COVID-19 Echo Chambers

The story I tell of echo chambers during the confinement period of COVID-19 is a story we all know, all too well. The majority of my audience will know the effects as seen on their YouTube, Facebook, and Google search feeds: the more you search for and click on videos, the more you see brothers, sisters, and long-lost cousins of those videos everywhere you open a search browser. The more discerning amongst my audience will be familiar with the cause: machine-learning based preference prediction algorithms that dictate user interfacing patterns. But only a select few will know how these algorithms are constructed. That is where the danger lies: our confusion on the internet lies beyond comprehensibility of ordinary techniques.

I, along with many other researchers and reporters, wrote on the effects of echo chambers in the outcome of the 2016 presidential campaign and elections. The 2016 aftermath on the internet was similar to the COVID-19 outbreak in that massive amounts of information, repeating sensationalistic stories were circulated en masse, some false, some true, most anxiety-inducing. It is a natural human tendency to feel anxiety when confronted with massive overloads of information, especially when the lines between right and wrong, good and bad, and truth and untruth become obscurer. It is the work of “data for good” to combat the negative effects of these search algorithms on mental health and anxiety to as to reduce information into digestible, factually correct, diverse, and non-repugnant forms (not guaranteed). Diversity in information sources is in fact the most important factor in your feed. An interesting analogy on why this is important is taken from machine learning: a computer system learns best from a diverse and large set of training data, humans learn the same way! Machine learning algorithms, ironically, work to reduce the diversity of this information. It’s as if machines are sabotaging the human learning experience to make us dumber, through entrapment in our echo chambers. If we are to talk about ethics in AI, this should be of the utmost concern.