Discover more from Thomas Paine's Blog
How media outlets are using tech platforms to keep us angry and afraid
Beware the 'emoji' - the single most pernicious feature of the digital age
On his Substack, Matt Taibbi recently highlighted the contrast between soaring corporate profits and the cultural disintegration of American life. At the very end of his article he makes the following observation:
Keeping the volk at each other’s throats instead of pitchforking the aristocrats is an old game, one that’s now gone digital and works better than ever.
It is important to understand how this is being done, because stripping media outlets of their ability to keep us angry and afraid is incredibly simple:
Stop using emojis.
Yes, emojis... the single most perniciously powerful feature of today’s social media landscape. Bear with me as I try to explain...
There are two basic theories behind storing and analyzing data. ‘Set Theory’ is where data is stored in tables of rows and columns. The data in rows are correlated to other data where one ‘parent’ records a table may relate to multiple ‘child’ records in another table. There are other relationships in this kind of data, but to simplify: Set theory is embodied in Relational Database Management Systems (RDBMS).
The other approach is ‘Graph Theory’. The term ‘Social Graph’ is used to refer to an aggregation of graph data from the various tech platforms. In graph theory, data is simplified by reducing everything to ‘nodes’ and ‘edges’ (sometimes called ‘paths’). A node is a noun: a person, place, or thing. Let’s reduce this even further: A person who uses a modern tech platform (including both social media platforms like Facebook and email platforms like Gmail) is a ‘node’ in graph theory. The information a person engages with is also a ‘node’. To simplify, I'll refer to a ‘post’ that appears on a user’s social media feed. The takeaway is both are considered ‘nodes’ on the social graph.
The social graph also contains ‘meta-data’ about the nodes. Meta-data is simply ‘data about data’ and are conventionally called ‘labels’. A natural person on the social graph will have highly detailed group of labels associated with them (or their node) to allow for demographic segmentation. This is also true of posts on the social graph. Only here these labels describe the information in the post. Understanding the significance of these labels, however, requires we understand the ‘edges’.
If nodes are nouns, edges are verbal phrases. The user node is the subject, the post node is the object, and the path verbally describes the relationship between the two. The edges that matter most have to do with emotions - fear and anger in particular. To bring this all together:
Subject [node (a person)] - is angered by [edge] - object [node (a post)].
Subject [node (a person)] - is afraid of [edge] - object [node (a post)].
Thanks for reading Thomas Paine's Blog! Subscribe for free to receive new posts and support my work.
Those who regularly use social media know well that the ‘is angered by’ is now expressed by associating the angry-face emoji with the post. What exists under the hood, and out of view, are the labels and a technically arcane fact about the emoji.
When we stick an emoji on a post we likely think we are using a graphic to express the emotion. The truth, however, is the emoji is more like a letter than a picture. For computers each letter corresponds to a numeric value. Originally these values were between 0 and 255 (the range of 1 byte). To accommodate non-Western languages, additional bytes were added to expand the available values. These standard characters are now known as Unicode, and use up to four bytes. Part of the available values have been allocated to a standardized set of emojis. Without diving too deep into the math, the angry-face emoji as used in browsers corresponds to an ‘encoding’ called UTF-8, and a decimal value of 4,036,991,136. The relationship between the subject (person) and object (post) is saved on the social graph by using the number 4,036,991,136 as the edge, or path between the nodes.
If we assume we use our spoken language to express emotion, it is clear that if there is one way to say “this makes me angry” there are hundreds - in just the English language alone. If we multiply that by the various languages used on the Internet, it becomes computationally impractical to think we are going to aggregate the data accurately enough to start building emotional profiles of people using social media.
The emoji solves this problem, not only for English, but for all languages used on the Internet.
The large number of possible emojis was not created to make computers fun. They were created to allow graph database technology to be used by machine learning to create finely grained emotional profiles of social media users, even across languages. It becomes possible to create (and sell) highly precise emotional segments and profiles of social media users. Facebook was caught red-handed using information like this to experiment on its users - without their informed consent.
