The Semeiotics of FacebookReasoning about behavior associated with Facebook. Facebook represents a new development in the history of human social interaction. It enables connected individuals to increase familiarity at a pace that was previously impossible without physical proximity. And it is the most effectivenote:1 representative of a new breed of Internet application in which familiarity is the new currency. In this paper I propose a model for reasoning about behavior on Facebook. In general, I am interested in systematic reasoning about the behavior of organisms. Here, specifically, I am interested in reasoning about the nature of interaction with Facebook and the effects that the marks signifying that interaction have upon behavior. The Basis of BehaviorNatural law The behavior of organisms have a wide range. They include, for example, the responses of bacteria in glucose gradients, the propagation of nesting behavior in animals, and the interaction behaviors of our species associated with the development of social order and productivity. There is obviously no gross equivalence between the behavior of our species and the behavior of simpler organisms, but I will take the position here that all such behaviors are the product of natural laws. More specifically, for reasons that lie outside of the scope of this paper, I observe that these behaviors are the product of the same laws. This is contrary to the position taken by many contemporary emergence theorists, who apparently believe that such behavior is the product of magic. That is, we are increasingly hearing claims that such behavior is not predictable or determined by natural laws.(ref.1)(ref.2) PersuasionThe objective. The goal of this paper is to inquire into the nature of behavior associated with Facebook and similar Internet applications. In particular, our goal is to ask in what way Facebook is persuasive. But what exactly is “persuasion?” Fogg (Persuasive Technology(ref.3)) defines persuasion as “an attempt to change behaviors and attitudes.” I take Fogg's use of the term “attitudes” simply to refer to the potential of an individual to behave in a particular way. A change in attitude is therefore a change in the probability that an individual will behave one way rather than another. We are concerned then with the behavior of our species and the probability that individuals will behave in one way rather than another. When we speak of “persuasion” we are simply referring to the effects that interactions within the social network have upon these probabilities. Therefore, the study of persuasion on Facebook requires the construction of some model of these interactions and to show how modifications to them change behavior. In the rest of this paper I will simply identify the factors that go into building such a model using the foundational theories that I am developing elsewhere(ref.4). The actual construction and rigorous systematic treatment of such a model requires a more extensive account and is outside the scope of this paper. My purpose here is only to illustrate, for the general reader, how such a model may be constructed. Reasoning About Persuasion on FacebookHow to build models of persuasion. There has been much research into reasoning with and about probabilities and inductive logic; the logic of inferring future events from past record. We will not elaborate upon them here.(ref.5)(ref.6) My goal is simply to suggest a method of reasoning about persuasion that can apply these techniques: where persuasion is defined as the actions necessary to increase the probability that one behavior will occur rather than another. We do this here in the context of Facebook and its functionality at the time of writing. I will introduce concepts that enable us to identify conventions embodied by individuals. These concepts enable us to characterize the probability that a particular behavior will occur, in a range of exhibited behaviors, when an individual apprehends a mark. All that would appear to be required to change behaviors is that we identify which individuals adopt a target behavior and the marks that caused them to do so. By applying those same marks to others that share the same conventions you will naturally increase the probability of the target behavior. The marks we are considering here are the elements of Facebook. The conventions are simply user behaviors on Facebook. It is important to note that both of these can be measured. We essentially ask to what degree is it possible to identify conventions on Facebook to enable the above model. SemeioticsGeneral theories of signs. Let's begin with a few fundamental concepts that give us a general framework for thinking about behaviors and measures for quantifying them. My model is a semeiotic one. A semeiotic model deals with “signs.” Signs, for our purposes here, are simply individuated experiences of any kind. The subjects of signs are “marks .” Whenever I speak of a mark I include reference to complex marks, marks that are composed of marks. So, from the point of view of a semeiotic model, the elements of a Facebook profile are marks that produce signs in the individuals apprehending them. Our goal is to reason about the behavior that these marks produce. For our purposes, the meaning of a mark is precisely the difference that the mark makes in the world; it is the unique behavior that it produces in its apprehension,(ref.7). This behavior includes any embodied behaviors that create references and relations. This view is a strict interpretation of Charles Sanders Peirce's (1839-1911)“Pragmatic Maxim” which simply states that to identify meaning look for changes in the world. We can now say that