Bad actors, engagement, and online communities.

Hi everyone.

This is not a vent post, although perhaps it should be.

I’ve decided to take an indeterminate length break from all posting on Twitter, with the exception of informational (likely locked comment, which is unfortunate, given comments/feedback are usually excellent and it helps me and others learn). It’s primarily due to the extremely unpleasant interactions I’ve had with one person in particular who I’d rather not name, which has escalated from vague online harassment to repeated, direct harassment to libel and real-life employer threatening with the intent to cause financial harm (I assume without repercussions it doesn’t usually stop there either).

But anyhow, the main topic here isn’t to vent, and it will not be mentioned more. The main topic is to look at the structural design issues of social communities, which have become en vogue both in relation to COVID-19 and the exponential growth of crypto and “anonymous” identities online.

---

Twitter is an interesting example of both a parasocial and a hybrid community, in which the algorithm tends to focus on two major filtering structures:

1) Subscription to the individual - This is the fundamental basis of the Follower/Followee model, which can be one-way (Twitter) or two-way (friendship, like Facebook). In general, even natively two-way subscription modeled Platforms (think a Facebook or LinkedIn, with a “friendship/connection” model) tend to introduce one-way features. This is due to asymmetry of value, or the rise of para-sociality over time — much as in real life, there are certain individuals, content creators or famous folks, who have a significant demand for viewers, while a minimal need to see others’ content (think a Joe Biden, for example).

2) Subscription to the content - This is a much newer structure, largely related to technological advancement. Content largely is unstructured, and it’s fairly difficult computationally to suss out the meaning of English, or worse, pictures (especially when it falls under irony, which most memes do). Subscription to content tends to be a more prized feature than subscription to the individual, but largely depends on how good a social network’s content aggregation capabilities are.

To make up for the shortcomings of technology for #2, we mentally take many shortcuts to “discover” new content, largely based on the idea of neighborhoods. This is known in machine learning as recommender systems, and is an active and highly in-demand area of research. The crux of a recommendation is identify the salient features that matter to an audience. The hard part, of course, is figuring out those features. We can label the salient features latent, in the sense that we can only directly observe the relevant outcomes (a purchase decision, for example), but we cannot directly train a model on the features themselves. That makes it a much more interesting problem in many ways.

But enough about machine learning. Let’s talk more about the heuristics we use, and the idea of neighborhoods. Engagement on social networks is extremely important. Twitter, as @noahpinion, suffers in part from structural flaws related to the weight tweets are given in recommendation by large accounts, without potentially accounting properly for the longevity of those accounts. In English — an older account producing content regularly will over time aggregate more followers, no matter how asinine or incorrect the content they post is. Worse yet, this can be compounded by “shock events” (almost like jumps in market prices) — individuals can find themselves in the center of a crisis or hot-button issue (Evergrande, Gamestop, etc), post an interesting thread or two, and build an audience quickly.

Noah Smith 🐇 @Noahpinion

One reason Twitter is a stagnant platform is how long it takes to build up a following. You can be 10x as good a tweeter as I am, and yet after a year of concerted effort you'll have only 1/10 the followers I have, because I've been on here for 10 years.

The issue is in the hybrid platform paradigm, there’s few endogenous factors which, on quick glance, provide credibility or aegis to the correct people. This is an absolutely pervasive issue in the context of many global events like public health or even recently market structure and the Citadel controversy.

As much as we love to believe in the adage that one’s words should stand alone from their person, it never holds in practice. We are, at the end of the day, more highly-evolved apes than gods of our own devices; we are easily swayed by biases, emotion, and the machinations of others. Very few of us take the time and energy to evaluate content on its own merits (especially when it does not agree with our biases); we rely on rules-of-thumb to do the mental heavy lifting for us.