Emotions, Machine Learning, and Media
As emails (each are a node on the social graph) arrive in an inbox, and the user marks them as spam, a label is being added to the email’s node, and it is compared to an ever-increasing set of previously labeled emails to discern patterns common to spam. These patterns are then used by machine learning to create spam filters to predict the likelihood an incoming email is spam, determining what ends up in one’s spam folder. It is a constant cycle of machine learning, fed by email recipients labeling incoming email as spam.
When a social media user sticks an emoji on a post, machine learning systems are similarly at work discovering patterns in both the subject node (the user, as described by the labels the platform has attached to it) and posts. These patterns in data essentially become a finely grained emotional dossier on the user, which then determine what future content makes the user’s social media feed.
The end goal is continued engagement on the platform. Knowing why a user engages on the basis of economic interests is of value to marketers and determines what kind of advertisement posts make one’s feed. Most users are aware of this kind of analysis. Knowing that a user is a likely voter (a label - information for which is drawn from voter rolls), and that the user is most frequently angered by posts about a certain subject, is digital gold for both media outlets and political consultants (these two fields are fast becoming synonymous). An outlet like Fox News, for example, knows its demographic segment are political conservatives (a label). Knowing which news topics make people in their segment angry or afraid then determines which stories make the news and how they are covered. This is true of left-leaning outlets like CNN and MSNBC as well.
The social graph and emojis have taken "if it bleeds it leads" to a whole new level.
Cable news is now about much more than selling the traditional 30 second ad. The television and Internet media spheres feed each other, with the social media component of the Internet sphere providing the critical information about our emotions. This, then, determines what stories make television news programming. We swing back and forth between television and social media, each tightening the other’s grip on our emotions. And we willingly make this possible when we use emojis.
The business model of Big Tech is simple…
At the heart of information technology is a discipline known to IT professionals as “Knowledge Management” (KM). This is the identification of meaningful pieces of data, then bringing that data into context with other logically related data to create information, and then employing statistical methods on the information to discover knowledge. The two most important concepts from KM are: 1) the textbook definition of the word “information” is “data-in-context-with-data”; and 2) the “social graph” is the context into which a very diverse world of data points are being brought.
If we think of data by using the metaphor of a reservoir, we can understand that on some subjects we have a deep reservoir of data. Hurricane forecasting, for example, has been the subject of KM data collection for decades. Various meteorological data is brought into context to create a relatively complete set of information about the subject. Knowledge (in this case, the likely path of a storm) is then created by an iterative process of back-testing the models.
(Perhaps of interest is the relatively shallow reservoir of data on SARS-Cov-2, resulting in sparse, incomplete information, and therefore volatile knowledge. The models were not “wrong” any more than a calculator can be wrong. The knowledge produced by shallow data and incomplete information was volatile rather than stable.)
To understand how we are being set against each other, we have to take these concepts and apply them to the reservoir of data about us as we use not only social media, but all parts of today’s technological platforms (email, web searches and browsing, etc.) The more we use these tools, the deeper the reservoir of data, and therefore the more robust and complete the information. This produces reliable, stable knowledge about us.
If nothing else is learned here, let it be this: The business model of Big Tech is the monetization of the discovery of knowledge about us.
There is no more valuable knowledge about us than what makes us angry and afraid. The ability of tech platforms to develop that knowledge relies on a universe of data brought into context. As to our emotions - as to what makes us angry and afraid - the part of that universe of data that correlates content to emotion is the emoji.
As complicated as the underlying concepts and statistics are, the point remains powerfully simple. If everyone stopped labeling emails as spam, the ability of machine learning to filter out spam would quickly be crippled. But we like our spam filters, so we have no problem providing that essential data point (the label). The same is true of our emotions: If everyone stopped using emojis, the ability of machine learning to predict what content will outrage and/or scare us - keeping us engaged in the platform - will very quickly atrophy to the point of uselessness.
If we all agree that “[k]eeping the volk at each other’s throats instead of pitchforking the aristocrats” is something we want to stop, we can. Stop using emojis and cripple the ruling class’ ability to keep us at each other’s throats.