Connectivism & Connective Knowledge

This is a summary of Valdis Krebs’s presentation in CCK08 on Networks. You can see the PDF version of his slides at http://orgnet.com/EmergentNetworks.pdf

1. How are networks built?

Each of us have our own networks. They become friends and ‘combine’ networks, by virtue of their having friends in common. So network combination is not additive; some of his friends are some of my friends.

Thus we can create simple networks, such as the ‘pipe’ network, developed by David Krakow, of Carnegie-Mellon. We can use this map to identify network metrics.

Networks can also be ‘multiplex’ networks. People can be connected in various ways. Consider The Simpsons. There’s the family network (Homer-Marge-Bart-Lisa–Maggie-Grandpa), friends network (Bart-Lisa-Milhouse), and work network (Homer-Lenny-Carl).

‘Steps’ – my immediate friends is the first step. Then the friends of my friend is the second step. We can see how quickly a network grows, when we increase the number of steps.

6 degrees (6 steps) can connect people, but it’s just to strangers who are connected. This happens even at 4 degrees; they are often strangers. The connections of interest are all at the 1-degree and 2-degree level.

We know what’s going on with people connected at one degree, and can influence them. We know something about people who are connected by two steps, but things are missing. By three steps, they are mostly a mystery, and we have very little influence. And after four steps, they are just the general public.

2. Organizational Networks

Consider a typical hierarchy. We see the boss (CIO) at the top, the directors in the next level, and the managers further below. This is a typical way of looking at an organization.

But we can take same hierarchy and represent it as a hub-and-spoke network. They are basically the same thing. It’s just a difference in perspective.

On this diagram, we can also draw other connections – not reporting relation in a hierarchy, but the actual working relationships between people in the organization.

If we take this same picture and the same data, remove the hierarchical lines, and left the software display the network according to how people are connected, we get the *emergent* organization. We see ow people are more or less involved in getting things done.

3. Examples of How this is used

a. Expert location – we find the experts in a group of engineers. We draw an ‘asymmetric’ or ‘directed’ network – it shows who goes to whome for expertise or advice. People take a survey and we draw up the network.

This is similar to the way we draw the web. The web shows one person linking to another. This shows one person asking another (as reported in surveys). Google uses a system called ‘page-rank’ which is a very similar person, to see who is an expert. We look not only at one-degree ties but also two and three degree ties.

b. Communities of practice – this demonstrates natural clustering that occurs according to common interests, common goals (we colour the nodes according to clusters). Interestingly, the outside contractors in the various nodes, they are the ones who gt the different clusters talking to each other.

c. Key Opinion Leaders – we can see who are the key opinion leaders outside the organization. Eg., showing who doctors go to for advice before prescribing a new drug or treatment.

We may think doctors just read the reviews, but that’s not how things normally work. They usually talk to each other about this. So it’s very important to a pharmaceutical firm to know who are the key opinion leaders or trusted advisers. If that person begins to apply something, usually others will do so also.

4. Types of Data Collection

a. Email clusters – usually we just ask people, or look at who they link to. But now we want to look in more detail.

Eg email clusters – not reading the email, just who sent email to whom. Note that we can have more and less frequent interactions. The less frequent interactions are a weaker tie. We can see clustering as determined by frequency.

b. Project management – we used network analysis of a project. We coloured the nodes based on the department in the project. We cal see eg. the blues talk to the blues, etc. And so we can see where there is siloed communication.

We can also see the *bottlenecks* in the project communication. So we had to realign the communication.

c. Spread of Contagion – ‘contract racing data’ – a picture of the spread of TB, for example. Software is used to map outbreaks – TB, SARS, HIV-AIDS. The network shows where the disease can travel. Black nodes specify infectious and infected. Freen infected and not infectious. Etc.

You can see that some of the people are more infectious than others. The person with the most connections – the ’social butterfly’ – is also the most infectious.

Another example – HIV spread in the porn industry. The gray links showed where they appeared in movies together. You can quickly see who the originator of this was.

d. Players on teams. Steroids in professional sports – the players (in green) and the red (drug providers) shows you the supplier customer relationship. What happened was that everybody in the network assumed that everyone else in the network was using steriods.

5. Comparing Networks

Can you distinguish between terrorist networks and business networks? Not possible just from the pattern; what you need to know is who is involved. There is no difference between the two networks otherwise.

Eg. Hutton Inquiry – there was controversy around data about the Iraq war. David Kelly was accused of leaking the data, and was later found dead. The network of interactions produced a map. (Not sure what that solved, bu…)

6. Uses of Networks

a. Government Corruption Scandal – a network map can be used to illustrate the spread of government corruption. Local citizens have created a site called ‘map the mess’ that display the relationships being exposed in news articles about the corruption scandal.

Again, you see this is a multiplex network – family networks, friends, business ties, employment, favours, etc. As you did through all these public news sources, it’s amazing how deep a map you can get, and what a picture of these relationships you can get.

b. Animal networks - how diseases spread between barnyard animals. Eg. ’social’ networks among cows. How are cows connected? They express affection for each other by standing next to each other when eating or drinking. And that is also a possible way of passing a communicable disease one to another.

c. Biological networks – we can look at networks of very small things, eg., a network of proteins and protein interactions. Eg., if they send chemical signals back and forth to each other in some living entity.

Notice that this biological network doesn’t look like the human networks. It doesn’t have the same clustering component that some of the human networks have.

d. Discovery networks – eg., the network map of the iPod. We can see the people from different companies that provided some technology for the iPod. We can see how this industry network has expanded ovr time, with companies providing accessories, enhancements, competitors, etc., as well as the content providers. We can see who is working with whom, and what role the company is playing in the industry.

e. Economic Development - colours identify roles – green for academic, gray for government, blue for business, etc. We then map two networks, the arrows showing who provides whom with resources, and bidirectional showing who is working with whom on projects.

