After graduating in 2019, I’ve always been mindful of my social media consumption. I had permanently deleted Facebook and Instagram accounts. I’ve been keeping search and watch histories paused on YouTube. It depersonalises the video recommendations. And I barely check Twitter as I find it emotionally expensive. So last year, when Ravi Soni proposed that we try to solve the social media, I didn’t need convincing to join him. I understood that we were setting ourselves up for failure. But I also think that it is one of the most urgent problems of the 21st century.
Core problems with present social media
We started discussing what we were building sometime towards the end of August 2020. During the discussions, we came across many problems. We focused on some while ignoring others for the time being. For brevity, I am organising them into a few coherent classes.
We are a social species. We need social validation and acceptance to thrive. The problem occurs when we seek too much of it, and social media helps us maximise it. Furthermore, it presents us with metrics such as the likes count. We use this data to paint a self-portrait from others’ outlook. We try to optimise our online behaviour based on our inference. As a result, we exhibit narcissistic personality traits.1
Social media helped us monetise, and so we did. We monetised our lives, kids, pets and even our acts of kindness.2 If it’s happening, it’s happening in front of a camera. We wouldn’t have to look far to find people who display distinct behaviour online and in real life. Everyone behaves a little different under the scope of social media. But the distinction in behaviour is pathological for some.
Disinformation, filter bubbles and political influence
Disinformation is a threat in general. And social media is an effective tool for spreading disinformation. But social media can not cause this problem.3 Neither does the problem have a technological solution. It is also simply not realistic to fact-check everything we come across.
Yet, social media causes an enormous problem by producing information bubbles. Mainstream services use recommender systems to generate personalised content feeds. The system guesses what information we would like to see based on our history. While a recommender system boosts user engagement, it has a critical side-effect. It effectively isolates us from counter viewpoints. And thus, it can (and does) polarise us to the extremes of our ideologies.4
Standard operating procedure for businesses
Businesses are amoral and optimise for revenue. Companies that offer software services for free earn revenue through advertisements.5 They social-engineer products to be more engaging and advertiser-friendly to maximise revenue. The strive for higher user engagement made social media chronically addictive. Optimal advertisement placement causes the abuse of our private information.6 Social media companies imposed a very high cost to the little their products promised - a tool to share personal updates with the people we know.
After some lengthy discussions, our efforts yielded HymnFeed. While working on it, we also participated in Y Combinator’s 4-week build sprint. Although we couldn’t build a functional demo in time, it had the following fundamental features on paper.
No influential metrics
It is perhaps the most addictive feature of the present social media. We agreed to remove the like, comment and share counts straight away. Removing them breaks the social validation feedback loop. If a user wants a count, they could see a list of users (or comments on a post) and count manually. Recently, Instagram also added a feature to turn off the likes count.
Infinite scroll but with a natural chronology
We decided not to prioritise items in a user’s content feed. Instead, we decided to use its chronological order. It ensures that content feeds don’t create filter bubbles. We did not want to guess what a user might like; we wanted them to decide it.
Consumable post tokens
Ravi came up with the concept of consumable tokens. When a user signs up, they get a fixed amount of tokens. Users can neither buy nor transfer these. They consume these tokens by posting on the network. On spending their tokens, users will lose their ability to post. This flow limits the amount of content a user generates. And so, it compels users to be more thoughtful of how to spend these. It is analogous to the philosophy of life: you’ve got limited time, so spend it wisely.
Starting over from scratch
We addressed most of the problems we were trying to solve. Yet, we couldn’t figure out a way to monetise the product. Advertisement based business models only work with highly engaging products. What we had was intentionally dull in contrast to present social media. We could try to boost its user engagement. But for us, that meant creating another Facebook, so it was pointless.
By the end of September 2020, we decided to pull the plug on HymnFeed. Ravi moved on to work at SuperTokens while I shifted my focus to a different project. Some time towards the end of November 2020, I rebooted the project from scratch as Soir. I carried forward the ideas we had discussed so far. But before considering these ideas, I started with the very basics.
The free flow of information is necessary for a society. But social media shouldn’t be the tool for propagating socio-political information.7 Instead, it should limit such opinions to circles where a friendly debate is viable, i.e. among the people we know. If one needs to expand their audience, they can seek other media targeting specific cohorts, e.g. blogs. Soir had WhatsApp like contact synchronisation to emulate this model. If a person is not in your contacts, you will not interact with them on the network. You can only view their profiles and exchange comments through a mutual friend’s post.
Replacing the like count with the view count
Should we care who likes what we posted on social media? An entire genre of philosophy says otherwise. In an ideal world, I shouldn’t even need the view count. However, I decided to keep the view count. I believe it can help in sensing the audience response similar to the like count. And unlike the like button, it doesn’t cause too much dopamine secretion.
Reconsidering consumable post tokens
Too much engagement is harmful; too little feels dull. Consumable tokens are a sweet compromise between the two. But unlike HymnFeed, Soir would have two sections. One dedicated for content from the contacts. And another one for hosting the public content. Posting on the public feed would need tokens, and reaching out to your contacts wouldn’t.
So, where’s the demo?
I stopped working on the demo in March 2021. I’m giving it another go now, picking off right where I left. I took a break because of the two dreadful questions that killed HymnFeed.
How will it make money?
Adopting an existing business model will imply adopting their issues. We know that social media and conventional advertising don’t fit well together. For a while, I was considering a subscription-based service model. But free social media have become a standard. I am exploring some alternatives, but everything is theoretical at this point.
The cost of switching and vendor lock-in
Most internet users are already using other highly engaging alternatives. So why should they consider Soir a viable option? It reminds me of two things. First, despite significant momentum against WhatsApp, users didn’t opt for Signal. Second is a tweet from Paul Graham — “In practice, good intentions rarely work as well as good incentives.” Considering that, my situation is most definitely gloomy. My initial goal was somewhat optimistic. I wanted to ship a functional demo to a small audience mindful of their social media consumption. I’ll let you know if I survive its first phase.
Thought of something constructive to share? Please feel free to write to me at ashutoshgngwr[at]gmail.com.
Pressing a like button is faster than typing a critical comment. So the data may not reconcile with reality, and thus inflating our inferences. But, we wouldn’t care about the quality of the data, as long as it gives us the illusion of optimisation. ↩
Here monetisation also includes non-monetary rewards. Many users will settle for an increased social media reach. ↩
Repeated information receives a higher truth rating based on its fluency (the ability to unconsciously and intuitively interpret data). The experiments suggest that people sometimes fail to search their knowledge base. Instead, they rely on fluency as a proximal cue because it is cognitively inexpensive. See “Knowledge does not protect against illusory truth [PDF]”. ↩
By revisiting these echo chambers, users repeatedly expose themselves to resonating viewpoints. They subconsciously reinforce their existing beliefs to their extremities. Moreover, as their ideology becomes more extreme, so will the content they consume and other users in their proximity. ↩
These companies don’t always use advertising. Sometimes they sell data to third parties (data brokers). Essentially when a software service is free, your data is the price. ↩
“In between the Internet user and the advertiser, the Journal identified more than 100 middlemen—tracking companies, data brokers and advertising networks—competing to meet the growing demand for data on individual behaviour and interests.” — The Web’s New Gold Mine: Your Secrets — The Wall Street Journal. ↩
Using social media for social and political debates trivialises nuances of the discourse. “The higher the Facebook use, the more the general political knowledge declines.” — Criticism of Facebook - Wikipedia. ↩