarding the Brexit referendum. We continuously collected relevant geo-coded tweets. A portion of the collected tweets (around 35,000) was manually labeled as "for Brexit", "against Brexit", or "no stance". The labeled tweets were used to train a supervised machine learning model which automatically labels the remaining Brexit tweets we acquired. We monitor the Twitter volume and expressed sentiment about Brexit in real-time (see http://kanarcheck.org/), and also evaluate the forecasting power of Twitter activities w.r.t. the actual referendum results.
We expect that the presented results will convince the audience of the great potential of Twitter sentiment analysis in addressing some highly controversial political issues.
[1] Cherepnalkoski, Mozetič. Retweet networks of the European Parliament: Evaluation of the community structure, Applied Network Science 1:2, 2016.
[2] Smailović, Kranjc, Grčar, Žnidaršič, Mozetič. Monitoring the Twitter sentiment during the Bulgarian elections, Proc. IEEE Intl. Conf. on Data Science and Advanced Analytics, Paris, 2015.
[3] Eom, Puliga, Smailović, Mozetič, Caldarelli. Twitter-based analysis of the dynamics of collective attention to political parties, PLoS ONE 10(7):e0131184, 2015.
[4] Mozetič, Grčar, Smailović. Multilingual Twitter sentiment classification: The role of human annotators, PLoS ONE 11(5):e0155036, 2016.
Home |
Calls |
Programme |
Registration |
Contact |
Copyright CCS © 2016