• Evan Warfel

How to Improve the Facebook Newsfeed


Towards a More Humane Social-Capital Slot Machine

Refresh the Facebook newsfeed too many times in a row and you’ll come into contact with the oleaginous feeling of capitalism. It’s the same feeling you can get when you see nothing you like on Netflix, when you walk into a supermarket with fluorescent lights, or when you inhale the toxic gas (aka the “smell”) in a new car.

This feeling is the product of being on the downstream end of what I call “denovation,” or frustrating incremental improvements that weren’t made with your well-being in mind. It’s the feeling that you are being used on some level while enjoying the process on another. And it is usually worse if the company is publicly traded. Look — if Netflix really cared about you, they wouldn’t have uploaded Marvel’s IronFace to their servers.¹


Everything you need to know about Marvel's IronFist is contained in the still above. Thin-slicing for the win.

As for the component of the oleaginous feeling specific to Facebook (FB), I suspect that this is related to three things: A) that browsing FB doesn’t require any skill, B) a constrained design process, and C) that browsing FB is almost exactly like gambling at a slot machine. (Except instead of gambling money for money, FB browsing requires you to gamble your attention for social meaning.)

The “lack of skill” part means that it is very difficult to get into Flow while using FB — there is no challenge and nothing to improve at other than expressing yourself.

While fixing this might require entirely rethinking the product (perhaps not the worst thing), in the meantime, I thought it’d be a good design-thinking exercise to articulate some ideas about how the Facebook newsfeed might be improved. As long as we are all semi-slavishly wagering our attention with what we hope are good odds, we might as well have a better experience while doing so.

Caveats: While I am neither an official product designer nor product manager, I work as Data Scientist and I have a background in Psychology and Cognitive Science. Like billons of my fellow simian descendants, I also use FB too much. And lastly, I’m skipping the whole Fake News / Cambridge Analytica angle (s)— that’s being written about in other places.

What problem is Facebook (and Instagram) solving?

At a high level, the problem FB ends up solving can be described with one or more of the following phrases: feeling more bored and unfulfilled than we give ourselves credit for; feeling disconnected; ennui; loneliness; not being fully integrated and self-actualized; operating with aspects of ourselves that we haven’t fully explored; being parented by people; wanting to have new things to focus on that don’t push us too much; being in an unfulfilling work environment, etc.

Though we consciously feel the desire to check facebook, it is much more difficult to be cognizant of what leads to that conscious desire.

What problem is Facebook (and Instagram) attempting to solve?

How to best capture your attention to sell ads while avoiding class-action lawsuits. Or as they put it, “Give people the power to build community and bring the world closer together.” ²

How does the newsfeed work?

The newsfeed is a lens that shows you a summary of the network of social networks you are connected to. To think about how it might be improved, let me first run you through how it likely works. Even though I'm speculating here, and no doubt simplifying a complex process, it's still a valuable exercise.

Specifically, every time I refresh the page, take all the recent content generated across my social network, and:

  1. Rank all the content by how much of an affinity I have for each person posting. Base the affinity on the history of past interaction, how much time I’ve spent looking at the other person’s profile pictures, mutual time spent looking at each other’s pictures, number of reciprocated likes, number of unreciprocated likes, chat history volume, etc. Bias the affinity towards more recent interactions.

  2. Create a second ranking of how much affinity I am likely to have for the category of content. Do this in terms of preferences (I don’t like videos, I do like long articles, I also like well-expressed personal updates…) and more mundane constraints (I just checked Instagram, so de-weight the FB- Instagram reposts unless I liked them, I’m checking the website via mobile, so don’t show me as many videos…), etc.

  3. Create a third ranking for how likely I am to appreciate the content itself. I’d be extremely surprised if Facebook didn’t know how to recognize wedding and relationship announcements, or when someone announces they are starting a business or have just gotten into an exciting program. FB analyzes the text of everything you post. They likely analyze the messages too. FB also scrapes the pages of everything you share. With all of this data, FB probably keeps track of how well you like various linguistically identifiable topics, which means they can calculate the “distance” between a given post and your preferences.

  4. Look at how much traction each piece of content already has, and calculate how similar am I to the people/subgroup that is giving that piece of content traction.(In this way, posts grow to be popular within a community before being exposed to people on the periphery.)

  5. Calculate a weighted average of these rankings, and discard all newsfeed items below a certain threshold.

  6. Order the selected items by some combination of the score used in (5), how recently it was created, how quickly a post is gaining traction, whether I saw it the last time I checked the news feed, and the intensity of the emotionality extracted from the text in the post.

