This article uses the term “graph” in a sometimes loose sense to refer to a subset of a population. That said, “graphs” are at the center of the social media revolution and sometimes the controversy over the influence of social media. Graphs are an area of innovation as the TikTok saga has highlighted.
Social & Interest Graphs
Many of the well established social media platforms started with the idea that their architectures should be oriented around a social graph. If A is connected to B socially, they will want to see their content. As business models evolved, and there was a desire to understand people’s interests, it was assumed, perhaps conveniently, that people connected socially have similar interests. It turned out, that this was a weak correlation. I can think of people who I am connected to on Facebook, for example, whom I had shared work experiences with, and therefore bonded, but with whom I have divergent political views, culture views, and leisure time activities. Silicon valley dwellers like me are exposed to some similar influences and experiences, which does mean we overlap in certain critical ways: affluent, embracing diversity, taste for good food, taste for a variety of food, etc. We may even enjoy some similar activities like hiking and traveling. So social graph as interest graph has some utility, but it is a little hit and miss.
A graph based solely on interest is a more precise interest graph, not surprisingly. Do any platforms focus on building an interest graph. One in the news lately does, TikTok. Over the last 20 years, the conventional wisdom about user interface (UI) design emerged: the central focus should be efficiency of interaction / reduction in clicks to get a task done. UIs adhering to this belief, often make many different tasks available on one screen. Facebook is one example of that.
TikTok takes a different approach. One video, not a grid of content, and few application options. This way, TikTok can know more precisely and with greater certainty, that when an eyeball spends time on a page, what the interest is. TikTok trades off more certainty about knowing interests, against the old conventional wisdom on UI design. TikTok is an example of what some are calling an “algorithm-friendly” UI. The TikTok UI supports data mining in a way that assists the algorithm – it is the combination of the two. In information terms, they increased certainty by increasing clarity (making the interest clear).
While TikTok users have a way of seeing a roll focused on who they are following, which might be considered a social graph of types, it also has the default “For You” roll which is used by TikTok to push you new content is believes you may have an interest in, and this is the main focus of the TikTok magic.
The use of a true influence graph, as opposed to a proxy graph (social), may already be a competitive advantage, especially if competitors find it structurally difficult to move to it, from a deeply embedded social graph. It may also be difficult to integrate a company via acquisition that works off a completely different UI and graph model.
Utilizing social and interest graphs to push users content they may be interested in, is one side of the content push coin. The other side is perhaps the aspect that has so many concerned, the desire to influence a person’s view of an issue, candidate, or action.
Target and Influence graphs
It is now clear that political campaigns, among other purchasers of social media services, use highly sophisticated analytics to create target lists. Lists of people in important electoral districts, that have views not yet set in stone, and may still be influenceable.
Once target lists are created, people on the target lists become targets for political information, some may say propaganda, looking to ensure that they take a desired perspective to the voting booth. How effective targeting specific individuals is, when done by either political or commercial entities, is not clear. It is the fear that it will be effective, if not already, that has many concerned about the impact of new data mining and targeting strategies.
Is there are more efficient and more effective way to navigate a list of targets? I don’t know if the following is already in use, but it does occur to me that it is one of the options, which I am calling an influence graph. The following is a theoretical speculation, IDK if it already exists in this way.
Say you have a list of targets. You tag certain targets as being influential / persuasive people. You make those people hubs within a graph (vertices with many degrees of connectivity to others) and you focus on influencing them (the hubs), and then getting them to broadcast their influenced view to the people that a) are also targets, and b) can be influenced by the influencer. I suspect some version of this idea has been used in TV and print advertising for many years.
Figure 1. Simple / crude influence graph
In artificial neural networks, the more a link between two nodes is used, the stronger it becomes (its weight increases). It is speculated that the human brain works in a similar way. So the idea behind the influence graph theorized here, is to not only expose targeted people with messages from a political campaign, but from people they are fans of, trust, or are otherwise influential with them. The theory being that the more a neural pathway is activated in total and due to messages from trusted/admired influencers, the stronger the pathway will become, and the quicker the pathway will become strong.
Clearly the desire to manipulate any thoughts or feelings of voters or consumers is a concern for many, even if print and TV journalism / ads have been doing it too for decades to centuries. There is a concern that computer algorithms will be more effective and more efficient than any previous media.
There are two broad types of graphs / uses of social media. Those that have the aim of sending consumers information about things they are already interested in, and those that aim to push consumers in the direction of believing, or acting, in a way they currently are not. Both have concerns, the first creating bubbles (not discussed in this article), and the latter, creating a plurality of opinion/action based on influencing small slivers of people (targeted populations in important geographies) and/or large populations of people. Concerns associated with these broad groups of graphs and uses of social media will be discussed in future articles.