Scribd
Upload
36 views1 page

NecSys Pagan

Uploaded by

Wenjun Mei
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
36 views1 page

NecSys Pagan

Uploaded by

Wenjun Mei
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

Emergence of Zipf’s law among social networks influencers

Nicolò Pagan1,*, Wenjun Mei2, and Florian Dörfler3


1Social Computing Group, University of Zürich, 2Department of Mechanics and Engineering Science, Peking University,

3Automatic Control Lab, ETH Zürich

*nicolo.pagan@uzh.ch

1 Introduction

Despite the interdisciplinary effort in the study of (online) social networks


formation process, researchers have so far underestimated the role of the
User-Generated Content (UGC) and of the recommender systems (RS).
Today’s most popular online social media platforms, such as YouTube or
Instagram, diverged from traditional social networks, e.g., Facebook or
LinkedIn, by allowing (and even encouraging) users to connect to real-life
strangers. In response to this change, some users specialised in creating
even semi-professional content that can attract more and more followers to
the point that they can make a revenue out of that. This trend has deeply
influenced consumers’ and companies' behaviour: more than 70% of US
businesses engaged Instagram influencers to promote products in 2017
3 Results and Discussion
(www.emarketer.com). Given the potentially profound impacts of the UGC-
based online social platforms on, e.g., information spreading, it is of
paramount importance to understand the statistical features of these One of the most relevant aspects of social network analysis is the in-degree
networks, especially in relation to the most influential individuals and their distribution as it immediately quantifies the reach of the content generated
UGC. Here, we report some of the most relevant findings related to our by a user in an online social network. According to our theoretical and
recent publication [1] in which we proposed a simple yet predictive network numerical analysis, the in-degree distribution follows a power law (see
formation mechanism based on the quality of the UGC. figures on the left) for both uniform (above) and mixed-PA (below) meeting
process. Furthermore, a striking Zipf’s law emerges when plotting the
number of followers as a function of the quality-ranking (right plots).
2 Model

We consider the unweighted directed network among N ≥ 2 agents whose


UGC revolves around a specific common interest. Each actor i is endowed
with an attribute qi ∈ [0,1] , drawn from an arbitrary probability distribution,
which describes the average quality of i ’s content. We denote the directed
tie from i to j with aij ∈ {0,1}, where aij = 1 means i follows j, and we assume
that each agent i can only control her followees aij but not her followers aji .
We then consider a sequential dynamical process, starting from the empty
network, where at each time-step t ∈ {1,2,…} each actor i
(i) finds another distinct actor j , chosen randomly from a probability
distribution on {1,…, i − 1,i + 1,…, N}, and
(ii) decides whether to follow j or not.
Depending on the recommender system, the meeting probability (i) of
finding a node j might depend (linearly) on the current in-degree of j , djin(t) ,
in a way similar to the preferential attachment (PA) process:
djin(t) + 1
, with probability α,
ℙ [Mij] = ∑j≠i djin(t) +1
1
N−1
, with probability 1 − α,

where α ∈ [0,1] controls the amount of PA in the meeting process. Clearly, if


α = 0, the probability is uniform and independent on the network evolution,
Fig. Above: uniform meeting process (α = 0), below mixed-PA (α = 0.5). On the left, the average in-
conversely when α = 1, the probability is degree distributions and on the right the in-degree pdf as a function of the quality- ranking.
time-dependent and grows proportional
to the popularity of the target node.
The tie formation process (ii) depends 4 Conclusions
QUALITY

on the comparison between i ’s current


followees’ and j’s qualities, i.e.,
Today’s social media platforms such as Instagram or YouTube heavily rely
1, if qj > Vi(t),

{aij(t), otherwise,
aij(t + 1) = on their recommendation engines that suggest users new connections
based on the User-Generated Content (UGC). Inspired by a meritocratic
where the payoff function Vi(t) := max qj INDEGREE principle, we proposed a novel network formation model in which actors
j∈ℱout
i (t)
form ties based on the quality of the UGC. We found that, independently on
measures the maximum quality in i ’s the recommendation strategy, the in-degree follows a Zipf's law with
i (t) := {j, s.t. aij(t) = 1}.
followees set ℱout
Fig. An example of the resulting network. respect to the UGC quality-ranking. This work is based on [1].

Reference

[1] Pagan, N., Mei, W., Li, C., & Dörfler, F. (2021). A meritocratic network formation model for
the rise of social media influencers. Nature communications, 12(1), 1-12.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy