Fake news revealed through artificial intelligence


Institutional Communication Service

3 May 2018

Social media has become one of the main sources of news. As the past elections in the United States and the United Kingdom have shown, however, they have also become a platform for sharing forged information, or “fake news”. This is a problem that undermines the very heart of the democratic system, blurring the ability of citizens to distinguish between fiction and reality. 

Michael Bronstein, professor and researcher at the USI Institute of Computational Science (ICS), together with fellow ICS researchers Federico Monti and Dr. Davide Eynard, has devised and developed a new method – based on algorithms and artificial intelligence – that could prove to be the most effective solution to the spreading of fake news through the Internet. Instead of focusing on content, the new system analyses the processes that generate news items, in automated fashion, identifying models useful to distinguish real news from fake ones. The innovative approach is based on a new class of automatic learning algorithms, so-called “Geometric Deep Learning”, developed by the research group of Prof. Bronstein together with other researchers at EPFL Lausanne, NTU Nanyang Technology University Singapore, and Tel Aviv University. 

The basic research that led to this new technology, which is patented by USI Università della Svizzera italiana and which will be marketed by the spin-off company Fabula AI, has been funded by a prestigious ERC Consolidator Grant, two Google Research Faculty Awards, an Amazon Machine Learning Research Grant, and more recently by an ERC Proof of Concept Grant – the fourth one received by Bronstein during his tenure at USI. Michael Bronstein is quite familiar with the world of innovation and technology transfer, as Fabula AI is his fourth startup, following several other successful initiatives developed between the Silicon Valley, Lugano and Israel. 


ERC Proof of Concept Grants 

Proof of Concept grants, which are part of the EU research and innovation programme Horizon 2020, are funds issued by the European Research Council (ERC), for the purpose of allowing ERC-funded researchers to explore business opportunities, prepare patent applications or verify the practical viability of scientific concepts. The grants, worth €150.000 each, can be used by researchers to set up new companies, file patent applications and attract capital to make their research marketable. In the first of three rounds for the current year, ERC has awarded 50 Proof of Concept grants to researchers in 12 different countries. 


Amazon Machine Learning Research Awards

The Amazon Machine Learning Research Awards program (MLRA) supports researchers in the field of machine learning, for the development of innovative algorithms and source codes, in a wide variety of applications and areas of interest. The amount of the prize received by Prof. Bronstein is $150,000.