While the field of artificial intelligence has explored a number of ideas for computers to act like a human, there has not been an investigation into how to act like a specific human, a known human. What aspects of a specific person's habits, preferences, and idiosyncrasies are necessary to generate their online identity? ShareAid parses personal informatics data captured live, including their location history, photo stream, browsing history, and sleep and activity tracking data; it automatically proposes a social feed comprising a series of published posts, and is evaluated similar to the Turing Test's Imitation Game, by testing a person's ability to judge whether the social posting activity of someone they know (the poster) was made by the poster themselves, or by ShareAid. Eventually, ShareAid aims to simulate the Facebook timeline, Twitter posts, or Instagram feed made by a specific person in a way such that a poster can adopt it instead of writing posts by hand, to understand what constitutes an intelligent online personality.
Here is a breakdown of the different folders in this repository:
The photos arm of ShareAid recommends captioned Instagram posts from a user's camera roll based on what they have posted and written in the past.
The weblinks arm of ShareAid recommends captioned Facebook posts of shared links from other websites by crawling through a user's internet browsing history as well as the links they've shared in the past.
ShareAid is a tool being developed as part of the Human-Computer Interaction Lab overseen by Jeff Huang at Brown University. If you have any questions about ShareAid, please contact [email protected].