Centralised Platforms 1/x
Next: Onion Rater
Before we jump into new forms of organization, its good to spend some time going over the key characteristics of our current models. This allows later to compare and see what are the invariants - things that will stay the same in future. But more on that later.
Centralised Platforms
TL;DR Centralised platforms are two-sided markets, hard to create but once successful, lead to winner take all phenomenon. They create trust and help sellers and buyers find and trade with each other. However, when growth slows down, they start competing with their users. Allowing users access to their own data would enable 3rd party apps tailor the user experience and mitigate some issues.
Welcome to a set of posts discussing the nature and effects of centralised platforms. The dominant design how we interact with each other today on the Internet.
The core topics covered in this set of posts:
Platforms are two sided markets matching need and supply. Hard to get started but with network effects leading to winner take all. Sooner or later the platform operators start competing with participants and they have unfair advantages with data that other parties do not have.
For communications platforms negative content easily starts to dominate due to the higher focus in our brains for potential threats. Various mitigations like unbundling and personalising filter algorithms and the personalised “lenses” for content.
Expert systems have their own flavour because legal practices and customs differ making them more local by nature.
Platforms today are in the middle between buyer and seller but nothing prevents platform formation for the endpoints – suppliers and customers. These would fundamentally change the nature of the resulting system. These endpoint platforms have the same two-sided market problem. In global world with very large number of participants with no direct way to communicate with each other, have not materialized so far.
We end by building a defence for social media platforms showing that focus on threats is a feature of the human mind and that the flat model of Internet amplifies it. Some part of the problems with social media are a natural feature of digital world irrespective of how platforms work. Neither was old media never such a bastion of truthfulness free of disinformation.
Nature of platforms
Today anything can be bought or acquired through centralised platforms. You can find jobs, educate yourself, buy cars, houses and home appliances, listen to music or watch videos, find spouses, rent rooms, find rides or repair persons etc.
From functional point of view, they are all web shops. As a buyer, you can search and browse offers and they have recommendation engines surfacing products that other people like you have consumed in the past. They take care of the payment transactions and make the revenue share between themselves and the seller. After the delivery feedback is gathered to create quality indicators. Sellers maintain their catalog and can see what orders they need to fulfill. Both parties see a history of past transactions. This is essentially all what a web shop does.
Good About Platforms
Centralised platforms are important in creating trust and making buying and selling easier between parties that do not know each other. If I want to buy say a book from a distant book seller from another country, how will I know if they are trustworthy so I have courage to send my credit card details to them. Especially if both parties know that most likely this is the only time, we make commercial transactions between each other.
Trust requires a trail of interactions where I have noted that the other part has always fulfilled their obligations; done what they promised.
Platforms enable that. I have multiple interactions with the platforms and should an individual seller on such a platform be fraudulent, the platform has strong incentives to compensate my loss and to take actions like warning or barring such a seller from their platform. The platform has no reason to protect either fraudulent buyer or seller. Also, my credit card details are known only to the platform, not to the myriad of independent sellers. I trust the platform and the seller trusts the platform. It is a trust mediator.
The platform works as trusted intermediary and I can safely buy and sell to parties I know nothing about. Everyone’s reputation is visible on the marketplace. This makes business faster and safer.
Let’s make comparisons.
When a small company tries to sell to a big corporation, you are hit with long legal documents. Take it or leave. Endless hours are wasted reading through them and trying to understand what liabilities we are signing to or as alternative, hiring expensive layer is needed. And this with every new large customer.
When two large companies try to buy and sell from each other, there usually is a room full of layers who try to make clauses for every possible eventuality that might happen in future. In today’s world seeing everything that can happen in advance is difficult and such lists are increasingly long. That’s why deals last months and can cost tens if not hundreds of thousands even in relatively small contracts. Not ideal.
The unpredictable, highly interconnected and increasingly chaotic world clearly favors platforms.
