When it comes to AI, correlation does imply causation!

This is something new that I’m trying with this blog, instead of your usual 800-1000 words and in some instances, 4000 word blog posts, I’ll be doing a few relatively short (100-200 word) posts that I’ll update throughout the month.


I’ve spent the past few weeks since the end of the semester, tweaking/hacking certain parts of my life, and I noticed that regularly (> a month) updating this blog needed to be one of them. Writing a blog post pushes me to go further than I usually do when it comes to research, although I do a fair amount of research in my day to day, the added pressure of someone else reading my work acts as a powerful motivator to go further. Hopefully it’s something that I can keep doing when the semester starts up again.


So, to get to the point, here’s how the post’s are going to shape up- The topics are generally going to be about something that caught my eye in the past couple of days. I’m going to write a down a couple of sentences talking about the broad theme and then reference sources at the bottom that y’all can refer to for further reading. I’m not going to be delving down into the tranches to explore the whole concept in the quick read (I’m going refer to it in the title), but that doesn’t mean it won’t be a regular blog post. Maybe y’all could help me decide on what I should choose to write on based upon the traffic the quick reads get. So anyways, here it goes:

Actual post:

Anyone that’s taken a statistics class usually has it drilled into them that correlation doesn’t imply causation. But what if AI changed that? For example, a Chinese payday loan company called Smart Finance has tried to plug the massive gap in China’s credit score system by asking you information that goes above and beyond just your financial history. An AI algorithm judges the likeliness of you defaulting by gaining access to your phone and checking for metrics such as battery percentage, call logs, how long it takes for you to finish the application, how many times do you change something on the application etc etc. Metrics that seem to have no bearing (at least not to most humans) on the causality of you repaying the loan are being used by AI to judge whether or not you should get a loan, to the point where the company doesn’t even employ risk officers.

The lack of an effective credit system has led to an epic ballooning in China’s payday/short term loan business, with unpaid debts currently at $392 billion according to BCG (Boston consulting group). Smart finance swears by these metrics despite concerns that they’re just a ruse to gain more access to your personal information so you can be threatened in the case of delinquency. Whether these metrics really do count is above my pay grade and for AI researchers to handle, but it really ought to start up a conversation among techno optimists about the Frankensteinian dangers of AI in the wrong hands.

Further reading:

NYT- https://www.nytimes.com/2017/12/25/business/china-online-lending-debt.html



Amy Webb, The Big Nine-

Kai-Fu-Lee, AI Superpowers-

Further Listening:



Leave a Reply

Your email address will not be published. Required fields are marked *