From individual cognitive biases to collective drifts on on-line social networks – COBCOD

Nowadays, social networks are strongly changing our social habits and constitute a new public arena of prime importance. The algorithms ruling the social networks are protected by industrial property rights and are only partially disclosed. However, it is known that they seek to capture the attention of the users as long as possible for commercial purposes. Users are thus drowned in a massive flow of contents and face an unprecedented competition to attract the attention of their peers. This double stress hinders the possibility of a rational thinking, favouring quick and emotional behaviours, more likely to be affected by cognitive biases. Then, the interactions in the social networks tend to transform these individual biases into massive collective phenomena, like information bubbles and echo chambers, fast propagation of fake news, herding effect, etc. This situation alerted the regulator and the Digital Service Package recently adopted by the European Parliament aims at mitigating the harmful effects of social networks. Better understanding the dynamics of these collective phenomena, particularly those emerging from individual cognitive biases, is essential for continuously improving these regulations. With this objective, the COBCOD project involves researchers from physics, cognitive science, computer science, social-psychology and proposes a rigorous approach combining online experiments, theoretical and agent-based models, computer simulations and analysis of large scale data sets collected on the social networks.