Information Spread and Control

и He who influences the thought of his times, influences all the times that follow. He has made his impress on eternity.

Hypatia

In the previous two chapters, an intuitive and fundamental overview of control systems was presented. Here, the same principles will be discussed, but with the goal of applying these principles to information spread. Because there is no one right method to use for any given information spread situation, a certain amount of intuition and creativity are required to design control schemes. To this end, socio-technical systems (of which social media systems are included) are presented as an intersection of social and technical subsystems along with external subsystem influences that must be considered when designing control methods for information spread.

By the end of this chapter, the reader should have a better grasp on identifying possible control methods for social media information spread applications, as well as potential control actions to drive the systems to the desired evolution.

Controlling Socio-technical Systems

There are three main concerns in socio-technical systems when applied to social media: technical subsystems, social subsystems, and external subsystems.

• • Technical subsystems include processes, tasks, required technology, structures, and technical specifications necessary to access and use social media. For example, many people use smartphones to create content for social media, and therefore the technical specifications of a smartphone may be a technical subsystem. The internet, of course, is a required technology for social media access.
• • Social subsystems are concerned with human factors, such as our physical form, skills, values, social structures, psychology, and relationships. Would a miniaturized desktop computer with communication capabilities be as useful or practical as a smartphone? Certainly not. Typing on a tiny keyboard is comical at best, not to mention wildly unusable. Though compliant with the technical needs, such a device would not be designed with people in mind due to our lack of skill to utilize or carry it. A humansounding voice for digital assistants is another good example, as it helps users feel more connected to the object.
• • External subsystems are outside influences, such as laws, regulations, business and profit decisions, politics, and more. Perhaps there are energy efficiency standards that must be met or laws requiring the smartphone developers to allow consumers to choose their own browser (and not be locked into one bundled with the phone). Designers are constrained by these external influences when developing the technical and social aspects of their devices or applications.

The simplest way to think of socio-technical systems is as a Venn diagram of social and technical subsystems with external influences, as shown in Figure 13.1. In this high-level interpretation, social and technical subsystems intersect as socio-technical systems. Note that external influences can affect one or both of the subsystems. Perhaps a law prohibits designing a social media site without notifying the user how their data will be used, while shareholders would like to monetize user participation. Maybe there’s politicized public anger about how the same social media site sells data or is used to spread misinformation without moderation.

In a more dynamic visualization, subset domains of social and technical systems can be viewed discretely. Notice in Figure 13.2 how in each domain interacts with and influences one another. Social domains such as people, culture, societal and individual goals, and societal norms display a coupled bidirectional interaction with not only each other but also with technical domains such as infrastructure and technology.

With this in mind, we can see how special considerations must be met when designing systems for information spread (such as online social media) and in attempts to control these systems. The reason for this is simple: they are incredibly complicated and non-deterministic. Not only are there technical elements to account for (which are largely consistent and understood), but also social elements. These social elements can vary considerably from group to group or person to person. Everyone has different values, self-identity, abilities, opinions, and personalities. A wealthy and elderly individual will see the world differently than a poor college student and interact with social media and the internet differently. Constantly shifting external factors share similar challenges.

Often, one’s goal is to understand the relevant social structures, campaigns, leaders, cultural positives and negatives, and ultimately influence them via control (in a constructive or destructive manner). To this end, one needs to know how the social aspects of the system fundamentally and mathematically

FIGURE 13.1: Social and technical subsystems overlap as socio-technical systems.

behave in order to control it. This behavior is captured in the system models. To apply control theory, a reasonable model of the system must exist or be established. Existing models are very good at explaining socio-technical systems to an extent, but context-aware (or scenario-based models) are sometimes needed. You can see several examples of context-aware models in previous chapters to handle specific instances of social media interactions.

Though some models are intuitive, they must still be run, tested, and validated before control is designed. Since we cannot control based solely on intuition, a model is required. Additionally, the application of well-established engineering control schemes requires closed-form expressions. As such, we must formulate macroscopic mathematical models in the form of equations to apply these control methods. For this reason, the previous chapters focused significant attention on existing closed-form models and their modifications for using them to social media information spread. It is essential to have these closed-form expression models to apply existing control techniques.

