Don't Shoot the Algorithm!

As algorithms make more and more decisions, any accusation of bias or an outright error in the decision gets redirected at the algorithm. Seth Godin cites a few such examples:

-         That important mail that landed in the spam folder;

-         Who gets stopped at airports for extensive checks;

-         Google’s search results;

-         Facebook’s news feed.

As Godin says, this begins to sound more and more like “hiding behind the algorithm”. But why blame the algorithm “as if it wrote itself”? Isn’t it obvious, asks Godin, that “someone wrote that code”?

What happens next when AI (an algorithm) begins to write new algorithms?!

“As AI gets ever better at machine learning, we'll hear over and over that the output isn't anyone's responsibility, because, hey, the algorithm wrote the code.”

While there are obvious dangers with algorithms making decisions, isn’t it also true that algorithms would make decisions that humans won’t make due to irrational and/or moral reasons like political correctness? Do you really want the next plane you travel in to be blown up because of political correctness? Or would you rather that people who match the profile of terrorists be vetted thoroughly?

On many topics, I’d pick an algorithm over the decisions of the likes of Angela Merkel. That, of course, leads to the next questions which are very hard to answer: which are those topics? And even more important, who decides the list of those topics?

Or can we write an algorithm for that too?!