The ease with which anyone can create online content for free, especially on social media, has led to superabundance of information being one of the defining characteristics of today's communication systems. This situation has resulted in increasingly intense competition for attention, which has become a scarce good. The researchers from the Complex Systems group (CoSIN3) at the UOC's Internet Interdisciplinary Institute (IN3) María José Palazzi and Albert Solé --professor at the Faculty of Computer Science, Multimedia and Telecommunications?--, led by Javier Borge, have participated in the design of an ecology-inspired mathematical model that makes it possible to break down and predict interaction patterns in a system as complex as the Twitter social network.
The model, published in the open access journal Nature Communications, is fundamentally based on two variables: the mutualistic (beneficial for both parties) relationship between users and hashtags, and competition for visibility, mirroring the situation of natural ecosystems with limited resources. According to the authors, this ecological framework "proposes a new and alternative way of understanding how Twitter works and can also be applied to other social media and communication ecosystems with similar characteristics."
The researchers considered various phenomena that went viral over the last nine years. One of the events was the 2012 UEFA European Football Championship, from which they recovered almost four million tweets from more than 1.3 million users, who used nearly 150,000 hashtags, from 19 June to 4 July 2012. The UOC research team also studied Twitter-based communication during the 2014 Hong Kong protests. From these demonstrations, they studied more than 800,000 tweets from almost 240,000 users, who wrote more than 30,000 possible hashtags from 27 September to 7 October. Another happening they analysed was the April 2015 Nepal earthquake, taking into consideration almost two million tweets from more than 810,000 users and contemplating more than 35,000 potential hashtags from 8 to 14 May of that year.
Parallels to the collaboration of flowers and pollinating insects
For decades, mathematical models have been applied to the fields of ecology and complex networks with the aim of describing the behaviour of natural systems to predict aspects such as the evolution of the abundance of species. Observing the behaviour of Twitter, the authors of the article identified similarities between some of these models and the characteristics of interactions on this social network. "Our intuition told us that Twitter users, understood in the abstract, compete for a limited resource (attention) in the same way as pollinating insects like bees compete for nectar. Hashtags, words and memes also compete to be the most used, in a similar way to how plants use their scents and colours to passively compete for the attention of insects," explained the authors.
Specifically, the new study adapts the type of mathematical models that have been used for more than 50 years in ecology to study natural mutualistic ecosystems (those in which species benefit from each other) to Twitter. "When a user chooses a hashtag, both agents benefit: the user because they believe it adequately expresses their desires and by using it they will obtain more attention, and the hashtag passively because it will be disseminated to more users, therefore reproducing the mutualistic relationship found between pollinators and plants. Our hypothesis was that if Twitter worked in a similar way to these ecosystems, we should be able to identify a certain correspondence and be able to predict the patterns according to which the social network is organized," they explained.
Two behaviour patterns: state of rest and collective attention
The results show that based on minimum ingredients (competition, mutual benefit and visibility maximization), this model makes it possible to capture and predict what actually happens. According to the researchers, Twitter has two basic patterns: when attention is fragmented, the system is structured like "a modular network, that is, organized into different groups according to the interests of the users around certain themed hashtags." But when there is an exceptional or viral event, which may be any extraordinary news story, such as elections, an earthquake or a TV show, "every user turns their attention to that phenomenon and themed communities disappear, entering what we have termed the nested state." In these cases, discussion centres around a small group of users who generate and use a large number of hashtags that are adopted by practically all the participants in the network. Once interest in the event wanes, the system returns to its normal modular condition: the state of rest.
One of the key aspects of this approach is that it is a simple model, given that with very few parameters it is capable of capturing the fundamental ingredients that drive the emerging patterns observed in Twitter and, moreover, it is "neutral" with respect to the users. That is, "the model does not need to assume anything about people's motives, biases or moods or about hashtag formats. The model's sole assumption is that both the users and the hashtags are in alignment with a subject of preference, which is why it works regardless of the communication event being analysed," stressed the researchers.
A model that can be adapted to other social media
The researchers indicate that this new ecological approach opens the door to modelling other social media and communication systems, as long as "there is competition for attention, through words or even images, as would be the case of Instagram." In this sense, the team that participated in the study aims to continue investigating this framework and is considering future lines of research, such as the possibility of using these models to intervene in communication events. "In the same way that ecologists use their models to try to intervene in ecosystems to, for example, prevent the extinction of a certain species, our idea is to conduct theoretical research into the conditions under which these communication events gain strength or wither away with a view to possible future intervention. For example, to make certain pernicious conversations or hashtags, such as those produced in fake news bubbles, disappear," concluded the authors.