a mark has precisely the same meaning for individuals in which it produces precisely the same behavior (what this actually refers to in biological terms we leave unstated here). However, the important point is that some portion of this behavior is recorded on Facebook. This gives us a way to reason in conventional terms about what people mean and, more generally, it enables us to reason about conventions. A convention then is simply the uniformity of behavior produced by a particular mark. So, we can now say, for example, that the profile image on a Facebook profile is a mark and that individuals that share a response to that mark share conventions.note:2 This provides a simple measure that enables us to categorize friends on Facebook. If two friends behave in the same way to a modification of any element in a Facebook profile then they share conventions to some degree. To determine the full degree to which any group share conventions will obviously require more than one data point. Applied SemeioticsUsing the simple model. This is a pretty simple model but how useful is it. Clearly, if we were able to observe all the responses of individuals viewing a particular mark we would be able to group them into categories of individuals that share the same conventions. Let's think of some examples. We could perhaps place an image of our favorite emblematic personality in the profile picture and measure the number of friends we lose and the number that we gain as a result, the number that make comments of a certain kind and so forth. Thus the friending process is a behavior that reflects the conventions held by the individuals joining and leaving in response to the mark. You can try this out by changing your profile image to one of Adolf Hilter, Ghandi or Karl Marx, for example, and measuring the behavioral responses among your existing friends and the new friends you acquire. Those having similar responses share conventions, at least to the extent that they are observable and relate to the particular mark. Considered use of this example may provide a useful mechanism for filtering a friends list. This model enables us to identify which individuals in a group share conventions. As noted, you need more than one data point. Conducting experiments of this kind, and gathering extensive data, you can build a landscape of the conventions held by any given population. It should be clear that if you are Facebook Inc., or any other entity with monolithic access to social data, that you have access to such data and can conduct such experiments. When you have a way to determine which individuals share conventions, and privileged access to behavioral data, you can construct a useful set of predictive categories. Comprehensive data within those categories of convention enables broad predictions and the ability to manipulate marks will enable the management of a society. What are the predictions that such categories enable? Individuals sharing conventions have a high probability of behaving in the same manner when exposed to a given mark. If you want an individual to respond in a particular way, then the simplest thing to do is to identify individuals in the same convention category that exhibited the desired behavior and apply the marks that produced that behavior to the individual you seek to “persuade.” Therefore, once you have assembled individuals into these categories of convention only a little data is required to enable you to change the behavior of any and all individuals in that category. Pragmatic Solutions to PersuasionKeeping it simple. This is all pretty simple stuff and we should not be surprised that technologists have tried to exploit ideas like this. The notion of “collaborative filtering” tends, as a practical matter, to be based on readily accessible statistics, like Amazon's: “those that bought A also bought B.” Let's briefly consider what is going on here in our terms because we see this approach used by a number of Facebook applications. The purchase by any individual of two products is a vote for the relationship between those products as marks and this is a useful fact. It reflects convention if it reflects broader behavior but its use is typically a pragmatic discovery. Such discoveries have obviously been valuable to Amazon. A variety of web services have had mixed results with techniques that attempt to apply probability techniques of the kind I have referred to. Often these techniques claim to possess some form of “artificial intelligence” and they attempt to solve persuasion problems. Simply put, by showing you some mark, for example placing an advertisement in your Facebook news feed, they seek to produce some behavior. Blanket offering of this kind work by sheer force of coverage; expose sufficient people to a mark and you will get the full variety of responses to it. But automated “intelligent” selections typically fail. The question is, why do they fail and can you, with your intelligence, do any better? Are there marks that you can place in your news feed that will produce some desired behavior in those apprehending them? If you can succeed where the probability models fail, what enables you to succeed? The services that try to apply advanced probability techniques discover that the data they collect is simply incomplete. They cannot know, for example, about the books that you buy at other locations online or offline. So, even if they succeed in making offers that are 100% relevant, making offers that are redundant is a problem. This is especially the case if the offers actually make claims about the “intelligence” of the selection. An Internet service that makes intelligence claims about redundant