Author: Hany Farid
Publisher: TEDx Talks
Publication Year: 2018
Summary: The following video discusses how predictive algorithms are widespread and have ethical issues. He gave one study example where random internet people were given prompts and asked to guess recidivism and they had nearly the same results as the algorithm, and though neither the people nor the algorithm knew the race, they were both making the same racially-biased predictions. A problem arises because we are biased and trust algorithms too much. For example, a judge would listen if a prosecutor said “the latest in machine learning algorithms predicted he would commit another crime” but if they said “a random man on the internet read a 1-paragraph summary and predicted he would commit another crime,” the judge likely would not care at all. Yet their predictive power might not be that different. It is dangerous to put all of our faith in algorithms that haven’t earned it.