Deconstructing the Quora Newsfeed

Last week, I wrote a post called “The Thick Edge of Quora,” where I looked into the last decade of search and made the argument that Quora, through its topic ontology, could conceivable build a new type of search engine and set of structured, interrelated web pages that will work toward building a better, cleaner, more efficient web, while also allowing for discovery.

I should stress again that Quora has the potential to do this, but nothing is written in stone. The scope of their ambition is so big that it truly requires venture capital, both in terms of financing but also partners in company building.

While the future is bright and full of hope, there is a near-term concern. As a power user, I’ve spent hundreds of hours refining my approach to how I use the site, yet I realize I am an outlier and that for many casual users or early adopters the site can be difficult to get a handle on and even navigate, despite the fact it has quickly become an excellent and trusted resource for people in the technology field, broadly speaking.

Disclaimer: For those who haven’t used Quora at least moderately, this may not make sense, so I make no promises that this post will be of interest to you. If you aren’t a heavy user or don’t care about the site, please, I beg you, do not waste your time reading this. However, for those of you who have used Quora and are addicted like I am, grab a coffee, close your other tabs, and walk with me for a little while as I try to explain.

To be fair as an observer, I’ve been thinking about one small issue with the current Quora experience and how, despite its seemingly harmless size, it can potentially distort the experience for many users.

In the age of Twitter, real-time information flows to us in newsfeeds. The good aspects of this transition to feeds is that we have the power to customize the information that appears in our feeds by following specific brands, sources, and people. Furthermore, we can constantly update, add, or prune our sources to our tastes, and each network has an interesting mechanism to help surface new content from those we are associated with. In Facebook, “liking” pieces of content can inform the site’s newsfeed algorithm, “EdgeRank,” to help us discover relevant content. In Twitter, users can “retweet” original tweets to their followers, and because it’s an asymmetric network, content has the chance to spread across further node points.

Quora’s version of a “like” or “retweet” is a “vote.” On Quora, users can either up-vote answers or down-vote them, too. You can vote on multiple answers within a question thread. You can up-vote some and down-vote the others. As you might expect with crowd behavior, there’s a tendency for viewers to vote in packs, most often to help surface great content and/or evidence of someone famous using the site. There are occasions, however, where the crowd wants to silence a voice, and down-voted answers are pushed to the bottom of the thread and usually deemphasized by being “collapsed” under a soft, grey-colored heading that hides the original text.

In the parlance of a Google search, the most up-voted answers would appear at the top of the “results page,” while the collapsed answered are buried in, say, the fourth or fifth page of a Google search. In other words, very hard to find. Ranking answers on Quora are not determined by the raw number of votes, however. Instead, each Quora user has a certain reputation rank based on their profile, history of use, social graph, and authority within topics. This Quora algorithm is currently referred to as “PeopleRank.” The idea in theory is that if the most authoritative person answers or votes on an answer within a thread where he or she is known to possess superior knowledge and/or experience, that person’s contribution would be given more weight.

The logic behind PeopleRank makes perfect sense. As the system scales and more users come on board, there simply has to be some type of mechanism to help sort and rank answers. One of the most base level concerns of early Quora users was the fear that, as the system opened up, the quality of content would diminish. In theory, the mechanism of voting tied to an algorithm based on reputation is a good idea to ensure future quality on the site and that the best answers are at the top of the thread, or at the top of the “results.”

Quora has built the product tools for users to assist in the surfacing of new content, with the help of machines that learn over time. And while these machines do mostly help ensure the best content surface to the top of the thread, it’s the effect of voting that has the potential to mess up a user’s newsfeed.

In Quora, your newsfeed is filled with content based on the following variables:

• Questions you’ve “followed” or subscribed to;
• Answers submitted to questions by individuals that you follow;
• Questions that have been “followed” by those individuals you follow;
• Questions submitted by users you don’t follow but are tagged in context within topics you follow;
• User activity pertaining to follows, such as Person X followed a certain individual or topic; and
• Answers from people you do not follow to questions you do not follow, but that have been up-voted by someone you do follow.

Despite what looks like a messy list of variables, it’s pretty straight-forward, and once I figured it out, I took the time to tune my settings accordingly. It took about an hour or so, but if you’re interested, here’s what I did:

• I had to un-follow a bunch of people. I had to ask myself through this process — “Would I want to answer or follow a question from this person?” If yes, then I remained a follower. I suspect I’ll continue to prune this list. (I’m currently following ~350 people.)
• I also had to un-follow a bunch of topics. When I first joined the site, I tried to only follow topics, not people. I got excited by all the great topics to follow. As I approached 200+ topics, it got noisy, messing up my feed. I shaved this down to 52 topics followed today, and again, I’m going to shave that down, but I’d prefer to get back to 200 if I could.
• Instead of consuming information on Quora through the newsfeed it generated for me, I instead opted to ignore the newsfeed entirely and used the old “hunt and peck” method of following certain people and topics to keep abreast of any interesting activity. (More recently, the site has added “groups” where a user can toggle between main topics they follow on the right-hand side-bar, strangely similar to the various tabs a user can use to navigate on the left-hand side of their Facebook page.)