This map, eg. in Ohio, showed that the universities tended to work with each other, and showed which organizations were at the center of economic development.

f. Geographic networks – we can overlay networks on maps. It can show how an organizational network does or does not fit geographic layouts.

For example, the U.S. electrical grid. This map looks different from social networks – they are all man-made, designed, and engineered. But they do have some emergent properties, as the grid changes to adapt to new development patterns.

These networks are not made for resilience as much as they are made for efficiency. There are not as many multiple paths from one place to another. Before 9-11 that was fine. But in an age of terrorism, the build-for-efficiency model maybe doesn’t work as well any more.

Or, for example, an oil pipeline network. This is a very sparse network – you could tell where it is vulnerable just by looking at it. Again, there are built to be very efficient, but as a result are very vulnerable.

g. Book networks – we can see information about books on Amazon, including what other books were bought along with this books.

Political books, for example, are related to each other. We see two very distinct clusters. We can see ‘right wing’ and ‘left wing’ clusters. There were a few in the middle.

Never in the last five years have these two clusters merged together. There have been different books, but the clusters remain the same. This tells us that the readers’ views do not change very much.

h. Musical Styles Network – some kinds of music belong together and some don’t. Love songs and folk don’t fit. But soft rock and 80s hits do. This is based on interviews and surveys.

i. Linked web sites. Just follow the links. Example, websites from the non-profit world focused on a particular issue. Mapping this way is very easy and shows the relationships.

j. Autonomous systems on the internet, eg., ISPs. We can see how these are connected to each other – tells us how information is routed through the internet. We map the path or paths that a packet takes from one ISP to an other.

We can see here that the internet is a very string hub and spoke model. Some nodes has a lot more connectivity than others.

This has the properties of some of the scale-free networks people have been looking at. These have very different properties from the social networks.

k. Wiki networks. We mapped people working on a wiki while on different floors in a building. Typically, we would see different clusters for each floor. But when they’re working on the wiki, they are working together without regard to floor.

l. Blogging network. We looked at the blog network in the Cleveland area. Green nodes are outside the area, red nodes are inside the area. We saw most of the people blogging in the area were whining to others in the area.

m. Online network. When we look at networks, we normally look at the part that’s connected. People are connected, eg., if they’ve interacted with each other. In this network, they interact by comments on a web page, internal email, a chat window, or leave messages in a guest book. So we looked to se whether people had a two-way connection.

What was more interesting was when we looked at everybody, not just the connected people. We see the hub of connectivity in the center, then very small groups of 2-6 people separated from the rest of the community, and then a very large group (60 percent of the group) that as disconnected. These were like the lurkers, taking things in, not really interacting.

We’ve done other surveys of online communities and seen this pattern repeated. In some communities, the lurking piece of the network can be 80-90 percent of the total population.

Questions

Q. Was the terrorist map on the left or the right?

A. Right

Q. Why was the on on the right the terrorist network?

A. It’s really hard to tell them apart. They both have a core-periphery structure. What they both are, are project teams. They’re both trying to get something done. Since they’re both human, they’re both going to have the same shape. There’s nothing in the shape that is going to tell you.

Q. This course has the same sort of structure as the communities at the end – a centre group with a lot of dialogue, and the vast majority being individuals who are observing and not engaging. Is it important that this outer group be brought in, or is this just a healthy network.

A. There’s no requirement that this outer group be brought in, but it’s nice to know where they are, and who they are. Going back to the terrorists – there’s the actual terrorists, there’s a group around that may help them, and there’s a group who may be passive supporters. They won’t do anything, but they won’t tell authorities anything. I think this pattern permeates most of our networks.

Q. My first reaction was, oh we have to make everyone connected. But we need to remember that these other people may be well connected in their own spaces, they don’t need to be connected to this one.

Can anything be expressed as a networks. We saw a breadth here – from Simpsons to food to diseases to proteins to the internet? Is the network a defining building bl,ock?

A. I think that any complex system that has interactions and interdependencies can be expressed as a networks, with links that either non-directed or directed, and there are various social network analysis practices that can be used to analyze hem. This network process helps us understand things. This current crisis – it would be useful to have a picture that tells what’s going on. It would be interesting to see how the money flows, how the investment flows, how one is doing business with the other.

We can do a lot of analysis around network maps, there’s a lot of math around them. But they can also be used to start conversations. Just as a map helps you make sense of a strange city, these network maps help you understand the complex social milieu you are trying to help or are embedded in.

Q. the brain as a complex system. Sporn – seeing the brain as an ecology with connections between neurons. There is some understanding developing. Some of the attributes we are talking about here are present in the neural network as well.

So – to learning. Are you aware of work showing that how we know things are networked. Networked attributes among concepts, for example?

A. I come from the position that the network you’re embedded in greatly influences how you do things and what you do – and what you know. I have started saying, what you know depends on who you know, and vice versa.

For example, suppose two of us graduate from the same place, and you network with many more people than I do, eventually you will be better off and you will be performing better as an employee, because you will have access to better information. So, your connections are very important.

You have to have certain skills and intelligence, etc., but that’s not sufficient. You need to be able to connect, who to rely on, who to work with. And that’s not just who you know, but who knows you. Often, your success in a company depends on your visibility in a network.

So, I think that learning is social and learning is iterative, so those with a better network have the potential to learn better and more.