And this presents a massive problem, as the world moves more into the “metaverse” of parasociality and online communities. While those with crypto-idealism believe the metaverse and a virtual reality future will equalize the playing field, and ideas and good solutions and merit will stand supreme - it will not. We can observe the future of virtual reality on Crypto Twitter:

🤠🤠🤠 loomlock.eth (🫐,🫐) @loomdart

I am hiring solidity devs Pls apply in dms or email loomdart@loomlock.com Bonus points if I like u Double bonus points if you don't hate me

I have nothing against folks like Loomdart, CL207, Hasu, or any other crypto anon. In fact, on chatting with them, I have to say I’ve found them much more humble, personable, intellectual, and helpful than most individuals net I’ve interacted with in finance online so far (this is not to say everyone is bad; conversely I’ve found many amazing people as well). But CT is in many ways even more top-heavy than Financial Twitter. While we have no idea the human behind our favorite blueberry or cat in a hazmat suit, there is one, and through artwork and digitized personhood they have reach far greater and more impactful than even the financial influencers we love to hate.

And in many ways, as the space gets more arcane and moves rapidly, we affix even more value to their words, even in an era or paradigm we aim to view as “trustless”. From my own experience, the most common seal of approval online is follower count. No matter who you are, or what you’ve spouted, people have a halo effect on those with a ton of followers.

I’ve affectionately talked about the CAP-T model before in passing; hearkening back to the corporate finance 101 CAPM, under CAP-T we view the most central characters of the space as the visible outputs of the most important latent variable on social media — attention and interest. With a bit of elbow grease and Pandas, you can easily observe a correlation between momentum and crypto price and growth of the accounts most salient in the crypto space.

---

This works in good times well, but its success is metastable. First and foremost, recent evidence has shown the negative externalities of social media usage, in more ways than one. Facebook and Instagram are known as the place where pretty girls can make amazing sums of money, but the sword of Damocles hangs ahead; those same girls pay for their success in lost self-esteem, productivity, and the vitriol of a thousand keyboard warriors hiding under the veil of anonymity to tear into them to ease the black hole where their hopes and dreams once were.

The metastability is apparent when you consider the market structure, and individual incentives. As I’ve remarked both on Twitter and in prior posts, the traditional incentive structure for social media was designed primarily with a, shall I say, skeuomorphic mindset — we tried to replicate on building networks what we observe in real life (popularity as its own reward). This works well in the two-way follower paradigm, where the primary utility of the social network is to replicate one’s friend group and social interactions, but falls apart like tissue paper at scale.

It’s undeniable twenty years later that two things have occurred:

1) Social media has become more, not less, present in our daily lives — reporters live on social media. COVID-19 pushed many to seek solace online. We’ve changed the name of the platforms (Friendster → Myspace → Facebook → Instagram → ???), but it presents a thin veneer over the basilisk. Facebook may die, but social media will not. I cannot tell you what the world or even what platform will remain supreme in 2050, but I wouldn’t be surprised in 100,000 years if my descendants rage-tweet at faster-than-light speeds about some anti-nanobot conspiracy theory.

2) The traditional compensation model has gotten more and more tired — As I’ve mentioned, NFTs and crypto are intimately connected to the highly insular and centralized nature of crypto-Twitter and other venues. One must divorce this view from the overall Long View of cryptocurrency, and its adoption in developing nations and use cases far beyond trading leveraged shitcoins. We can observe the “traditional” model evolving in stages:

1. Payment in influence and audience - In the beginning of these networks, the primary currency of attention was fungible solely for audience building. This would be co-opted by some to create businesses or advertise existing ones, but the path to do so was fairly unknown. This also explains in part the relative latency for “social media strategy” to arise in business — in the greener stages of the social media revolution, it was hard if not impossible to translate the impact of a tweet on a company’s bottom line. Thankfully or not, most platforms have built tooling and accommodated this need, or perhaps we wouldn’t see such philosophers like:

   Wendy's🍟 @Wendys

   I am tired of lunch. These clowns. I am tired of people caught in the freezers of their beef.