Despite these extensive Quora manicures, I am still learning how to navigate my own newsfeed and overall experience. And if I’m struggling with it as a Quora geek, I can only imagine how occasional or new users are processing the constant flow of (usually high quality) information.

And, there’s one variable that Quora doesn’t give users control over that I feel is the most intrusive to my overall experience: Voting.

When someone I follow on Quora up-votes an answer to a question, even if I don’t follow the question or follow the author of the question, that content appears in my newsfeed. I believe the intent behind this makes sense logically and, for the most part, those actually up-voting answers have good intentions to help promote content to the site that makes the overall experience better and ads knowledge to the system.

The effect of all the voting, however, has some deep implications for user experience today and the future of the site tomorrow.

I have written about this issue in pieces about my own personal view around voting in Quora. Basically, I don’t vote often. When someone votes on an answer, that answer also appears in the voter’s activity feed. Therefore, if someone is checking out my profile and it’s literred with random up-voted answers, what kind of signal does that send to the person viewing my profile? Furthermore, if I keep up-voting any answer, that isn’t very conscientious toward all the people that have graciously elected to follow me on the site.

The site also doesn’t tell me anything about another user’s voter behavior. While I can draw some inferences from a user’s profile about their follower/following counts, their ratio of questions asked to questions answered, and how much free time they have (through edits), there’s no mention of voting statistics. I’d like to know about a user’s voting history for two reasons: (1) if they vote too often, I may elect not to follow them out of fear of messing up my feed; and (2) if they don’t vote often, I’d know that a vote from them is perhaps a stronger signal.

Given these conditions, I choose to up-vote only occasionally with the hope that my vote could mean something more than just an easy click or a “like.” In lieu of voting, I simply send “thanks” to someone if they answer a question I posed or follow. That way, the author gets the satisfaction that someone has read and liked their answer and taken a bit of time to show appreciation. Sending thanks is a private signal within Quora that only shows up in the recipients notification stream. If I really love an answer, I will up-vote it and oftentimes broadcast it through Twitter.

All of this minutiae regarding voting begets another, more important, question for the site: Why is voting handled this way on Quora?

There is already precedent on other major social networks that allow their users more control over their newsfeed. In Facebook, for instance, you can group people into lists or mute certain individuals who are really loud without having the awkwardness of unfollowing them. In Twitter, the analogy is clearer. If you follow someone, that person’s retweets also show up in your stream (though not through list mode). If someone is a rabid retweeter and messing up my feed, I can simply go to your profile and un-check the little RT symbol, thereby quietly taking that privilege away from you.

The common thread in these two examples above is that these networks have features that enable users to control the signal from within their own newsfeed. Now, given how remarkably efficient and fast Quora’s underlying technology stack is, there’s no doubt in my mind that they could implement a feature like this in a very short period of time. I also have to believe that they’ve considered this feature and, for whatever reason, decided against it for now. One way to implement the feature would be to allow users to visit a Quora user’s profile page and disable the vote signal button, much like one can do in Twitter.

So, then, let’s assume Quora has thought about this element and for now, held off. Then, we must ask “Why?”

Here are a few possible reasons as to “why”:

1. Up-voting signals incorporated into the newsfeed helps users discover new content, users, topics, and questions. It creates an opportunity for discovery.
2. Up-voting helps Quora begin to map out not only which users are interested in certain topics, but also which users are perceived as more knowledgeable as it builds its PeopleRank algorithm. Therefore, sending these signals into your feed help signal to the user which users may be perceived as expert by others.
3. Viewing an up-vote in the user’s feed may encourage that user to engage with the question and offer a new answer.

All this voting is good and well-intentioned, but I think it’s a dangerous, double-edged sword: on one hand, up-voting builds a community within Quora and around topics, and helps people and ideas get discovered, but on the other hand, each vote ends being an editorial act, where someone I follow explicitly says to me that “hey, this is good content,” or “hey, this is a good answer to the question.” In some cases, this is true, but in many cases, it is not true — and that takes away from my experience.

Q: Why did Semil spend all this time and blog space writing these thoughts down in detail?

A: If Quora is going to continue to grow (and I believe it certainly will), I am not sure that the current newsfeed mechanism and voting behaviors of power users will be received well as it spreads beyond the early-adopters. From the majority of people I talk to about it, they first express love for the site but couch that statement in saying its difficult to navigate, overwhelming given how much information is on there. Voting noise amplifies this problem. As the service scales to bring on new users, the same thing may happen to users as has happened on Facebook — that they ended up friending so many people and “liking” so many pages and brands that their feed turns into an unruly customized feed of Google-searches, difficult to manage and find signal within all the noise — and, isn’t that what Quora is all about?

Oh, and if you liked this, feel free to up-vote it <smileyface> “Thanks”