CMN 150V study

Wait! How much do some top celebrities get paid per tweet?

US$ 10,000 or more

What does the "two-step flow model" say? Information flows:

from the media, to influencers, to the public

What are the two complementary aspects of Computational Social Science Prof. Lamberson from UCLA will talk about?

empirical work + computer simulations

The only truly influential people in society are celebrities and famous people, like Justin Bieber, Barack Obama, and Kim Kardashian.

False

What trick did Prob. Lamberson use in order to track who re-posted a URL-link to a story on Twitter?

He tracked short URL, which are unique to the post

What are 'Twitter cascades'?

An information diffusion process on Titter in which a number of people make the same decision of passing along information in a sequential fashion

What is the first thing researchers found when looking at the empirical evidence about Tweets that go viral?

The vast majority of posts never get retweeted, but a small fraction of links go viral

Does this sound familiar? In a data science framework, what are the first and second part of the data refer to?

Training set & test set

If it turned out that one of the top-25 most retweeted URLs came from an account that had little influence in the past, and has a small number of followers, where would it be located in this graph?

Bottom left

If it turned out that one of the top-25 most retweeted URLs came from an account that had MID-LEVEL influence in the past, and has a small number of followers, where would it be located in this graph?

Mid-bottom

Let's assume that if you have this disease, you will surely have these symptoms. Now, 1 in 6 million people have this disease. Let's assume that 1/3 of the population has these symptoms. What's your chance of having the disease, given that you have the symptoms?

1 in 2 million

Sticking to our previous example, we said that the probability p(disease | given the | symptoms) was 1/2,000,000 = 0.00005 %. What had we assumed is the probability p(symptoms | given the | disease)?

100 %

If something went viral, looking at it, there is a high probability that it has been sent by an influencer: p(influencer | viral) What about p(viral | influencer), the probability that something goes viral, given that it was sent by an influencer?

Low probability

The proposed model consists of the following:

Take a hypothetical network, select some nodes, and simulate a contagion process by assuming that neighboring nodes have a fixed probability of getting infected

The horizontal x-axis presents:

If a given network (of the simulated networks) has few or many links

What is the lesson learned here? Whether someone is influential depends on:

the general structure of the network

Even tough we cannot predict who is influential, why is it still worthwhile for companies to pay large amounts of money to celebrities to send out messages?

Because chances are that among large number of people reached by them, some turn out to be influential, who will then influence others

What is the difference between this lecture on Social Network Analysis, and the previous lecture on the same topic?

Today we will look into dynamically evolving networks + start to simulate theoretical networks

Closeness centrality is calculated as:

the sum of the length of the shortest paths between the node and all other nodes in the graph

What is one common way to scientifically test whether there's something special about your network?

You create a large number of random networks, and compare your network with it

Multiplex network:

entities are connected to each other via multiple types of connections

Look at this pseudo "theory" (on the left) and then at the "hypothesis" on the right. Can you see how the scientific hypothesis makes an informal verbal idea more concrete? What aspects are equivalent in this translation? (check all that apply)

"who communicate" = "communication network"

"more friend" = "higher degree of centrality"

" have....friend" = "friendship network"

If you found that your network is not (statistically) different from the random networks you created, what would the conclusion be? (check all
that apply)

It is likely luck of the draw if I find something special in the network. I would have found it in any randomly drawn up network of that kind.

I can claim that my network is just another random network

I cannot claim that there's anything special about my network

Let's assume a very simple network with three nodes (A, B, C) and one (undirected) link: G(n,M) = G(3,1). How many different networks can you form with that?

3

Wait! I thought you can do 3 graphs with 3 nodes and 1 link! What's the connection here?

All possible G(3,2) become all G(3,1), when exchanging the "missing link" with the "existing links"

Wait again! What was the difference between the "numerical solution" and the "analytical solution"?

For numerical solutions, you enumerate the options and basically count, for analytical solutions you use math to derive the results

In network analysis, a component is:

a part of the network in which a path can get you from a node to any other node

How many of the 50 people are in the "giant component" at this point (in the largest connected subgraph)?

3

Wait! What happened here?

two connected groups of 4 merged into one giant component of 8: 8/50 = 16% of the nodes are now part of the giant component

Why do you need at least 1 connection per node in order for the giant component to dominate?

everybody can have one friend (on average), making a chain of friends

In network analysis, a cycle is:

a walk that ends where it began

Who will accumulate more links over time?

the people who were there first

What does "preferential attachment" mean?

the probability of a node to connect with new nodes, corresponds to the number of existing degree of a node

What does preferential attachment say in this case? The probability that a new node:

connects to each one of them equally likely

What does it mean that something is distributed according to a power-law?

exponentially few, have exponentially much , and exponentially many, have exponentially little