   

2. Influencer capitalism — Fairly quickly later on, brands in certain verticals started understanding the power of social media marketing. This tended towards lifestyle items, especially given traditional advertising has proved time and time again — we love to see pretty people doing things, even if we say we don’t. Similarly, the rise of social-adjacent networks with a video component greatly increased the value-add of influencers. Most consumers will not, for no recompense, watch an advertisement for free. They will however, watch “honest” reviews of cosmetics, even under the veil of sponsorship. The influencer model, to borrow economics, is largely a game of exchanging credibility for currency. Brands can borrow the influencer’s clout and “honest social signaling” capabilities (for which they built an audience — at least the valuable part of their audience) in exchange for hard currency. Suddenly it became directly profitable to be popular enough online in the right ways.

3. Web3/Creator Economies — the issue for #2 is while it worked well for some, the risk to reward is all wrong for most. For starters, the influencer market is not a fair market — aside from the backdoor deals and follow-for-follow (or outright purchasing followers), it’s no secret that it rewards in general attractiveness and other genetically-endowed qualities. Moreover, it only provided significant value add for certain verticals (e.g. cosmetics, streaming) while almost none for others (e.g. financial education or geopolitics). Lastly, the risk of building the audience for an individual with sufficient credibility is massively asymmetric to the reward, which only really reached scale at a late enough stage in influence-building.

In the Web3/Creator paradigm, we are seeing increased demand from creators to receive adequate compensation for their work from all verticals. As a Twitterati (or perhaps former), it is a significant time investment. When I write threads, unlike perhaps many, I fastidiously checked my sources, and could spend hours writing the proper type of post to best inform my audience. My compensation for doing so, other than media exposure, thus far has been minimal or negative (opportunity cost).

The only direct monetization path for the traditional non-lifestyle influencer, especially on less "brand-valuable” platforms (such as Twitter) is auxiliary services — producing enough content (or clever marketing) in order to sell auxiliary services with your reach (newsletters). This is where platforms like Substack saw substantial market opportunity, and caught the zeitgeist.

For myself, I’ve been very clear since the get-go that my dream, despite certain accusations, has not been to enrich myself from social media or capitalize on engagement. I have had the fortune through platforms like Substack and Discord to help raise significant sums of money for charity, and hope to continue to do so (including through the NFT series I’m designing, which will give all proceeds net cost to charity). I do not mention anymore a trading signal that I personally use that has been monetized (by the creator of the site, although I receive a fraction). That said, per my demand to the creator (Sean), the site is closing shortly to any new subscribers (we will not kick existing users who enjoy it off, but I anticipate it will slowly die over time, as these things do). I have no interest in continuing to monetize anything, to be quite honest.

That said, we can view the advent of NFTs and alt-cryptocurrencies as a natural fill for a gap left by the poor economics of social media marketplaces. While attention is nice, at the scale and content generation needed to make an experience “worthwhile” to the end user, the marketplaces present a terrible offer to the actual content creators that power it. While they support and profit off of advertisements and engagement, creators without business interest in creating will find themselves besieged by ever increasing cost as they scale in viewership (emotional, mental). This of course impacts anonymous and named accounts quite differently — in the existing paradigm, there’s negative edge to being non-anonymous at scale, unless you have a commercial interest or need to (e.g. a lifestyle influencer).

