Digital Smarts - Machines Learn Language Biases

A recent study at Princeton University found computer software that uses algorithms to learn human language can also pick up on prejudices among words that even humans may fail to recognize. Researchers used nearly 1 trillion words taken from the internet and tested word association using machine-learning algorithms – those that are learned by example. They found that machines ended up mimicking many human biases after finding word connections to genders or traditional masculine and feminine traits.