User manipulation markers

Me and my colleagues were analyzing a large data set on the user behavior of a whole linguistic-cultural group the other day. We were identifying the changes in the priority objects in processes of gratifying the needs. Kinda like at first everyone wanted to eat and were looking at food pictures and now the majority wants sex, so they’re downloading porn).

We’ve found an interesting phenomenon but first I want to expand on the concept. Linguistic-cultural groups (simply put, the residents of the same country, fellow citizens) periodically change their priorities. This happens due to the fact that the conditions of their existence are changing. Plus there’s some internal dynamics where there are certain correlations between original objects of need/desire and the ones they were replaced with. The change of priorities has an impact on the network behavior. Something gets less readers, something gets more. The period of change can be both no more than a day and last for years.

If you want a somewhat simplified example, let’s take the large group and consider it as one person. When he wants to travel he reads about beautiful places and trips. When he has a newborn child he begins to read about parenting and education. Now we have a change of current need represented in the network behavior. And we are faced with the fact that we’ve found another reason to change current objects – a media pressure.

The kind of pressure when a certain subject, object or idea is being strongly cultivated in the information space. Back to our example: it is as if our user is strongly encouraged to read about some socialite’s ups and downs. And he adds to his parenting website clicks – in this case – some tabloid clicks regarding the news about those times she broke with her boyfriend, how’s her dog getting on, and what color her last party dress was. The persistent pressure will force our user to read about it sooner or later.

Since important part of the task was to separate the behavior caused by the objective environment factors and inner dynamics from the behavior caused by the media exposure, we started looking for a way to figure out the solution. And very soon we’ve found it. Turned out it was piece of cake.

The average visit depth rate of the website related to the objects meeting the requirements is much higher than that with the objects with artificial actualization. In other words – on average – country citizens read about the problems of saving and making money 2.3 pages per user, and they read on the adventures of socialites about 1.2 pages per user – regarding our data set of course. Surely, these figures depend on the period, data normalization, particular qualities of the original data set. But the double difference would remain.

Your results differ? Let’s discuss ;).