offers (offers that would interest you if only they had not already been fulfilled) is viewed negatively and this negativity propagates to the entire service. In the worst case, a service that makes intelligence claims and makes offers that are relevant anything less than 100% of the time is considered “stupid” regardless of the fact that it is right most of the time. note:3 So, even though we have a compelling theory, we are limited in our ability to collect the full range of data required for it to be useful as a comprehensive persuasion tool. There are practical matters that limit what we can know and this limit the application of the theory. It remains an open and Orwellian question as to whether these are hard limits. The simple pragmatic “those that bought/liked A also bought/liked B,” used by many Facebook applications, is not the product of a foundational theory. It is a pragmatic discovery. It makes no claim to intelligence, it is simple, informative, and allows the acceptance of redundancy. In this case redundant offers become affirming experiences. We say “Oh yes, I have that book and if people that bought that book bought this book too, then this book is one that is useful to me!” This same response encourages you to consider those other offered books that are not familiar. A General Theory of BehaviorExtending the model. What can we do with the minimum data that can be collected? How are our offers (the marks we use to evoke desired behavior) best presented to be effective? We are looking for two things: the target behaviors to which the theory can be applied, and why certain applications of it will be ineffective. What we need is a general theory of behavior that will give us a more complete framework. Using this framework we may be able to identify reliable and futile mechanisms of “persuasion.” While a semeiotic model deals with the operation of individual sentient entities (any kind of organism, including Facebook users), we need a more general way to consider the behavior of sentient entities in groups. I call such theories “Natural Ethics.” Natural Ethics deals with the inevitable behaviors of sentient entities in groups (it deals with what they will do, not with what they should do). Facebook users are such a group. Natural Ethics have two components: genetic disposition and convention. We discussed convention earlier, now let's consider the dispositions of individuals that are defined solely by who they are physically and, in particular, ask which of these dispositions are reflected in behavior that is influenced by interaction with Facebook. There are a couple of obvious candidates, so for the sake of brevity let's focus our attention upon them. They are:
These are the two indisputable genetic dispositions. We can potentially add more but let's keep it simple. Note that I have said nothing about broader social behaviors here. This is essentially because the model treats individuals as the element of any broader social behavior and takes the position that any consideration of community is the consideration of individuals. By this means we avoid metaphysical pitfalls often associated with reasoning about social behavior. We reason about broader social behavior when using the probability mechanisms described earlier. These allow us to work from the basic assumption that individuals that embody like conventions behave similarly. I then take the position that in the absence of convention the behavior of all organisms in species is inevitable, determined by their physical form in the environment in which they find themselves. This position always raises the question of “free-will.” So for clarity I will note that our definition here of free will is simply that it is the navigation of ignorance; if we always knew the right thing to do, then there is no doubt that we would do it (and this applies regardless of a hesitant disposition. I am referring here to an absolute knowing of “the right thing”, as certain as any natural law). Our “choices” are determined by the behavioral model I have described: genetic disposition mitigated by convention. It is reasonable to anticipate, therefore, that in the absence of convention organisms will mate and eat. Or more generally, they will reproduce and sustain themselves. Conventions serve to make this behavior more orderly, more efficient. How are these dispositions manifest on Facebook? The most obvious genetic disposition is, perhaps, the mating disposition which is reflected in the range of interactions around reproductive behavior. But we also find behavior designed to meet the eating disposition. The outcomes then, from the point of view of genetic disposition only, that conventions present on Facebook potentially mediate are:
Without convention we can reasonably assume that these dispositions prevail and that the drive for them without convention is determined only by the degree to which these needs are currently met by the environment. Facebook, like all media, is a vehicle of convention only. It does not fulfill these dispositions directly. So to speak about the behavior surrounding Facebook we have to ask what role convention plays in the characterization of these dispositions. Quite simply, in a general theory of organism behavior, conventions (as we have defined them in the foregoing) mitigate genetic dispositions. Here “mitigation” refers to the increase or decrease in the behavioral effect of these natural dispositions, in the range of behaviors that are the possible products of the disposition. Conventions do not bring new behaviors into the world. They only modify behaviors that are the product of genetic dispositions, behaviors associated