We browse as we move: returners and explorers in the cyberspace

Ours is a life of perpetuum movement. We do not move only physically, however, walking on the streets of our crowded cities or flying the ocean on a fast plane. We move also in a more abstract way, especially when we enter the limitless virtual universe defined by the Internet. Maybe you are now idly sitting on a comfortable sofa with a smart phone in the hands, but your mind moves continuously in the cyberspace, jumping from a website to a social media application, from a virtual friend to a pirate movie website. Presumably you were on Twitter when you found a link to this post, maybe you will read another blog post on bigdatatales.com, then jump to my personal website lucapappalardo.com or randomly fly to a completely different domain, such as nytimes.com. Whatever your abstract virtual path, your mind moves, without any limit in space and speed.

What is really surprising is that the patterns of abstract mobility are similar to the patterns of physical mobility. We browse the infinite cyberspace as we move on the finite space. This is what a team of data scientists, including my colleagues Hugo S. Barbosa and Ronaldo Menezes, discovered in a recent scientific article. From the previous post you know that movements define your mobility profile: you are a returner or an explorer, no middle ground. If you are a returner you limit your mobility to a few frequent locations, if you are an explorer you live the mobility space in a more erratic way. Well, Hugo and Ronaldo’s team discovered that the same applies to your abstract mobility: you are a returner or an explorer, no middle cyberground.

Defining an Abstract Mobility Space

In the real world we move on a physical mobility space: in the streets of a town, in the corridors of a university, from the kitchen to the bedroom in a house. There are precise coordinates in the real world to quantify the length of our displacements, how long we travel every day, how fast we move. The Web, in contrast, is shapeless. How can we define the coordinate of a web page, and the length of a virtual movement?

Hugo and Ronaldo’s team think that, after all, abstract mobility is not so different from physical mobility. Every time you move from an Internet domain to another, you are performing a virtual, abstract movement. So the path bigdatatales.com –> lucapappalardo.com –> nytimes.com defines a movement in an abstract mobility space. Following all the virtual displacements you made on the browser, your virtual road graph can be reconstructed, a map describing your movements in the cyberspace. This is your abstract mobility network, and you can see an example in the Figure below. The Chrome Visual History app provided by Google provides a way to visualize your abstract mobility network.

Abstract Mobility network of an individual generated from the virtual trajectory S = [s1, s3, s4, s2, s1, s3, s1, s5, s1]. Nodes in the network (circles) represent Internet domains, edges (lines) represent virtual movements between Internet domains. The size of nodes is proportional to the visitation frequency of a domain, the size of edges is proportional to the frequency of the virtual movement.

From the abstract mobility network, Hugo and Ronaldo’s team reconstructed an abstract mobility space using a force-directed layout algorithm, which arranges the nodes of the network based on how frequent they are visited in sequence. The nodes of the network are placed on a virtual space according to the weights of the edges connecting them: domains that you frequently visit in sequence receive coordinates so that they are close in the abstract space; domains that are not visited in sequence are far in the abstract mobility space. When the algorithm is applied on the abstract mobility network, every node receive a pair of virtual coordinates. Once we have coordinates we have a reference system, and we can compute distances, displacements and mobility habits of individuals in the cyberspace. In other words, we have an abstract mobility space, a road map of the cyberspace emerged from the Web browsing displacements (see an example in the Figure below).

A visualization of the top 423 hosts generated based on the browsing behaviors and and visitation frequencies (high resolution available on-line http://my.fit.edu/ ̃hbarbosafilh2011/webmap.png). Label sizes represent the PageRank of the vertices in the abstract mobility network.
A visualization of the top 423 hosts generated based on a dataset of Web browsing behaviors and visitation frequencies. Label sizes represent the relevance of the vertices in the abstract mobility network. You see that domains that are usually visited in sequence, such as google.com and mail.google.com, are close in the abstract mobility space. In contrast, domains that are unlikely to be visited in sequence, like amazon.com and youtube.com are far in the abstract space.

Returners and Explorers reloaded

Starting from a dataset describing the 4-years-long Web browsing history of 524 users, Hugo and Ronaldo’s team reconstructed the abstract mobility space for every individual. Then, they synthesized each individual’s abstract mobility by computing her abstract mobility radius, that is the characteristic abstract distance traveled by the individual during the four years. A small abstract mobility radius means that the individual typically travels within short abstract distances, a high mobility radius indicates an individual covering large abstract distances in the cyberspace. Following the methodology we proposed in the previous post, they compare the recurrent abstract mobility radius of an individual (computed on her routine abstract displacements) with her total abstract mobility radius (computed on all her abstract displacements). A surprising result emerged: as for physical mobility, people naturally split into two well-separated groups, returners and explorers, with rather different abstract mobility characteristics. Returners’ recurrent abstract mobility radius is very similar to the total abstract mobility radius: their mobility in the cyberspace can be reduced to the mere displacements they do between a few preferred domains (Facebook and Google for example). In contrast, explorers’ recurrent abstract mobility is just a small fraction of their overall abstract mobility: they do not reduce their mobility to a predefined routine, presenting a high degree of exploration of domains in the cyberspace.

Abstract mobility hence resembles physical mobility, a surprising result suggesting that the tendency to return or explore can be something universal, embedded in our brain. Can other complex systems, like social networks, music listening, or book reading, be characterized by the returners/explorers dichotomy? I think yes, and this is what I am trying to discover in a collaboration with Hugo and Ronaldo’s team.

Did you find the same pattern in another context, or would you like to investigate that? Do not hesitate to contact us (see contacts below), we will be happy to collaborate with you and explore deeply this exciting phenomenon. As scientist, after all, we all are cultural explorers.

Article by Luca Pappalardo

This post refers to the following publication:

Hugo S. Barbosa, Fernando B. de Lima Neto, Alexandre Evsukoff, Ronaldo Menezes,
Returners and Explorers dichotomy in Web browsing behavior — a human mobility approach,
Complex Networks VII, Volume 644 of the series Studies in Computational Intelligence, pp. 173-184.

You can download the paper here.

Please, contact us if you investigate (or would like to investigate) aspects related to the returners and explorers dichotomy:
Luca Pappalardo: lpappalardo @ di.unipi.it





Be First to Comment

Leave a Reply

Your email address will not be published. Required fields are marked *