January 8, 2013 at 17:38Predicting informational epidemics by analyzing a social graph
At some point, one of the TED lectures stirred up the quiet haven of my perceived internet.
Description of the lecture from the site:
Attention.
To understand the further text, viewing the lecture is required.
It lasts only 18 minutes and has Russian subtitles (thanks to Nadezhda Lebedeva ).
The ability to predict the behavior of network actors exclusively by analyzing the distribution of content turned my mind! For whatever reason, I decided to repeat the experiment, even in artisanal conditions. Cooling my head and thinking, I had two complaints against the professors:
1) The behavior of biological viruses (influenza, T-Virus), if you can predict the distribution statistics, is a huge error, because in addition to the contacts of the actors, many other factors will influence the distribution, from the physical health of the characters involved to flashes in the sun.
2) The concept of the Paradox of Friendship (a random person calls his friend and in most cases will have more friends) sometimes seems completely absurd to me. For verification, I asked ten of my friends to name some friend at random, and the choice of only two of them was in agreement with the theory of the Paradox of Friendship.
To solve the first question, I selected a request for help as an information virus. A scooter was stolen from one of the sample participants. This information line was convenient in that the victim conducted searches using social networks and regularly reported on the results in them.
To solve the second question, instead of the Friendship Paradox theory, in order to identify the most popular users in the sample, I simply calculated the number of connections of each participant within the sample.
It is also necessary to mention that, in contrast to the original study, the VKontakte social network was chosen as the medium.
The sample was collected from 100 students from different faculties of one Moscow university.
The timeline is divided into 20 marks, where one mark is one day (the period from July 30 to August 18).
By “Moment of infection” is meant the first contact of the user with the information line (repost, commenting on the record, etc.).
All data collection and processing took place manually (I am a humanist).
The result of the dreary calculations was the table below.
(three of the most popular and not popular actors are shown)
Based on the data of the popularity of actors within the sample, I divided them into two groups:
Group A - the most popular 20 users
Group B - the remaining 80 users.
And the final result was a schedule.
Despite the fact that this research of mine confirmed the initial theory, it is necessary once again to recall that the experiment was carried out under conditions of artisanal conditions, and therefore it is impossible to rely on its results too much. But I myself firmly believed in the possibility of predicting information viruses by such methods.
I would be grateful for the advice and recommendations.
Description of the lecture from the site:
After mapping tangled social networks, Nicholas Christakis and his colleague James Fowler explored ways to use this information for good. And now Nicholas Kristakis will unveil his latest discovery : social networks can be used as the fastest method to detect the spread of any epidemic: from innovative ideas to socially dangerous behavior or viruses.
Attention.
To understand the further text, viewing the lecture is required.
It lasts only 18 minutes and has Russian subtitles (thanks to Nadezhda Lebedeva ).
The ability to predict the behavior of network actors exclusively by analyzing the distribution of content turned my mind! For whatever reason, I decided to repeat the experiment, even in artisanal conditions. Cooling my head and thinking, I had two complaints against the professors:
1) The behavior of biological viruses (influenza, T-Virus), if you can predict the distribution statistics, is a huge error, because in addition to the contacts of the actors, many other factors will influence the distribution, from the physical health of the characters involved to flashes in the sun.
2) The concept of the Paradox of Friendship (a random person calls his friend and in most cases will have more friends) sometimes seems completely absurd to me. For verification, I asked ten of my friends to name some friend at random, and the choice of only two of them was in agreement with the theory of the Paradox of Friendship.
To solve the first question, I selected a request for help as an information virus. A scooter was stolen from one of the sample participants. This information line was convenient in that the victim conducted searches using social networks and regularly reported on the results in them.
To solve the second question, instead of the Friendship Paradox theory, in order to identify the most popular users in the sample, I simply calculated the number of connections of each participant within the sample.
It is also necessary to mention that, in contrast to the original study, the VKontakte social network was chosen as the medium.
The sample was collected from 100 students from different faculties of one Moscow university.
The timeline is divided into 20 marks, where one mark is one day (the period from July 30 to August 18).
By “Moment of infection” is meant the first contact of the user with the information line (repost, commenting on the record, etc.).
All data collection and processing took place manually (I am a humanist).
The result of the dreary calculations was the table below.
(three of the most popular and not popular actors are shown)
Based on the data of the popularity of actors within the sample, I divided them into two groups:
Group A - the most popular 20 users
Group B - the remaining 80 users.
And the final result was a schedule.
Despite the fact that this research of mine confirmed the initial theory, it is necessary once again to recall that the experiment was carried out under conditions of artisanal conditions, and therefore it is impossible to rely on its results too much. But I myself firmly believed in the possibility of predicting information viruses by such methods.
I would be grateful for the advice and recommendations.