Machine learning enables early-warning signals to be recognised in the event of social instability
The Cyber-Defence Campus has developed new techniques of machine learning in cooperation with the start-up giotto.ai and ETH Lausanne. Thanks to these new techniques, early-warning signals for socially disruptive events such as riots, wars or revolutions can be recognised and identified in social media.
A new technology for recognising early-warning signals in the event of social instability has been developed. These warning signals are recognised and identified via social media.
Social networks are playing a major role in the general formation of opinions as well as the emergence of social movements. Socially disruptive events such as demonstrations, protests, revolutions, riots or even wars can have their beginnings in social media. Today, for example, it is recognised that social media played a leading role in the emergence of the Arab Spring in 2011. It is therefore of crucial importance to be able to recognise the emergence of such events early on.
For this reason, the Cyber-Defence Campus and the company giotto.ai have developed new techniques of machine learning to establish early-warning signals for recognising such events.
These new methods are innovative in many aspects. First of all, they are robust. This means that a recognition model trained on a particular data type will be able to recognise similar but not identical events. Secondly, only a small amount of training data is needed to achieve results that are at least as good, if not better, than is the case when using large language models. Thirdly, Deep Learning and its lack of explainability can be avoided through a geographical approach in conjunction with the use of traditional algorithms of machine learning. Fourthly, these methods do not require any access to personal data. Looking forward, it can be said that the methods used are promising, but need further tests and developments to be validated.