The Best of Both Worlds: Mixed Systems with ML and Humans in the Loop to Combat Fake Information

Abstract

Dealing adequately with misinformation is one of the societal challenges of our times, since misinformation has been proven to be harmful for people, societies, and democracy. Improving Artificial Intelligence algorithms underlying information retrieval and recommendation systems is a path that must be encouraged; however, this is not the only path ahead and, above all, it can be combined with other approaches. In addition to the known limitations of Machine Learning model results, which cannot be guaranteed to be 100% accurate, ethical issues are raised when an algorithm acts as a censor of what information a human being can and cannot access. This paper discusses some recent initiatives during the COVID-19 pandemic to improve the quality of information delivered to users that have two characteristics in common: firstly, they are technically simple, hard coded, and do not involve any AI; secondly, they represent a preliminary step in a broader perspective that goes beyond technical improvements to promote critical thinking among those receiving the information. Although they can be seen as preliminary cases of how to deal with misinformation, they seem to be effective and they point towards more interdisciplinary solutions to the contemporary issue of misinformation, possibly bringing other developments to ethical, Human-Centered AI.