I’m skipping over things like RSVPs to events and such as well as any notion about how consistent someone is in terms of these rankings. I’m also assuming the whole thing has not just been thrown into a {deep, convolutional, recurrent, recursive, subversive} Neural Network.

Let’s turn to how FB might be improved. One sentence summaries of each point are provided in italics where appropriate.

Saving Grace

Due to the constant refresh, FB intentionally makes it hard to find content that you've seen once. An experience that was more closely aligned with the human needs of its users would allow us to save and bookmark content. Maybe it's a good idea to ban "refresh the view on each new page load" behaviors.

Like it like that

A “like” and other non-verbal interactions on Facebook (and Instagram) indicate both social support as well as an expressed preference. Sometimes I like something because doing so reflects my reaction to what the post means for the poster, rather than because I prefer the content of the post over other things.

One way to solve this problem would be to allow people to express their support and preference at the same time with two different interactions, but only communicate the support (aka the “like” and not the “degree thereof”).³

Those frequent posters

Allow me to de-weight, but not completely “snooze” people who post frequently.

Broadly speaking, each piece of content in the newsfeed communicates two types of stuff — information about the poster, and the information contained in the content itself.

It also happens that some people use the newsfeed as if they were the editor of their own tiny magazine. More power to them. However, it’d be nice to filter that kind of thing out when I don’t feel like wading through all of their content, and un-filter it when I do.

I say this because seeing Alice post seventeen times a day doesn’t tell me too much more about Alice, other than they are the type of person who posts frequently. Presumably, I already know this from yesterday. I’d go for “unsubscribe” but I don’t know if that means I’ll never see that person’s posts again; and if that is what I wanted, I’d de-friend them.

Further, because there are more opportunities for me to engage with frequent posters, FB may be mistaken about either my relative level of interest in said person, the “confidence” they place in that estimation, or both.

Here are two ways to solve this problem.

First, FB could allow people to manually influence the frequency (high, medium, low) with which people appear in our newsfeed, without indicating that we never want to hear from them for 30 days or unfollow them completely.

What I’d really like to be able to do is “snooze people except for their updates which are likely to be important or get a relatively good amount of traction.” This could be done by making the threshold for whether I see a person’s next post increase every time I see a post of theirs within a certain time period and have this threshold decay back to neutral otherwise, assuming they are are in the relative top 5–10% of most frequent posters in my social network.

Second, FB could allow people to indicate how important their post is to them — some posts are more personally meaningful than others.

Yes, people can do this in words and “*important*” tags. But often, when someone has something important to say, they communicate that non-verbally as well, which opens up interesting design/UX considerations. For example, every person could post with three categories of importance, and the subscribe feature could give people the option of seeing everything that a person posts, or only their most important posts, and so on.

Pavlovian Engagement Curves

Allow me temporal control over my newsfeed experience.

There are times where I find myself quickly checking and then quickly exiting FB, because I have no idea if it will be worth my time. With over two billion users, spending money on infrastructure that supports me (and everyone else) quickly exiting FB must be a low a return on investment.

The same thing goes for people who are feeling a bit depressed and spending lots of time on FB— are depressed people more likely to act on ad over and above non-depressed peers? Should an impression count as much? I couldn’t find a marketing study to answer this question, but it doesn’t seem out of the question — depressed individuals have been found process information differently than non-depressed individuals and frequently report less willpower.

Assuming this hypothesis is true, if browsing FB while a bit depressed means a person is less likely to view ads with interest and click on them, and doing so does nothing to improve their emotional state and may even worsen it, then FB is wasting money due to how “addictive” its newsfeed is when advertisers pay per-click. ⁴

We all know that Facebook’s success is predicated on the idea that the ads on the site have value. But ads won’t have value if they just sit there and no one pays attention, engages with, or chooses to act on them. Why would anyone knowingly pay for a TV spot if they were aware that the TV owner muted the ad or wasn’t paying attention?

FB might get more revenue-making engagement if they let users pre-determine their own “level of engagement over time” curves. I would love to be able to make FB get more boring at certain periods of time. Or, if it’s a slog of a day at work, for me to dole out the quality posts over a longer period of time and not take the hit all at once.

Giving us more control over when and how we check our newsfeeds would help solve this problem, as well as give us a greater sense of autonomy.

Search

Allow me to search for types of content popular within and recommended across my social networks.

If the FB newsfeed is a lens into the network of social networks you are a part of, then there is plenty more FB can do. What if I want to see nothing but long-form articles posted across my entire social network over the past month? Or videos? Or everything but videos? Or the most popular blog posts shared by all my tech-friends? In other words, Facebook could be a search tool, too. Maybe I want to go on a nostalgia trip see what people I knew in college and my first year out are up to. Maybe I am going to New York soon and want to see posts from the people I know who are there.

