How to create conspiracies in the age of machine learning?
UnBox Festival 2019
Machine Learning, Workshops, Co-Creation

With digital access increasing for all, a main source of news for the masses is the internet — social media like Facebook, Instagram, Snapchat, Whatsapp and content intermediaries like 9gag, 4chan and reddit are only a few digital spaces to be named. While such platforms open the gateway for communication in ways not possible prior to the digital age, they also leave much room for error and falsehood i.e. conspiracy theories, fake news, hate speech, power and political manipulations, etc.

Especially with the rise of image-based content — take memes for example — the power of persuasion has found its largest audience, no longer entirely bounded by socioeconomic classes, literacy levels, amongst other factors. Such content spreads information quickly, not only through the little effort required for its assembly, but also through the influence of ‘seductive’ participatory and collaborative efforts. Further, the meme-format of information, often humorous and prevalence, adds a frame of normalcy to the idea which it represents— thus, acting as an ever-growing expansive echo-chamber, especially for those who are susceptible, and especially for those who seek out such affirmation of their beliefs.

We hosted a workshop at the UnBox Festival 2019 called Conspiracy Creation to start conversations around this topic. To prepare content for the workshop, I collected all the headlines from The Hindu, from all of 2018. I then used a simple LSTM machine learning algorithm to learn from this corpus of data to generate new headlines. Participants were given these headlines, and asked to prove these headlines. They split up into groups, and started to create speculative worlds to prove that these headlines could be real. The groups then proceeded to create memes, or videos as an artefact of their new speculative worlds.

Team: Hugo Pilate, Natalia