AI and Social Sciences – The Facebook Experience

A friend of mine just announced he is closing his Facebook account. He is not the first one; he won’t be the last one either. The recent Facebook data breach has rattled many social media visitors. However, I am surprised at the diverse groups of professionals that were caught off guard by this news break. It seems they missed the implicit understanding that every time you visit, post or ‘Like’ something on social media, that information gets stored and is used to profile and build your persona. In short, you (the visitor) are an observation in this massive big-data lake. It so happens that Facebook got caught in this crossfire, but it could very well have been a Twitter, Instagram, or other such visitor interactive forums.

Now, looking at this data breach from the underlying analytics point of view, it appears that some data science organizations have been crawling through Facebook users’ data, gathering personal information and using it to build predictive models about human behavior in political arena. But these ‘observations’ are not some inanimate numbers or phenomena and therefore should have been dealt with differently. Data Science community seems to have missed this vital insight.

The field of Artificial Intelligence / Machine Learning evolved predominantly in the pure sciences arenas of information technology, medical, space, manufacturing etc. It has been making significant inroads into areas that deal with consumer behavior like finance, telecom, retail, insurance, and healthcare by carefully avoiding the pit-falls of identifiable data. But when it entered the social media space, such caution seems to have not been exercised.  Humans are social animals that demand privacy and hence the rules of engagement when dealing with human subjects should have been different from those in pure sciences. Unless this distinction is clearly understood, such scandals will continue to recur in future. It would be like stitching an AI solution for pure sciences and forcing  it on a social science mannequin.

 In order to generate accurate predictions while protecting human privacy, the fields of AI and ML must embrace their social science siblings and provide holistic integrated solutions.  Social sciences (Economics, Political Science, History, Sociology) study human and consumer behavior, and tell us why we humans behave the way we do. In fact they go a step further and capture the randomness or uncertainty in human behavior as well, all while carefully respecting the sanctity of privacy. Sociology tells why our society is what it is, History is politics-of-the-past, and Economics is an important outcome of politics. Social sciences do have an array of tools, theorems and techniques to understand, quantify and provide insights of human behaviors, and have the capability to predict and forecast them as well. Further, thanks to the 1940s ‘makeover’, Economics is heavily quantitative, thereby rendering itself very well for advanced analytical methodologies. Its somewhat slower sibling, Politics, is just entering the phase of a quantitative makeover. Unfortunately it found itself mildly bruised with these social media breaches. We just saw one such bruise called Cambridge Analytica.