A team of Stanford University researchers has developed an AI that predicts human behavior by reading modern fiction from the online writing community Wattpad.
Augur, the AI developed by Stanford University researchers led by Ethan Fast, learned about human behaviours by analysing 1.8 billion words of modern fiction from Wattpad.
In the paper entitled “Augur: Mining Human Behaviors from Fiction to Power Interactive Systems,” Fast and associates wrote that “it is possible to create a broad knowledge base of human behavior by text mining a large dataset of modern fiction.”
Contrary to popular notion that fictional writing contains a false record of humankind, Fast and associates said, “Fictional human lives provide surprisingly accurate accounts of real human activities.”
By analyzing more than a billion and a half words from Wattpad, the Stanford researchers found predictable human behaviors such as blushing after being complimented; turning on the lights after entering a room; and not answering a phone when in a meeting.
Fast and associates said that they will integrate the information gathered by Augur “into the kinds of activity trackers and machine learning models that developers already use.” The Stanford researchers added that they plan to create a broader suite of Augur applications and test these applications in the wild.
This is not the first time that AIs have been trained to predict human behaviors by reading books.
Mark Zuckerberg announced that Facebook’s very own AI has been trained to predict missing words from reading children’s books that are freely available on Project Gutenberg.
In the paper entitled “The Goldilocks Principle: Reading Children’s Books with Explicit Memory Representations,” Facebook’s AI research team led by Felix Hill wrote that existing AIs are already excellent predictors of prepositions and verbs. AIs, however, fall behind humans when it comes to predicting nouns, Hill and associates said.
“Using children’s books guarantees a clear narrative structure, which can make the role of context more salient,” Hill and associates added.