Informally speaking, triadic closure / clustering coefficient assesses:

what fractions of my friends are friends

Why are most of my friends connected to highly connected people?

popular people have many connection, inevitable also including my friend

What is the stylized factual concept about the number of degrees of separation in social systems?

six degree of separation

One of the characteristics of a small world network is a high clustering coefficient. How is that calculated again? =>

with the number of closed triangles

When you find a high clustering coefficient, it means that people group together in different densely connected clusters, what does that mean?

many people hang with their group, with little connection with other groups

What does it mean to have a small average path length?

the shortest path between all nodes is short, on average

A small world network is characterized by:

a small average path length + a high clustering coefficient

When social scientists say that societies consists of small world networks, what do they mean?

people have both close connections in tight groups and quick access to everybody

What has benefits and also costs?

creating and maintaining links with other

What is often meant when social scientists say that a dynamic social configuration has become stable?

nobody could benefit from changing anything in the configuration

If there is almost no cost to create and maintain connections, what network will evolve?

a clique

If there is a very high cost to create and maintain connections, what network will evolve?

a "network" without any links

What is the average path length in a star network configuration (assume MANY nodes around a single hub)?

almost 2

What is the aggregate net benefit for the entire network?

4*0.6=2.4

That's an important point (it will influence the entire efficiency and stability result): why do (according to our model assumptions) indirect connections have benefits but no costs?

the lower left node still needs to maintain one connection (to the upper left node), and maintain connections are to assumed to have costs

OK, wait: so the direct benefit of connecting to another node directly is 0.8 (let's say 80%). Now the indirect benefit, of going through another node to a third node (two steps of separation) is what?

0.8*0.8(80% of 80%)=0.64 (64%)

Why do we get "six benefits" and "six costs"?

there six degrees (incoming/outgoing arrows)

How many indirect benefits do we get now?

6, two for each pair

How many indirect benefits do we have? How can nodes benefit by indirect paths, reaching other nodes through intermediaries?

......mmmmmmhhh... none of these previous two, as there are also indirect costs of two degrees of separation, but we didn't yet talk about how to go about that...?

Has the social efficiency (the total sum of the efficiency of all nodes/agents) increased or decreased with this last reconfiguration of the network?

Decreased

According to our assumptions (see previous lecture: https://www.playposit.com/play/781527/UCCSS_SNA2_08:-Growing-Efficient-Networks-(18min) ), what is the net cost-benefit of the middle node?

3*0.8 (three direct benefits) - 3*0.2(three direct costs) = 3*0.6 =1.8

According to our assumptions (see previous lecture: https://www.playposit.com/play/781527/UCCSS_SNA2_08:-Growing-Efficient-Networks-(18min) ), what is the net cost-benefit of the pendant node?

1*0.8(direct benefit) - 1*0.2(direct cost) + 2*0.8^2 (two indirect benefits) = 1.88

In the previous lecture (https://www.playposit.com/play/781527/UCCSS_SNA2_08:-Growing-Efficient-Networks-(18min) ), we said that this star-configuration is the most efficient network structure for our network. However, besides all economic efficiency, what is the danger we just discovered? (check all that apply)

the network might not be stable, as some nodes might want to leave or change

the total social net efficiency is not fairly distributed among all members of society

the network is not equilibrium, as some node(s) could improve their personal situation

since direct links are more closely than indirect links, some nodes more benefit than others

Following the logic outlined here, if some members could still increase their status by changing the network, what could be done to stabilize the existing network structure? (check all that apply)

redistribute resources until everybody is equally well off

use subsides balance costs and benefits among different members of the network

take away from the ones that have too much and give it to the sones that have too little

What was our definition of "social stability"?

nobody can improve the personal standing by modifying the existing social structure

What was our definition of "social efficiency"?

the overall net-benefit is maximized for the entire network

What was the characteristic of a scale-free network that is grown with a preferential attachment logic?

exponentially few nodes have exponentially many links, and exponentially many links, and exponentially many nodes have exponentially few links( power-law)

Why do you need less steps to diffuse something in a scale-free network, compared with a random network?

Hubs can reach many nodes very quickly

What vaccination strategy would you suggest in a scale-free preferential attachment network?

Vaccinate the catalyzing central hubs

If you ask some random members of a network to name a friend, is this friend more likely very popular or not?

on average, when asking many people, a popular people will be named more often

Check all correct statements:

With lots of hares, the number of lynx can increase

With lots of lynx, the number of hares will decrease

With few hares, the number of lynx will decrease

With few hares, the number of lynx will decrease

With few lynx, the number of hares will increase

What do models NOT help us with

They describe the world exactly as it is

Check all rules followed by flocking birds in the model:

avoid collision

be where most others are

go where most others go

What are the two main assumptions of Axelrod's "culture model"

People interact more likely with others who are similar

Interactions cause one individual to change to match the other

If "1" stands for liking and "0" for disliking, and the first cultural feature represents "R&B" music, the second "dreadlocks", the third the "49ers", and the fourth "Facebook": what could describe person C:

Liking R&B, bald, Dallas Cowboy fan, just posted a status update

The percentage of similarity defines the likelihood of interaction. If a person likes cats, Top Ramen, ML, and cat memes; and you happen to like dogs, Top Ramen, SNA and cat memes, what is like likelihood that you will interact?

