Intrinsic Reward andSocial Reward  Cost         Fraction of the population
Threshold and Cascades
Threshold and Cascades
Threshold and Cascades
Threshold and CascadesLinear Threshold Model      3       2          3       2                         4              3
Threshold and CascadesIndependent Cascades Model          0.7          0.5                 0.4                0.1    0.2
Valid Points• Triadic Closures• Clusters force members to conform to  prevailing behaviors.• Weak links introduces new inf...
Valid PointsTriadic Closure
Valid PointsClusters on emerging Behavior
Valid PointsWeak Ties
Bias•   Vocal minority•   Time of day is ignored•   Social Network bias•   Forward Rewards
BiasTime of Day is Ignored
BiasSocial Network Bias
BiasForward Rewards
Recent Studies• Purchase Susceptibility and  Quantifying Trust• Maximizing Information Diffusion within  a Social Network
Recent StudiesPurchasing Susceptibility• E-commerce websites.• User to user communication upon  purchase important.• Triad...
Recent StudiesMaximizing Diffusion• Given a social network, how should we  choose a set S so that we will be able  to maxi...
Methods applied• Bayes theorem for information  cascades• Games for modelling cost and reward  within a social network• Na...
Methods AppliedGames         p* = r(z)f(z)
Methods AppliedNatural Language Processing
New Problems•   Data Completeness•   Definition of Friendship•   Modeling negative Links•   Emergence of new Models of    ...
Research Directions• Exposure Maximization• Analysis of Animosity• Contagion Coordination and  Competition• Rewards as Rei...
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Introduction to Social Network Influence and Contagion
Upcoming SlideShare
Loading in...5
×

Introduction to Social Network Influence and Contagion

142

Published on

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
142
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
3
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide
  • For the rest of her friends would either be scared to risk going to jail, as it was quite an oppresive regime. Or not know that there was even an uprising. The Mass Media wouldnt give it a time of day or maybe even show a cute human-interest story about a mom loitering around EDSA Shrine. Needless to say, all these people will go to jail.
  • But let’s say that we fast forward Marcos to now, and have my mom post a status update in facebook, letting her 500 friends, most of which are acquaintances, Since Dunbar’s number is only at 150-230. However, similar minded individuals caught fire with just my mom, and empathized with her, which caused them to join her cause.
  • Not these four people, may have affected someone else who are easily empathic to the cause of small groups. They may not be in any of their close relationships. So they join the “uprising”
  • And since they have gathered in a considerable amount of support, other people chimes in.
  • Reaching a tipping point, a deluge of people gathers in EDSA shrine to oust the Marcos Regime.
  • Mass Media took notice, and covers the story, broadening the reach.
  • Soon enough, we may have recreate the glory of EDSA not by a yellow cladded wife of a dead influencial senator. But by a simple housewife from Dipolog.
  • President Ben Ali
  • But let’s say that those three people, werent as easily affected by the plight of a mom, they would not have converted to the behavior she introduced and would keep on doing what they were diong. The tipping point is not reached, and the uprising would have also been a failure. Such are the delicate dynamics of Social Network Influence and Contagions. And this is also apparent in some less activistic examples like.
  • Buying a fax machine.
  • Or getting a new job. This and the previous examples have a “cost” that has to be taken into consideration, that is where thresholds come about. Some instances, cost is not really an issue, like...
  • The emergence of viral videos.
  • Explain Intrinsic RewardExplain SocialRewardExplain the GraphExplain z’ and z’’ and p* and their relationExplain the movement to and fro z’ z’’
  • AnatoleRapoport - If two people in a social network have a friend in common, then there is an increased likelihood that they will become friends themselves at some point in the future
  • Kleinberg - Clusters are the only obstacle for cascades
  • Granovetter – New Employment found in Acquaintances
  • Vocal Minority - Mustafaraj, et alA social Network is comprised of the active members and lurkers.
  • Studies only consider the reward of having K people adopt to a behavior before you. However, they did not consider the effect of having people adopt the behavior because of you.
  • Leskovec et al
  • Bayes Theorem – used in modeling uncertainty in networks. A classic example in showing cascades involves a jar of balls comprised of blue and red balls. We say to a class that the jar contains either a majority of blue OR a majority of red balls. We have a class that will one by one grab a ball from the jar. Upon picking a ball, they will not show the class the ball they have picked BUT they will announce their guess to the class.The uncertainty behind the guessing of the jar, given the information/guesses received by previous players is frameworked using Bayes’ theorem.
  • Data CompletenessDue to the massive nature of the social networks, and ongoing issues on privacy, researchers are met with the problem of incomplete data, wherein the information needed to fully assess persons in a study are not available.Researchers of Social Network Analysis have resorted to creating a much controlled means of data gathering, specifically on creating applications in Facebook and asking for participants within that network. Information are not as details as one would have in a formal research interview, and links in between people may not exist in the online social network but is apparent in real life. A new issue arises in that the process goes back to the analog, style of data gathering, with smaller samples and even affected by local biases.Definition of FriendshipAs online social network research has only been fully blown only recently, modelling other aspects of friendship and general relationships has been modeled only in its most basic idea. Some research consider a relationship existing between two people if they have talked only once. This can be related to the issue of the incompleteness in data.Modeling Negative LinksNatural Language Processing techniques has only been recently integrated with Online Social Network Analysis as a way to assess user similarities within a network (homophily). However, we have yet to analyze negative relationships within a network. In order to represent a social network effectively, one must also take into enemies into consideration. The problem with this idea is that in social networks, the concept of friends automatically removes any enemy even though connected people consider themselves as enemies. Furthermore, enemies within the social network may not even be friends, thus analyzing the possibility of animosity between users is hard.Emergence of new Models of Friendship and RelationshipNot only does the idea of social network evolved due to the recent advances in technology, but also the types of relationships and friendships comprising a network. Pyramid or Network marketing gives rise to the higher rewards received by a predecessor to have his “children” conform to the behavior (of buying in to the product). Shares and Liking schemes present a new way to reward the sharer with generally no cost. Anonimityin social networks give rise to a more nuanced and chaotic view of friendship and relationship, wherein no formal ways to model links or influences within a network, but a distributed concept of leadership.
  • Exposure MaximizationAn important question to pose for word-of-mouth marketing is how to maximize the number of people exposed to the information they want to spread, given that their reach is only to the nearest nodes around them. There already has been research on inferring the exposure curve, a curve that predicts the probability of “infection” from the time the user has been exposed to the behavior.Analysis of AnimosityTriads has been used to detect animosity towards social networks. However, this is only apparent on anonymous social networks. Literature posits that as with strong links, if a person views another person negatively, his friends with strong link also has a high probability of viewing this person negatively. Contagion Coordination and ContagionContagions tend to have a cooperative or competitive nature. The problem with recent studies is that they only study information or behavioral contagions in a microscopic point-of-view, taking it exclusively from the rest of the existing contagions. However, a study by Leskovec shows that in some cases, when a contagion is related with another contagion (for example, a news about a sports team winning, and at the same time a news about a member of that sports team getting marries) starts concurrently.Network Rewards as Reinforcement for Behavior RepetitionThis is my targeted Thesis topic, the idea behind this is that in considering future susceptibility to spread a new behavior, an individual is also affected by the reactions or adoptions of people around him. A good example scenario would be on the behavior of a pyramid marketing scheme. A node is more likely to join a pyramid marketing scheme if a considerable fraction of his neighbors join in as well. However, upon joining, he will be rewarded if the behavior is spread further down his links. Making finding new recruits, despite the cost of possibly alienating relations, a viable behavior.Another example would be the susceptibility of people to post links in their respective social networks. If the user was rewarded by his neighbors through likes and reposts, he might be more susceptible to share more links of the same nature in his network. Making him a proper target for early adopters or initial exposures.
  • Introduction to Social Network Influence and Contagion

