CMN 150V Final Study Guide questions and defined correct detailed answers|

Twitter networks, influencers, disease vs symptoms, cost/benefit/spread in networks - correct answers Session 6 Case Studies people in networks who can 'tap in' to communities of interest - correct answers Influencers and influencial people a famous person tweeting on behalf of a company to promote a product - correct answers Example of influencers/influencial people Influential people find out about trends and spread the word through social connections. - correct answers What is the Law of Few? They create cascades to get people on trends. - correct answers How do influential people contribute to trends according to the Law of Few? info -- media -- influencers -- public - correct answers Two-step flow model of influence influencers decide what should be consumed, then the public bases their decisions off of the influence - correct answers what happens in the two-step flow model of influence influencers are consistently popular, regular people randomly go viral - correct answers Difference between influencers and viral but regular people tracked shorted urls that were shared and reposted to see how links propogate through networks; the tracked links created twitter cascade models - correct answers Twitter study of cascade an information diffusion process in which a number of people make the same decision of passing along information in a sequential fashion - correct answers Twitter cascades split data into past and future 'past' data - model - predict - compare to 'future' data - correct answers Predicting influence through twitter reposts training set - correct answers 'past' data testing set - correct answers 'future' data number of follower and past influence predict one's influence - correct answers findings of twitter study outcome factors are statistically significant but poor fits to the model groups actually going viral meaning the ability to go viral must be because of an outside factor - correct answers what is wrong with the twitter study findings symptoms do not equal disease but disease does equal symptoms - correct answers googling symptoms study occurs when the likelihood of an event is inaccurately assessed by ignoring the general frequency of the event itself - correct answers base-rate fallacy P(A|B) = P(B|A)P(A)/P(B) - correct answers Baye's rule the probability of an event, based on prior knowledge of conditions that might be related to the event - correct answers what does baye's rule describe P (viral | influence) does not equal P (influence | viral) - correct answers Baye's rule with influencers P (disease | symptoms) does not equal P (symptoms | disease) - correct answers Baye's rule with disease create hypothetical networks, select nodes, simulate a contagion process by assuming that neighboring nodes have a fixed probability of getting infected - correct answers Computer simulations and social networks determines if/how things are absorbed and spread among people/institutions - correct answers Density of a network determines influences - correct answers Network Structure degree distributions - correct answers example of network structure influence - correct answers another word for social contagion how things are spread/diffuse in networks; infinite number of ways of distribution - correct answers social contagion/influence used to find out if there is anything special about the network or if it's just random - correct answers random networks test own networks against random ones to see if your hypothesis holds (uses induction and glass-of-red-wine theorizing) - correct answers hypothesis testing in random networks G (n, p) or G (n, M) - n nodes (form independent links) - p probability (for each node) - m independent links - correct answers random (erdos-renyi) graph benchmark to see what m (links) connect with what nodes - correct answers purpose of Erdos-renyi benchmark calculate average degree of network - correct answers Use G (n, M) for degree/links = avg. degree (simulate and count) - correct answers numerical solution for average degree of networks (n-) x p = avg. degree (mathematical derivation) - correct answers analytical solution for average degree of networks nodes connected in one group (not a clique) - correct answers component tipping points of emergence and change; non-linear phase transitions emerge - correct answers threshold functions all nodes are given and links are grown/added - correct answers nodes and links in erdos-renyi graphs will have one unique, giant component and cycles - correct answers network with p 1/n will be fully connected - correct answers network with p = 1n(n)/n growing networks following various rules - correct answers scale-free networks start with number of nodes fully connected and add new nodes with number links to existing nodes with equal likelihood; older nodes will have more likelihood - correct answers adding nodes with uniform likelihood in scale-free networks kids at highschool table and new kid comes and has equal probability of connecting with kids at table; kids at table have previous links cause they are not new - correct answers example of nodes with uniform likelihood some nodes have many more links than others (fat-tail distribution/power law preferencial attachement) - correct answers adding nodes with non-uniform likelihood in scale-free networks few nodes have many links and many nodes have few links - correct answers fat-tail distribution and power law the probability of a node to connect with new nodes corresponds to the number of existing degrees of a node; this is unique to scale-free networks - correct answers preferencial attachment at time (t) there will be t(m) links - correct answers new nodes adding m(links) to existing nodes t*m*2 - correct answers total number of degrees in a scale-free network degree of node(j)/ (2*t*m) - correct answers probability of attaching to node (j) in scale-free networks exponentially few have exponentially many and exponentially many have exponentially few - correct answers distribution according to a power-law new nodes creates a fraction of its links with other nodes uniformly at random; other fraction is created by looking at 'friends of friends' connected nodes - correct answers hybrid models = uniform models + preferencial attachment guesses what fraction of my friends are friends (the likelihood of triangles closing through connections) - correct answers clustering coefficient and triadic closure nodes with more connections are easier to find and make more connections - correct answers hybrid models and preferencial attachment they easily satisfy the mechanism you're after - correct answers how hybrid models are useful people have both close connections in tight groups and quick access to everyone - correct answers small world networks unlikely to quickly reach others since networks are mostly closed - correct answers networks with high levels of clustering likely to quickly reach all others on average - correct answers networks with small average path lengths the shortest path between all nodes is short on average - correct answers small average path length each node links to both neighbours (high clustering and long average path length) - correct answers small world networks and netlogo simulation no node can be added or take away for benefit - correct answers stable equilibrium has costs and benefits - correct answers creating and maintaining links most stable and efficient networks - correct answers hub and spoke/star networks when nobody benefits from changing anything - correct answers when a dynamic social configuration becomes stable clique will evolve - correct answers no cost to create or maintain connections a network will have no links - correct answers high cost to create or maintain connections benefit + benefit2 cost (complete network) - correct answers low cost network star networks with all agents - correct answers medium cost network benefit + (n-2)*benefit2 cost (empty network) - correct answers high cost network 2*(0.8-2)*0.2 = 1.6 - 0.4 = 1.2 - correct answers 2 benefit and 2 cost network direct links are more costly than indirect links - correct answers c