simply with biological structures in their environment. FamiliarityThe role of the familiar and how we deal with that which is not familiar. In terms of Facebook then, individuals respond as they would to any other mark in their interaction with the world. Facebook is more effective than a traditional form of social communication, say by posted letter, because it provides a vehicle for rich media and interaction mechanisms that are always available, immediate and reflect interaction behavior between proximate individuals in the world. This rich media increases our familiarity with their subject. It will be argued, of course, that other services provide rich media and interaction mechanisms, so what is different about Facebook that enabled its wide adoption? Fogg and others note correctly that Facebook has provided an environment of “trust.” It has done this by enabling transparency and eliminating anonymity. This is important because it more accurately reflects the familiarity protocols of the natural environment. The environment of trust arises from the greater familiarity that Facebook enables. Familiarity is a property of “semeiosis”, which for our purposes here we will simply define as the operation of the mind; it is the processing of signs. It is what is going on inside the individuals in the model I have described so far. Familiarity is simply the degree to which we have adopted a behavior in response to a mark.note:4 This is initialized by the innate familiarity to the environment provided by genetic disposition and subsequently modified by convention. We know intuitively how to deal with that which is familiar, by definition. We rely upon convention to help us deal with that which is not familiar. In this case, the marks of convention identify categories that we become familiar with, and those categories modify our behavior in the variety of circumstances that involve that which is not familiar. So, for example, we trust strangers to treat us in times of crisis because they are identified by the mark “Doctor” or “Nurse.” We select strangers to attend to our teeth and teach our children, all by convention; common responses to marks that help us deal with individuals whose services we need but who are in all other respects not familiar to us. These same dynamics apply on Facebook. Facebook enables us to reinforce our familiarity with those we are already familiar with. And it enables us to strengthen our familiarity with those we are only weakly familiar with. Since we rely upon familiarity and conventions that allow us to navigate that which is not familiar, Facebook's environment as far as it ensures transparency and eliminates anonymity is a good and valuable social tool. And in general, this mechanism of convention provides social animals, such as our species, with extraordinary advantages. Navigating the UnfamiliarThings to watch out for. However, the mechanisms that provide us with these advantages are also the source of great risk. A mechanism provided by Facebook, or any similar platform that undermines our ability to rely upon these conventions, inhibits our ability to navigate the unfamiliar. Conventions are undermined when unexpected events occur that force individuals to reassess their response to a mark.note:5 In these circumstances they will pass from predictable probabilistic behavior to random behavior when the mark appears. Alternatively, if other conventions were previously embodied, they will revert to them. Imagine this situation in a non-virtual environment. You arrive at a party and there are many people there that will not tell you their real name. A few claim to be qualified doctors and nurses, priests and elected officials. If you are a conventional Westerner you won't be able to help yourself. You will trust the “qualified individuals” and distrust the anonymous individuals. Such are our conventions and our innate dispositions. However, the true situation at this party is that the people that are anonymous are in danger. They are simply protecting themselves and they have something important to tell you. The qualified individuals are imposters. Before the Internet it was the convention of “free speech” that anonymity was the haven of last resort for people at risk. Individuals claimed anonymity only as a cloak of protection for their safety and that of their families. It was a valuable convention because it enabled them to say what could not otherwise be said. It enabled them to alert us to betrayal. When the conventional cloak of anonymity is made ineffective because it is so readily available it becomes impossible to detect the sincere alert to danger. It denies protection to those that truly need it. As a consequence it undermines the freedom of speech in a society. Wikipedia is an example of these dangers. It is impossible to determine the real identity of contributors and therefore their conflicts of interest and the true merit of their contributions. Wikipedia is useful at points in its uncertain history and this lulls us into a false sense of security. The sources have neither the advantage of being truly familiar, nor are there attributable conventions that would allow us to trust the unfamiliar source. The great strength of the Facebook model is that it provides transparency and thus a relatively secure environment for identity; people are who they say they are. This encourages individuals to strengthen their familiarity with each other. It should be obvious that it takes a lot to undermine a convention and that familiarity has great power. Someone that poses as a qualified dentist and later turns out to be qualified as a car mechanic is unlikely to be trusted to care