This line of thought also leads to another idea — is there any way for me to see what my friends see on facebook? What articles are trending across my friends’ social networks? I may not be able to see who has liked or disliked them, but I would be interesting to see an aggregate summary of how articles and other content trickles down to me, especially if the activity of my social network could be effectively visualized. ⁵

Taming the Shrew

There is information in the quality of a click and of a view. If I watch a video at 2x speed, should my “view” should count for half as much?

One of the biggest problems with the facebook newsfeed is an assumption that most of us make; one so fundamental to our usage of modern technology that we rarely notice its downstream effects.

Ladies, Gentleman, and everyone else, consider the mouse click, that binary indicator of anticipated interest.

A click is just like a movie ticket. Purchasing one doesn’t quite reflect your “captivation” or “engagement” — it gives you the option of reaching those states.

But even if a movie strongly appeals to a niche group, say, 90% of a subculture, the way ticket prices are currently set up means that the niche film, though loved, will not make Big Bucks. Conversely, Marvel Superhero Movies try to appeal to as many people as possible by appealing to multiple taste subgroups, but not enough to turn off other groups. The result is an “appeal to the bland middle” effect. All because we are using a binary instead of a continuous measure. But what if there was a mechanism where at the end of the movie you were charged a price relative to how seriously it engaged your attention? ⁶

This sort of thing is harder to do with movies, and much easier to do with computers.

Imagine the widespread use of a new type of mouse, one with three levels of “click depth” that indicated “predicted interest” — burning curiosity, peaked interest, and passive interest. And as part of this adoption, further imagine that the click-depth was always and consistently communicated across the browser.

Instead of competing for your click, websites, ads, content producers, and so people would be competing for your strong-interest clicks.

In some sense, everyone could win. Users could see more relevant content, product managers could figure out who is strongly vs passively interested in what they are managing, and advertisers would love to know how people are clicking on their ads. Click-bait headlines might disappear, and the “appeal to the bland middle” effect might weaken. (I am willing to bet that the popularity of clickbait is due in part to the fact that most people who click on the title are moderately to passively interested in the headline for different reasons.)

“Strongly interested clicks” could become more valuable for both content producers and the advertisers over and above “total number of clicks.” Webpages could be customized to a user’s expressed “level of interest.” Spam posts asking people for clicks would loose some of their power.

But though it seems like all of us would be better off with this solution, there’s a coordination problem in the way. Unless a large giant coalition of companies (including the New York Times, Washington Post, Facebook, Google, Apple, and the other computer makers) all band together and willingly adopt a new protocol, and then we all get on board, this ain’t going to happen… Even though there ways to approximate this with “length of button-down time” and pressure sensitive taps.

How hard is it?

Not very.

We all can see the effects of more and more FB and Instagram screen time. Adults watch their phones in many of the spaces between the moments where they would need to watch their kids. At the same time, I’m willing to bet that most people don’t love every second of doing so. My intention here is to make suggestions that improve the wholesomeness of the everyday FB experience.

Broadly speaking, the way to do this is to make interacting with facebook feel more organic by mimicking how people non-verbally communicate. Due to confirmation bias and capitalism, these are sadly seen as counter to FB's business "needs."

Another thing FB could do is be more transparent about how the newsfeed is put together, as well as the consequences of the various actions you can take to alter your FB experience.

Currently, not a whole lot of the time spent on facebook is, in fact, “time well spent.” The same is true for checking Instagram, binging Netflix, or watching TV. Does Facebook promote vitality, growth and richness of character? Maybe. If you’re careful. Can it be a way to relax? Well, sort of. There’s a kind of relaxation you can achieve if you walk meditatively along the beach for two hours, and that’s not what you get when you look at Facebook.

But while Facebook has certainly contributed to our eyes getting glued to the screen, at the end of the day, it wouldn’t exist if the demand weren’t there to be exploited.

What does the existence of the demand for FB mean? Maybe the demand means our jobs could be more fulfilling than we think they are. Maybe we aren’t as happy (broadly construed) as advertisers and pop culture would have us believe. Maybe we only think we are engaged in what we are doing when we aren’t checking our phones.

How many of us check social media to prevent otherwise awkward lulls in conversation? Or feelings of emptiness? We likely aren’t introspectively sensitive and articulate enough to perceive and identify all of the upstream, internal triggers that cause us to want to check FB (or Instagram, or Twitter, or our favorite news site, or play a game on public transit, etc.)