For our purposes, macroscopic dynamical models are beneficial for applying these accepted control techniques. Because each individual in a group will have varying whims and tendencies on social media networks, they cannot be

FIGURE 13.2: Social and technical domains interact as part of socio-technical systems.

accounted for individually. As such, they must be looked at in aggregate. In other words, the people must be treated as one group (or at least as a manageable number of subgroups). To do this, we base assumed behavior around “typical” behavior, for example, users reposting stories they are interested in. Obviously, by generalizing and aggregating a population, unavoidable error will be introduced, but the same fundamental control principles will still apply once a good macroscopic model is formulated.

The Control Action and Social Media Systems

As with standard control application, there are two types of control actions commonly used: open-loop and closed-loop. As before, for open-loop control systems, the control action is independent of the process variable. Likewise, for closed-loop control systems, the control action is dependent on the process variable.

But how does one apply a control action to a social media system or information spread? Let’s take some examples from our earlier case study. Consider the previously proposed social campaign model. The process flow diagram is shown in Figure 13.3. Recall the mathematical model developed for the context-aware social media marketing model:

FIGURE 13.3: Flow diagram for event-triggered social media chatter model.

The control input (trigger) here is represented by u(t) and is an external influence driving the system evolution. In other words, each major event sparks social media chatter about the topic. Specifically, the control trigger is acting upon those unaware of the event, hence the u(f)[l — X,] term. Indirectly, via online social interaction, this control action then influences the rest of the system. Remember to always identify what terms are interacting with each other. Do all interactions make intuitive sense?

Naturally, the same principles can be similarly applied to other situations for a more deliberate control effort. For example, a music artist might slowly release songs to the public in order to build and maintain an online media presence to better sell an album. A tech company can tease small specifications of their latest gadget (through rumors or interviews) to keep it socially relevant and discussed on social media. If any of these scenario-specific models have special limitations or differences, they must be taken into account at the modeling level. Advertising the latest book from a prominent author may have more impact before release than after being consumed by the public. Once the book is released, any rumors concerning plot details will be resolved, and the interest will naturally end.

Optimal Control and Social Media

Optimal control methods are particularly well suited for social media-based applications. As with most forms of controlling a system that evolves along a natural trajectory, resources must be spent strategically and compromises must be made. The resources used to control social media systems can typically be reduced to capital expenditure (money spent). However, sometimes resources take other forms such as a grassroots effort for a cause, where emotional appeals evoke more attention than neutral ones. In our case, optimal control methods used on social media systems will focus on capital resources since it is quantifiable and easily visualized. Money is spent as a resource to run internet banner ads, sponsored posts, send samples to review bloggers, sponsor YouTube channels, and hire social media marketers.

The performance measure (or cost function) J expresses the strategy by which these resources will be spent. While everyone would surely like their influence over a social media group (be it for a political campaign, product, or simple public announcement) to be swift, precise, and inexpensive, reality forces us to choose some elements we favor over others. Would you rather quickly educate the public about how to protect themselves during an emergency or would you rather give slower, more detailed guidelines? Is it better to run a set of political campaign ads over inexpensive online ad space for a longer period of time or to flood the public with them close to election day at a higher cost? There is no right or wrong answer, only context-dependent answers. Examples of optimal control in social media will be further explored as concrete examples in Chapters 14 and 15.

Exercises

• 1. Explain what is meant by a “socio-technical system”. How is it different from conventional engineering systems?
• 2. What are some of the challenges to be overcome when trying to apply control schemes to socio-technical systems?
• 3. Choose an example of a socio-technical system not already discussed in this chapter. List some of the elements that make up its social and technical subsystems. What kinds of specific external influences must be considered when dealing with the system?
• 4. Some examples of control actions for social media systems were presented in the text. Give another example of a control action that can be applied over social media groups. Are they practical to implement? Explain your reasoning.