Monday, October 05, 2015

Recommendations 4.0

Have you ever gotten a recommendation from Netflix or Amazon or Spotify for a book or movie or song that you didn’t think you would like, but took a chance and really liked it?  Did you think, “Wow, Netflix knows me better than I know myself!” Or at least their data science/collaborative filtering/content filtering algorithm does. 

This really shouldn’t be surprising.  Algorithms like Amazon’s product recommendation and Pandora’s music genome project can consider thousands of variables, millions of users’ behaviors, and millions of products to find just the right ones for you at that time and place.  If the algorithm is designed well and implemented with the right UX of course.  There are many disasters as well.  But we create disasters of our own too – I can think of a few movies I really wanted to see that turned out to be horrible. 

Now jump ten years into the future when the targeting gets an order of magnitude better.  It adds emotion monitoring (which has actually just been launched as an app by Affectiva), activity monitoring (through your fitness tracker), and whatever else gets invented in the near future. 

This would still not be perfect, but might be reliably better than we are ourselves. Enough that we trust it to be as good as we would be, and more convenient because we don’t have to bother with the search, sort, and choose process. 

Now add another step.  Skip the middle man and let those companies actually make the decision for us. If we read a book a week, Amazon can simple decide what we are most likely to enjoy and ship it. We log on to Netflix and the queue is full of TV shows and movies that it knows we will enjoy.

If that works so well, why not broaden our thinking?  The food vendor knows just what we feel like eating for dinner based on our past eating history, taste preferences, activities that day, current stress levels, and nutritional needs. So as we are getting home from work the drone is dropping off the ingredients we need (or for those who don’t like to cook, the takeout drone delivers dinner ready made).  On stressful days we get our favorite comfort food. On workout days we get a balanced protein/veggie/carb combo.  You get the idea.

Michael Sandel at Harvard gave me an even better application. The frequency of bad dates and the high divorce rate prove that we are not particularly good at finding romantic partners. Maybe Match or eHarmony’s algorithms will get good enough that we can absolve ourselves of that responsibility as well.  Instead of suggesting dates, they will jump right to arranging marriages. 

That has kind of a “full circle” vibe to it, doesn’t it?