You can also compare platforms to insurance. When something bad happens, insurance can compensate say 80% of the damage ensuring that nothing bad really happens. Platforms are better as they actually reduce risks by making reputation visible and by having large incentive (being cut off) for both parties to honor the contract. They make trade easier and faster, something that insurance does not do. Some platforms compensate fully at least small losses.
The Bad
Platform vendors are in the business for themselves, not for the sellers or buyers. This leads to some negative consequences.
Low price increases customer traffic and platform profits. This may be the reason they have not been very effective in finding and removing counterfeit brand products from their catalogs. At least in one case a major brand found that over 90% of its chargers were replicas on a large market place. As a customer you buy something with small discount and get potentially lethal products. Platforms protect the transaction (safe transfer of money), not the end product you get.
For the seller the case is similar. Publicly listed platform companies have a built-in need to grow to keep the share price up. Prices reflect the belief of growth. When growth stalls, prices drop sharply. At some point a platform cannot grow any more by selling more stuff and growth threatens to slows down. The only way to grow is to start selling competing products under own brand cannibalizing sellers’ revenue and profits. This is not very hard as there are plenty of contract manufacturers willing to produce goods for anyone and due to volume, platform company gets good discounts.
The platform knows exactly who the customers are, what product categories are good sellers and what are margins for different products. They can pick the best categories for their own brands and leave all unprofitable categories to “business partners”. In other words, they start attacking their own sellers. They are in a good position because they have full visibility, do not share customer information with sellers and can decide what customers see and in what order see as they browse the catalogs. This is like being both a team playing in a football match and acting as the referee at the same time. An enviable position to be in.
Dynamics on social media platform is different but equally bad. Platforms can maximize their profits by keeping the users engaged for maximal amount of time. This allows them to show more ads and earn more. Human minds are wired to react much stronger to negative information than to positive news. This is an evolutionary adaptation. For our ancestors a single bad outcome could have killed them or injured severely, while positive outcomes only have fairly short-lived effects. Ancestors with more cautious and stronger reaction to potential risks had the edge. On social media the built-in focus on danger and negative information leads to a state where prioritizing emotionally strong, cynical and aggressivity generating information keeps people logged on longer, generating more revenue up to detrimental levels for mental health. This behavior is called doom scrolling.
There are also state level players on social media platforms spreading false information. The reasons for this are political. Authoritarian regimes want to show to their own population that democracies do not work anymore and they do not have alternatives but must accept the rule of the current regime. They are systematically trying to cause political instability, drive people away from each other and play down any news surfacing about their abuses with counter information.
Content platforms like video sharing have their own problems. They operate based on advertisement and small portion of that money can go to the content creators when they sign up for revenue share. The proportion and generated revenue are small. This means that the more content you can create, the more money you will create. As content creation is expensive, the best approach is to write computer programs that generate synthetic content. For example, videos for children with very small changes like same video just changing main character from lion to cheetah to hyena to dog to…
On big content platforms over half of the content can be computer generated. All platforms try to detect fraudulent and low-quality content and press it down in search results. When overwhelming majority of content is not human made, the machine learning models can flip over. They think that synthetic content is the gold standard and this highly irregular stuff coming from humans is suspectable and should be severely demoted. This moment is called ‘The Inversion’ when normal becomes unnormal and vice versa. Since real humans do not get positive feedback anymore as no one views their content, they get demotivated and stop producing content and visiting the site. Soon real users follow. Business grows no more.
There are mitigation tactics available, but before it a word on two sided markets and how platforms get gamed today.
Two-sided markets
Nature of Two-Sided Markets
Platforms are example of two-sided markets where supply and demand meet. Building a well-functioning two-sided market is a chicken and egg problem. To get sellers onto your platform you need a lot of buyers and buyers are attracted only to platforms with lots of sellers with lots of variety on offer.
There are many examples of two-sided markets – newspapers bring together readers and advertisers, stock exchanges bring together investors and companies, web search brings together information seekers and content providers (with advertisers) etc. Standards also work in similar manner bringing together manufacturers who get a larger markets and buyers who want to buy from a bigger selection of vendors.