50 %

What does the color of these boundaries mean?

How similar people are to each other

What will we explore today? How we can:

create theoretical models of societies that might not even exist in reality

What is a model?

An idealized abstraction that represents something of interest

What is something always involved in modeling?

Omitting details and deciding which aspects to leave out

If you find two completely different models that claim to model the same phenomena:

both of them can be useful, representing different aspects

As we have just seen with the example of the London Tube map, one ambition of modeling is often:

to be able to grasp the general gist behind an aspect of reality

The key takeaway from the anecdote from Feynman is that scientific modeling usually:

involves a bi-directional process between human thoughts and formal thinking structures (equations, concepts, code, etc)

It is important to choose the right language to describe something. In Computational Social Science, we will use:

Computer code

Prof. Dawn Sumner from UC Davis Earth & Planetary Sciences (https://dysumner.faculty.ucdavis.edu/ ) was involved in determining the location to land the Mars Rover Curiosity. What kind of "language / tool" did she use to extend her thoughts about this problem?

Virtual Reality, where code runs models of Mars

What is part of the model set up here? (check all that apply)

We randomly distribute a certain number of agents on a grid

About half of them are of one type, the other half of another

Unhappy agents move randomly to a free spot

What is the social result of this random movement once every individual is happy?

Society as a whole organized in 'neighborhoods' of reds, and 'neighborhoods' of green

Schelling's segregation model is an example of social emergence, since the outcome on the social level, is different than the motives of individual people.

True

What do we find here?

Seems like unhappiness and similarity are stuck at their level

What happened when we moved from individuals having the ambition of having 26% of similar neighbors, down to 25%?

Segregation on the societal level dropped a lot, much more than the 1% move on the individual level. The relationship between the individual and society is "non-linear".

What happened when we moved from individuals having the ambition of having 75% of similar neighbors to 76%?

The outcome was qualitatively different on a societal level.

What does it mean that a model finds an equilibrium?

Opposing forces or influences are balanced and nothing changes anymore

What could contribute to the tendency that a society of quite tolerant people becomes more polarized? (check all that apply)

Schelling's segregation model shows that even quite tolerant individuals, end up in very homogeneous communities

In homogeneous communities, people mostly hear the aspects and arguments they already agree with

Mainly hearing what you agree with, strengthens your existing beliefs ("echo chamber")

The "other" rarely hears of the opinion of your group, and becomes less convinced that it matters

Segregated communities drift further and further apart, as they are rarely in contact with the "other"

What are some of the simplistic modelling assumptions in our version of Schelling's model? (check all that apply)

Caring about exactly 4 neighbors

Only consider those neighbors that live horizontally or vertically directly next to you

Having only two different kinds of people in the world, not more

Not being able to see where one moves next when moving around

Considering a larger circle of neighbors, will we end up with more or less segregation? Why?

Social science is difficult and not intuitive!!! ...dunno...

Considering a larger circle of neighbors, will we end up with more or less segregation? Why?

More segregated, since averages of neighbors involve more people, so one needs more homogeneous surroundings

So in modern societies, which count with wider social circles and have more mobility --according to this model-- the conclusion is that societies tend to:

be more segregated

By the way, what is it called when similar people tend to connect to similar others?

homophily

What is true about the NetLogo modelling environment (https://ccl.northwestern.edu/netlogo/ )? (check all that apply)

You can find different models in the "Models Library" under "File", ready to explore

One tab shows the visual modelling "Interface"

One tab shows the "Code" behind the model

One tab shows the "Info" that describes the model verbally

In NetLogo, what is a "patch" and what is a "tick"?

A "tick" is a unit of time, and a "patch" is a unit on the spatial grid

Where are the agents moving to?

toward the nearest unoccupied location with the most sugar within their range of vision

On Sugarscape, "wealth/income" is defined as the amount of sugar that agents collect. Why do we see these four different "income-groups" here?

Agents are located on 4 different sugar-levels of the Sugarscape

On Sugarscape, different agents have different types of long- and short-vision. What ability does better vision give an agent?

search for patches with more sugar in far off places (more effective)

On Sugarscape, different agents have different metabolisms. What ability does a faster metabolism give an agent?

burn less sugar and survive longer with the same amount of sugar (more efficient)

Why is the average vision of the society (as a whole) increasing and the average metabolism decreasing?

The society evolves and only the fittest survive