Need Some Chaos in User Profile Algorithms

Razeeb Mahmood
An Attempt at Writing
3 min readJul 11, 2018
Containing Chaos by Michael Lang

Because of how current user profile modeling work we are getting more and more siloed to our biases. It affects what we watch, what we listen, what we read etc. Though its human nature to side with things we agree with we should experience a little chaos in terms of what we expose ourselves to. Otherwise we will stay stuck in a constant feedback loop that doesn’t really help us experience different things and grow. Whether we agree with these things or not.

This is how current “User Profile” algorithms work (parts in blue). Need to add the part in orange, or at least have the option.

Music

I enjoy listening to music from The Killers, Snow Patrol, One Republic, Coldplay — they are considered Alternative Rock. I use Spotify and when I listen to them it thinks based on my play history that’s my “taste”. Now when I use their Discovery Weekly (their machine driven suggestions) feature it mostly feeds me Alternative Rock music. All it’s doing is querying what a lot of users are listening to that I haven’t, that fit my “taste” and just builds me a playlist.

Usually this is great, I get to discover a lot of good bands and songs I wouldn’t have discovered otherwise. But it is also bad because Spotify has now limited my exposure to mostly my “taste”. Unless I start listening to some Backstreet Boys, NSYNC or One Direction, Boy Band music is not popping up anytime soon in my Discovery Weekly.

Not ideal because I shouldn’t have to alter my habits to then alter algorithms that a company like Spotify relies on to feed me new music. Spotify should expose me to good music from genres I haven’t listened to much, or at all. Challenge is not to send me music I already listen to, challenge is to send me music I haven’t listen to but may.

News

This is probably the most important area where chaos is needed. Most people get their news from social media. It’s not ideal but we mostly also happen to be friends with people online and offline with whom with we share similar ideologies. This creates a big problem as we get stuck in an echo chamber. If you are friends with people with similar ideologies and you share and like news with similar ideologies you are in an echo chamber that constantly validates your confirmation bias.

Facebook, Twitter even YouTube delivers you content that fits your “User Profile” this way. So if you are a liberal you will get fed liberal news, conservative, conservative news. Animal lover? Lots of cute videos from The Dodo.

Now what if in our timeline we received news from the opposite side from time to time. Just enough to challenge us to at least hear their voice. I always wanted Facebook to test this idea. To have maybe a slider that moves from very liberal to very conservative. And based on what is selected our timelines would show respective news and posts. But I also believe platforms like Facebook should built into their current algorithms some news intentionally intended to challenge our biases, things that don’t necessarily fit our “User Profile”.

Video

Same problems exists in platforms like Netflix and YouTube. What you watch dictates what gets recommended to you, which then leads to what you watch, feedback loop. I will be writing another post related to Netflix discovery sometime soon hopefully but essentially the idea is to recommend me content I wouldn’t necessary find otherwise. Just because I enjoy murder mysteries don’t think that’s my only “taste”, I mean seriously my friends who has seen my account probably think I am a psycho. Or just because I watched some episodes of Gossip Girls (to see lovely Blake Lively) please don’t think I enjoy Young Adult shows.

Suggest content that we haven’t seen or that doesn’t fit our current “taste”, but is good and we should probably give a try. Help us expand our “User Profile” by continuously adding some chaos to the loop.

--

--