Two-sided markets have economics of scale. Bigger network is usually better for the participants. They normally have a same side effect to boot meaning that people on the same side benefit from each other. Having a big community using a product like say a game console usually means you can share experiences with friends or post of community forums and get problems resolved quickly. This is not universal however as with niche products with limited set of customers sellers may not prefer presence of competitors or buyers for collectibles do not appreciate other bidders driving prices up.
Two-sided markets can in principle charge both sides of the service as they also have costs associated with serving both, but it is more common to subsidize the more cost-conscious side.
As the network grows, so do returns for participants. This tends to create fierce competition between platform operators. This leads to a winner-take-all end result. Even a small lead means that the current platform leader can use their higher profits to invest more into lowering prices or improving the platform thus accumulating their lead. The outcome often is that platforms end up with a near monopoly. You can clearly see this in operating systems and various Internet services, although on the Internet the cost of changing is for most part very low so new entrants have possibilities to topple the leaders. But usually, the toppling a dominant player comes from the side – offering something quite different that replaces it. Social media companies eat away the profits of newspapers by providing something different.
Misusing Platforms
Most interactions between people are happening today through digital platforms. This is true both for personal communications on social media platforms as well for commercial interactions on business to consumer (B2C) market places and business to business (B2B) procurement platforms where companies buy from each other.
All on these platforms have trust issues. When people can act anonymous there are smaller psychological barriers restricting anti-social behavior such as aggressive tone on social media and various monetary frauds on commercial.
In other words, users can easily game the current platforms and platform operators cannot really solve these underlying structural problems without knowing customer identity solution that due to costs and large scale they are unwilling to implement.
Many well-known methods are used:
Computer generated content for various social medias – videos, short postings.
Paying customers for good reviews, self-evaluation or agreeing mutual reviews (swapping praise)
If platform requires that you have bought the products you review, vendors can compensate the product price (i.e., give you free product) as payment for good reviews. The reviewer can then sell the product on the same platform to turn it into money
For mobile apps there are free download days and during these days phone farms download the apps pumping download numbers up and taking the app to the top of the most popular apps category
Spreading disinformation online for political reasons and using bots to make it appear very popular on the services. When this continues long, all types of beliefs with no connection to reality can become popular affecting decisions. In the past the role of art was to imitate reality. When online dominates, the role of reality will be to imitate “art”, this is the core of identity politics.
Dilemma of online information
There is a dilemma at the heart of many community and content services on the Internet. Social media and search are the main areas where this is present today.
People have right to privacy and at the same time there is right to information. Both are mentioned as key human rights in European Convention in Human Rights but in different articles. The question is to what degree for example search engines can make available and show information about individuals that is damaging to them even when this information as such is public and true. And to what degree this depends on for example how old that data is.
Right to privacy says that people have right to be forgotten if information about i.e., their past actions or present orientation is really damaging to them but at the same time right to freedom of expression says that people have right to voice their opinion on true and public matters.
There is no simple answer and the interpretations differ between say Europe and the US. For example, in EU people can submit requests to global search engines to have links to their past behavior (bankruptcies, insolvencies, criminal cases etc.) removed from results and a large percentage gets accepted. At the same time it’s legal for newspapers to have local search engines that return those results. Building a meta-search engine that fetches results from a fleet of those local searches then again may or may not have the right to show those results.
In principle there is a difference between availability (the information is there) and accessibility (it is known to exists and the whereabouts are known) and in offline world this difference has been used in balancing these two rights. Online this difference no longer holds. People access information via search engines and if the link is not there nor are you allowed to share it, that piece of information vanishes completely in practice.
The there is also the meta-question: should delinking itself be public. This dilemma is a topic that truly keeps giving.
Many people just assume human rights support each other without realizing that sometimes they do and other times they don’t. Everything begins with admitting there is a conflict.
If you want to read more on this topic, “The Right to Be Forgotten": a Philosophical View by Luciano Floridi is a good start and it is legal to link to it: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3853478
In the next post we look at ways to mitigate negative effects of centralised platforms with among other personalisation, and The Onion Rater in particular.