Identifying and Tracking the Spread of Memes in Social Networks

The Defense Advanced Research Projects Agency (DARPA) has put out a request for proposals for software/algorithms that would be able to identify and track in social networks like facebook — or the internet in general — memes of interest to the US government including the military services. (Search for “Social Media in Strategic Communication” or the SMISC program.) The algorithms should be able to identify communities of people with the same interests; it should be able to identify propaganda efforts of adversaries, etc. They actually used the word “Memes”! To do this they will need a working definition of meme — a definition that can be implemented in software. This definition must be in the form of a mathematical algorithm. I have developed such a definition. This definition identifies and defines a meme and automatically, over time, tracks its evolution. The algorithm is easy to implement but it might require excessive computational resources in its raw form. This is probably true of many algorithms that aim to work over all the information on the internet, phone network, etc. They won’t want to identify every possible meme. They couldn’t anyway since the number is way too big. The request gave one example of the sort of thing they want to identify and track and that was rumors on a social media network of the location of some important person (probably Mubarak) where the rumors led to spreading the meme to storm that location. This was all detected by human observers who then spread contrary information about this person’s location, and storming of his location was averted.

So they want (at first) to identify and track the spread of memes relevant to military and police activities. They want memes having to do with violence and threats of violence. Of course, later, the threats of violence can be abstracted so far away from actual violence that political opponents could become adversaries. Or anyone not adhering to a specific body of beliefs — an ideology — could become an adversary in the minds of the adherents of that ideology. This is likely an inherent problem with such algorithms. They can be used for other purposes than those for which they were originally designed. And they cannot be easily controlled because these algorithms can be developed by anyone with just a little of the right mathematical background. 

In addition to determining who has heard or expressed a meme of interest, it will be necessary to assess the operational status of the meme in the mind of the individual. Having a meme in your mind is one thing; doing something with it is another. How strong is the belief in the meme. In general (but not always) the stronger the belief, the more likely the meme leads the individual to actions. The amount and type of emotion and feeling associated with a meme is some measure of the importance of the meme to the individual and thus also some measure of its probability of leading to action. Here is one place where it gets complicated. We can see the emotion (bodily movements) in face to face (more completely, body to body) interactions. Picking up typed words off the internet leaves all these direct observations out; the only emotions we can pick up are those that are expressed by typed words (forget emoticons; they aren’t used enough to make a difference), and of course typed words can express emotions — otherwise novels wouldn’t work. But most internet users are not describing their emotions as in a novel. And real people go through different moods. Some people have predominant moods that are there almost all the time: Some people are shy or subdued, others are agitated with anger close to the surface almost always, and so on. And of course anger about one thing spills over into anger about subsequent things. So estimating the amount and type of emotion associated with the expression of a meme by an individual at a particular time and place via typed words on the internet is not easy. So the measures of the type and amount of emotion and their corresponding internal feelings will be approximate. And how all this contributes to making a decision to actually do something is not known in detail.  But part of the content of a meme in a mind is the amounts and types of emotion it has associated with it in that person’s mind.  The outside informational content of a meme and the inside emotional/feelings content define the meme instance in that person’s mind. We can imagine a multidimensional metric space of memes. The types and amounts of the various emotions would be a multidimensional subspace; the subspace of outside informational content would be also multidimensional but would be defined via a linguistic semantic analysis.

A meme can be defined as a radial category of its instances in a population of people. All the instances of a meme would be within a certain distance from the category founder and/or within a small distance from another instance. Memes defined by such categories would also form a metric space which would evolve over time. Thus we could get the velocity of evolution for a particular meme.

A community is a group of people communicating and cooperating together for one or more goals. We can define communities in terms of the memes of the members of that community. It is much more than a group of people having memes in common. There must also be structural memes that keep the community together and cause it to continue.

The government would like proposals to develop the details of all this. Such algorithms and software would be useful for people in general for the same purposes the government says it wants them, namely, to detect propaganda, to detect and track incitements to violence, and to effectively counter these memetically. These tools will be useful for anyone wishing to spread their beliefs, really any beliefs — memes — throughout a population. Think targeted advertising. So others besides the government will develop such tools, if they haven’t done it already.

What does this mean for the revolution most people will want? The understanding provided by such tools will also help our progressive revolution to be more effective, since what we want is to spread the memes for our goals and methods: nonviolence, human necessities for every human person, nondestructive and supportive interactions with our earthly living and nonliving physical environment — social justice.

And it suggests that some people in the government are beginning to see the value of memetic science. And they are going to pay people to work on it. As scientific investigation of memetics grows, the appreciation of the value of our memetic nonviolent revolution will also grow. Although the value of nonviolence is appreciated by some people, most people are still trapped by the idea that force and violence are the only ways to get effective long-lasting change in individuals and societies. They do not understand what force does, or what they think it does, or what they want it to do, which is to change memes in people’s minds in order to change their behavior. Memes determine behavior. Those who use force don’t see this step in the process. They simply think force changes behavior. Through the scientific investigation of memetics, people will understand that lasting behavior change requires prior meme change, and they will understand all the various ways in which people change their operational memes, and then they will realize that force, coercion, threats, and violence are only a few — and not very effective — ways to get people to change their memes, to change their minds. They will see that our nonviolent revolution is not only possible, but is a much better way, and that a revolution must be nonviolent if its goals include getting to a mostly non-coercive, non-violent society.