Perhaps The Man in the High Castle — Phillip K. Dick’s meditation on mass delusion and consensus reality — was really about advertising and marketing and the power of believing in what one thinks one’s experience is.

Why would Facebook do any of this?

Though the suggestions here might harm FB's earnings in the short-term, the damage they are doing to their brand is high. Researchers are increasingly turning their attention to the effects of social media. Most of us are aware that we are "wasting time." Long term, this means that what FB does in the future will be increasingly less credible.

In the meantime, Facebook, as long as the demand for your product exists, the least you could do is make your user experience better and more wholesome for a large swath of humanity.

________________________

Endmatter: Addenda and Footnotes

A. Opting out.

If you are still reading, you may and/or should concerned about how much data FB has on you.

Here are links to opting out of the Six Data Brokers FB buys your data from:

https://isapps.acxiom.com/optout/optout.aspx;

https://us.epsilon.com/consumer-information;

https://www.oracle.com/legal/privacy/marketing-cloud-data-cloud-privacy-policy.html;

https://solutions.transunion.com/facebook-opt-out/;

https://www.i-behavior.com/opt-out/; and http://www.experian.com/privacy/opting_out.html.

B. Footnotes.

  1. One of the best explorations of denovation was penned by Scott Alexander, who among others, has pointed out that if society seems inhumane, well, it wasn’t exactly designed by humans. Instead, society is designed by groups of people responding to incentives, or, he argues even those incentives themselves. Goodhart’s Law and Campbell’s law are more specific instances of Denovation. Academia is also great for solid examples of denovation, like when seemingly well-educated individuals spend their lives shoddily studying how people order pizza; or when they fail to realize that their Stanford Prison Experiment says more about them than it does about human nature.

  2. It isn’t clear that FB’s official motto distinguishes them from “email” or “fiber-optic cable” or even “the telegraph.”

  3. It’s hard to be certain without data, but I’d wager Facebook sees a population difference where younger people are much more prone to like their friends’ content because they are friends with them and older people are more prone to using “likes” to express preferences.

  4. I once had an idea for a short story component that involved the black-market sale of illegal newsfeed ranking algorithms which had been deemed by the government as “too addictive.” The board responsible for determining the legal was jointly chaired by the Nevada Gaming Commission and the Federal Reserve — every time the economy needed a boost, the feds would adjust interest rates and tech companies would be forced to make their newsfeeds more boring.

  5. Maybe it’s my background in data, but I think an interesting use case for the Oculus Rift would be to visualize your social network (and maybe statistically approximate and anonymized friends’ networks.)

  6. Finally, a non-medical use case for EEG hats.

C. References.

Hooked: How to Build Habit Forming Products. Nir Eyal with Ryan Hoover.

Flow: The Psychology of Optimal Experience. Mihaly Csikszentmihalyi.

D. Addendum on Advertising.

Why do companies buy ads for multiple slots in the same online television shows? From a Bayesian or Information-Theoretic perspective, perceiving the same ad twice shouldn’t do anything to change your propensity to buy something. Clearly, it is highly unlikely anybody’s need for a product changes while watching a television show.

It’s hard to be sure from a distance, but here are six possibilities:

First. Maybe the advertisers think we aren’t too attentive and miss parts of the ad the first time, so repeat viewings are necessary for it to sink in.

Second. It is currently a societal norm to either a) think you know better than other people and reason people out of whatever emotion they may be feeling, or b) interact with people like this. Perhaps this sort of thing is what makes celebrity endorsements common — being reasoned out of your current emotional state into buying a product is one way ads get you, no? And if it works once…

Third. Maybe the people who choose to buy the same ad-space at each commercial point in streaming televisions shows are clueless as to how ads work, or just don’t care.

Fourth. Maybe those people are being lied to.

Fifth. Perhaps the real point of advertising is to force you to associate emotional states with products, and the more times you view an ad, the more you’ll form this association. Assuming this is true, I have to imagine there’s a more optimal spacing than every commercial break on The Expanse.

Sixth. Maybe ads play a vital function in reinforcing the notion that we all live in the same society. Maybe one reason Americans seem to think America is falling apart at the seams is because some of the glue that most holds the country consists of Television, Advertisements, and (possibly payola-ed) Pop-Music. Thus the multiplicity of the same ads is nothing short of patriotism on the part of marketing executives merely looking to help the social fabric of America at the expense of more effective campaigns.

...

Thanks for reading! Feel free to comment or connect with me; I would love to explore what I’ve mentioned here further.

#Facebook #HumaneTechnology #SocialMedia #Product #DesignOrientedThinking