Se esta mandando tu mensaje...

Case Study

5 ways we’d like to deepen our relationship with algorithms

In our previous piece, we outlined some typical ways of interacting with recommendation algorithms on video and music services like Netflix, Spotify, and YouTube, based on the results of a small study we conducted. What implications does that have for content distributors, web and app designers, and other services employing these algorithms? Below we make some suggestions about how algorithms could step up their game and really win our affection.
 
 
#1 Let me talk to you (and talk to me back)
 
People want to be able to give algorithms more direct feedback on their recommendations. One user suggested,
 

“I wish there was an option in which the algorithm asks if it’s doing a good job giving you recommendations. And if it’s not, it can ask you some questions … so that the algorithm can understand you better… I feel like this is a lot easier and less frustrating than going out of your way to click on shows you like in hope that the algorithm adapts to your interests.”

 
Algorithms should have more straightforward ways to elicit and respond to feedback. Say you watch a horror movie one night with some friends, but you don’t actually like that genre. In most instances now, there’s no easy way to tell an algorithm to remove that behavior from your history and adjust your recommendations accordingly. Ideally users could avoid having to engage in a content navigation balancing act to “fix” their algorithms:
 

“Whenever I watch something weird that I normally wouldn’t and see that my recommendations have been suffused with related content I’ll watch more of the videos I usually watch in order to bring things ‘back to normal,’ so to speak.”

 
#2 Let me expand (or contract) my content horizon
 
Algorithms operate on the basis of similarity. They recommend content that is by some degree similar to what we’ve already consumed. For the Netflix newbie (or any new user on a content provider’s site) this isn’t a problem, because all the shows seem fresh and exciting. However, more experienced users may find themselves looking at the same set of suggestions every time they log on. One user lamented,
 

“Recommendation algorithms should exist to broaden your horizons, not narrow them. Too often the balance is shifted toward only displaying recommended content without an easy way to search for content using more generic attributes… Recommendation engines should not imagine you as a passive consumer.”

 
What if there were a way, instead of relying on an algorithm with the same set of rules, to either shrink or expand its recommendation scope? Users should have the option to calibrate their algorithms either more narrowly (suggesting content very similar to their consumption history) or broadly (suggesting content only tangentially related to what they’ve watched before). We picture a sliding scale, that on one end would result in more familiar recommendations, and on the other end, less familiar recommendations.
 
This additional flexibility might encourage users to engage with more diverse content – and satisfy more adventurous users who are constantly looking to broaden their horizons.
 
 
#3 Let me get to know you better
 
They know what content we will like before we like it – yet, we have no idea how they work. We are eager to understand more of the recommendation process – not necessarily the nitty gritty technical stuff, but the basic building blocks of what goes into each suggestion. One user offered:
 

“More transparency could be key to recommendation algorithms for me. For example, access to a disclosure or report that gives a window into why certain types of music were suggested based on previous listening choices. The clarity and transparency … tailored to my unique listening history, would be extremely helpful.”

 
We envision a drop-down option that would show which of our previous choices led the algorithm to make its suggestions. This kind of toggle would help us understand the magic behind the recommendation process and boost trust and user satisfaction.
 
 
#4 Let me change you based on how I’m feeling (current mood: hungry for reco’s)
 

“I listen to very different music when I work out versus when I am studying or hanging out with friends – yet nonetheless, I would love recommendations in all these categories.”

 
While algorithms make useful content suggestions, their ability to parse and separate those recommendations is lacking. Most folks aren’t interested in the same genre all the time. You may have different preferences on weekdays and weekends or depending on your mood.
 
Separating recommended content into different genres or moods would improve user experience and allow for more specific suggestions within individual topic areas. Here’s a straightforward example from a Spotify fan:
 

“If I like both rap and alternative music, I want the algo to suggest me both rap songs and alternative songs, not songs that are the ‘average’ of rap and alternative.”

 
#5 Let me split my personality
 
Many people called for more categorized recommendations, but also for more flexibility within their user profile. The fundamental problem with one comprehensive algorithm is that our tastes vary, and there is no way of curating multiple algorithms. One user wrote,
 

“I purposely ‘like’ or interact with algos so that they’ll learn my tastes. Unfortunately, my tastes change frequently and are quite diverse, so this results in a lot of ‘interaction’ so to speak.”

 
An improved interface might include the option to add sub-profiles to your Netflix or Spotify account – each with its own distinct tastes and “personality” – in which unique and independent algorithms would operate. This would streamline the experience for those with highly variable consumption patterns.