Andrew Schwarts:"Towards AI-Based Language Analyses for Saving Lives"
Abstract: Your mobile device knows a lot about you and that may be in your best interest. For the first time in human history a substantial portion of our daily behaviors are being recorded. While this presents legitimate concern for exploitation, with care taken for privacy, security and transparency, this abundance of personal data is enabling more accurate and less obtrusive human assessment at scale.
Here, I will go over recent and on-going work toward accurate AI-based human assessment from social media and other daily language patterns. On the individual level, Facebook and Twitter have been found predictive of mental health diagnoses, cognitive impairment, personality, demographics, and occupational class (among others).
At the community-level, Twitter has been found predictive of flu and allergy outbreaks, life satisfaction, atherosclerotic heart disease mortality, health behavioral risk factors, excessive drinking rates, and even real estate price changes. hile these techniques have shown robust links over a plethora of important aspects of human life, the most beneficial applications may be concerned with better decision making rather than simply more accurate assessments.
I will discuss related work, analyzing the linguistic features of recommendations in an online political forecasting game, ultimately suggesting machines may be able to better suggest quality information for decision making than people themselves. Finally, I will bring these two works together to discuss a developing vision for a future where AI improves decisions and ultimately helps to save lives.