    1. 1. Intrinsic Reward andSocial Reward Cost Fraction of the population
    2. 2. Threshold and Cascades
    3. 3. Threshold and Cascades
    4. 4. Threshold and Cascades
    5. 5. Threshold and CascadesLinear Threshold Model 3 2 3 2 4 3
    6. 6. Threshold and CascadesIndependent Cascades Model 0.7 0.5 0.4 0.1 0.2
    7. 7. Valid Points• Triadic Closures• Clusters force members to conform to prevailing behaviors.• Weak links introduces new information but not a change in behavior.
    8. 8. Valid PointsTriadic Closure
    9. 9. Valid PointsClusters on emerging Behavior
    10. 10. Valid PointsWeak Ties
    11. 11. Bias• Vocal minority• Time of day is ignored• Social Network bias• Forward Rewards
    12. 12. BiasTime of Day is Ignored
    13. 13. BiasSocial Network Bias
    14. 14. BiasForward Rewards
    15. 15. Recent Studies• Purchase Susceptibility and Quantifying Trust• Maximizing Information Diffusion within a Social Network
    16. 16. Recent StudiesPurchasing Susceptibility• E-commerce websites.• User to user communication upon purchase important.• Triads take a role in the sale of items or seller discovery• Relationships are integral part in the decision making process
    17. 17. Recent StudiesMaximizing Diffusion• Given a social network, how should we choose a set S so that we will be able to maximize behavior adoption?• Modeling after the Cover Set problem• Use of approximation via Hill Climbing
    18. 18. Methods applied• Bayes theorem for information cascades• Games for modelling cost and reward within a social network• Natural Language Processing in assessing Homophily
    19. 19. Methods AppliedGames p* = r(z)f(z)
    20. 20. Methods AppliedNatural Language Processing
    21. 21. New Problems• Data Completeness• Definition of Friendship• Modeling negative Links• Emergence of new Models of Friendship / Relationship
    22. 22. Research Directions• Exposure Maximization• Analysis of Animosity• Contagion Coordination and Competition• Rewards as Reinforcement for Behavior Repetition
    1. Gostou de algum slide específico?

      Recortar slides é uma maneira fácil de colecionar informações para acessar mais tarde.

    ×