for your teeth again; no matter how good a job they actually do on your teeth. But this is unlikely to stop you from seeking out a new dentist on the same basis as you used to select the first. Conversely, if you become sufficiently familiar with the car mechanic then the strength of that familiarity allows you to reject convention and allow the car mechanic to continue to treat your teeth. We all know how to deal with that which is familiar, by definition. You are more able to navigate that society, and you are more productive in it, the more familiarity you have. If you are not at all familiar within that society then you are totally dependent upon convention. If you are familiar you will have less dependence upon convention. Our dependency upon familiarity and conventions for managing that which is not familiar makes us vulnerable in environments where convention can be undermined and familiarity falsely acquired. Celebrity, incidentally, is simply the product of familiarity. The power of familiarity is readily seen. For example, in the election of the California Governor the policies and performance of Arnold Schwarzenegger were and are irrelevant. This is limited to some bounds that essentially require that Schwarzenegger maintains familiar behaviors. His broad familiarity then makes his re-election as California's Governor simply inevitable. To beat Schwarzenegger has nothing to do with policies. It simply requires a candidate that is more familiar or uncharacteristic behavior on the part of Schwarzenegger. ConclusionI introduced foundational theories of behavior and applied them to Facebook as a medium of social interaction. I observed that the primary strength of Facebook, and this may be the cause of its success, is that it enables the currency of familiarity by providing a platform for its development and use. Specifically, I first described a simple model of behavior: individuals act in response to marks. I then generalized this idea to enable the identification of convention as the uniformity of responses to marks in groups of individuals. I then noted that we rely upon convention to inform us in our dealings with that which is unfamiliar. According to this model Facebook's advantage is that it enables the strengthening of familiarity. That, in turn, strengthens social interactions and enables mechanisms of persuasion. It should also be clear that the dependence upon convention is reduced for those rich with the currency of familiarity. The establishment of “conventions” by groups of familiar individuals assists in the navigation of that which is not familiar. I noted that the measurable outcomes from interaction through Facebook include dinner and the birth of a child. These are measurable outcomes of any social environment and under any circumstances (by definition) satisfying and productive social environments, such as Facebook, have improved probability of these outcomes. They will occur where they would not otherwise have occurred. I have suggested that with broad and sufficient historical data it is possible to identify what you need to modify on your Facebook profile to achieve these outcomes. I have also added a note of caution. While there are clear social benefits for individuals to the increased liquidity of the currency of familiarity, a platform for that currency, according to this model, enables social behaviors to be observed, predicted and modified. Such a platform is potentially the ultimate tool of mass persuasion. | ||||||
End Notes
note:1 As measured by its growth and commercial success.
note:2 We, as a culture, use the notion of “meaning” as a way of speaking about the behavior associated with marks. We say “What does this mean?” or “What do you mean by that?” In all cases when we speak of meaning, though it may not be immediately obvious, we are speaking about the behavior associated with a mark, the behavior a mark produces in its apprehension. We would do better to say “What conventional behavior does this mark suggest?” or “When you exhibit that behavior what response do you expect?” Some readers might object that this does not allow us to reason about relations or references but this is not the case. Both are a behavior embodied in the engineering of sentience and they are the products of apprehending marks.
note:3 These observation are the product of the author's experience with users while building such services for RCA, Microsoft, Oracle, and others.
note:4 Charles Sanders Peirce would have referred to this degree to which we have adopted a behavior as “habit.”
Concepts
genetic disposition: I am referring to the inevitable behavior of an organism in its environment as determined by its physical structure which, as we know, is a function of genetics.
inevitable: I define “inevitable behavior” as behavior that is the product of the engineering of sentience. It differs from “deterministic behavior” only in so far as engineering by classical mechanics differs from the engineering of sentience. I will not go into the full reasons here why I draw such a distinction but it should be clear that there are, in fact, such distinctions at least as far as it is evidenced by the engineering of organisms according to genetics.
mark: The subject of a sign.
meaning: The difference a mark makes in the world.
Natural Ethics: The inevitable behavior of sentient organisms in groups defined by genetic disposition mitigated by convention.
semeiosis: The operation of “the mind,” the mechanics of apprehension, of sign processing and response, of sense and motility
semeiotic: A general theory of signs. The study of the foundations of logic and apprehension.
sign: An individuated experience.