More interestingly, NFTs represent perhaps the newest and final evolution of the advertisement (and associated cryptocurrencies). The issue with traditional advertising is, while a small fraction of consumers benefit (from discovering a product they may enjoy, for example), the cost is largely spread across the audience who has to view (and suffer) through an ad. This represents an equilibrium to discover for the content creator — in extrema, you could monetize for example 100% of your content, and temporarily make a ton of money, but eventually destroy your audience. At the other extreme, you can monetize 0% of your content, and suffer only the costs of producing it (e.g. dealing with emotional and mental toll, opportunity costs otherwise). NFTs as ads are interesting because unlike traditional advertisements, the consumer of the NFT/ad often thanks (at least during this weird bull market) and appreciates being advertised to. Unlike an ad for Gucci or Channel, if I follow an influencer who pumps an NFT, I can resell and benefit from the popularity of the influencer as well. In essence, everyone wins (well, not the last guy holding the NFT, potentially). This seems to solve the equilibrium problem mentioned above — as an influencer, if I create NFTs and other resellable content, in essence my audience can benefit from following me as well, instead of just bearing cost on my monetization strategy (tl;dr - traditional influencer monetization strategy is nearly zero-sum).

---

We still haven’t talked about, however, why this leads to metastability. As aforementioned, the value of a social network is not unlike a currency — it becomes from adoption and engagement (network effects). A Facebook without users isn’t a Facebook.

In many ways, Facebook and Twitter and most traditional social marketplaces present an amazing value proposition for the end user, but a limited value proposition for the creator. For an honest social signaler who doesn’t monetize, as reach reaches scale, the costs increase continually (“haters”, trolls, vitriol, reputational risk, etc) while the benefits tend to plateau. While engagement has some self-esteem value, it tends to be marginal. Compounding this, bad actors in the community who engage in harassment or antagonistic practices tend to weigh more heavily than positive commenters. This is further worsened by the psuedonymous nature of many online communities and low expense to create new accounts (called Sybil attacking, in network jargon). It’s trivially easy to disrupt online, and much harder to rebuild.

Conversely, if all influencers monetize, the costs are now borne by the end user, and engagement (and the platform’s value) dies with it. While influencers may be happy monetizing and ignore the costs associated (or, ideally, have a rep or publicist who reads the inane vitriol and spares their mental state), the opportunity cost for the end user using the platform increases. Few like to be advertised to constantly, and these communities similarly die rapidly.

In closing thoughts, I’m reminded much of the Tragedy of the Commons. Financial Twitter is an amazing resource, and I credit much of my knowledge and value from entering the markets and finance from it. There are many truly gifted, intellectual, and humble people on the platform, who give much more than they get.

However, attention with few exceptions, comes from the niche, not from the person. The advent and focus on content discovery algorithms means once you define a “niche” (for example, mine is apparently quant/volatility/crypto/shitposting), your account growth tends to be constrained by the interest in the niche. No matter how affable a Leveraged Loans or Henry Hub Twitter is, chances are they will never reach a mass appeal and large enough audience.

In the current model, the content creators that power niches like Financial Twitter come from two categories. Either they truly enjoy the craft and want to network with like-minded individuals (and learn), or they are looking to use it as a clever brand marketing strategy. I do not fault either. However, in the latter, certain methods of doing so impose cost on the audience, and lead to a decrease in niche attention. As previously mentioned — if we all start funds and use Twitter to advertise said funds (within confines of relevant laws, of course), eventually our aggregated audience will die down. Attention is a scarce resource, and we must all be stewards to see a vibrant community grow.

But the issue of course is divergent value distributions, much like our asset pricing discussion on cryptocurrencies. Some of us find value in educating and learning, as I often believe I do. Some of us find value in the financial benefits, and want to rep our knowledge to encourage fund inflows or sell newsletters or tooling. Others, unfortunately, find value in tearing others down, attacking others, and demonstrating simply their superiority.

In the end, no matter what your fundamental value for social niches is — we all lose from overgrazing. Online, especially under the veil of pseudonymity, there tends to be limited cost to selfish resource-taking or overt antagonistic behavior, and unfortunately some reward (engagement and audience-building). While we must weigh contents on its merits, at the end of the day, without trying to encourage building a good community, the community will fall apart.

We must all be stewards of the online experience, and our actions - whether via engaging with a creator’s content, helping monetize or support them, or amplifying information we choose to retweet and share - will shape the aggregate experiences and destinies of our online communities.