Predicting the Unpredictable — Human Behavior
What would you say if I told you that every action you take and decision you make could be predicted? You might immediately retort that human beings are unpredictable — that our fundamental nature is random and sporadic. After all, our daily behaviors seem to follow no patterns, and there appears to be no regularity to the actions that we take. Such a question is one which has been frequently debated amongst many scholars. Various intellectuals ranging from behavioral economists, who attempt to quantify and rationalize the decisions made by consumers in markets, to mathematicians, who strive to create new theoretical frameworks, have addressed this question — yet no definitive conclusion has seemingly been reached. While many assert that the inherent volatility of man would make it impossible to accurately predict the choices that humans make, such a statement is only true to a limited extent at the individual level. I would argue that given the emergent properties of human interaction, it is not only possible, but very simple, to predict the behavior of any given group of individuals.
Just four months ago, M.I.T. researchers Max Kenter and Kalyan Veeramachaneni devised an algorithm which enables a robot to instantaneously predict human emotions and even actions (1) to an accuracy of 93% through a meticulous analysis of complex data sets pertaining to any given individual (3). Information such as one’s facial expressions, tone, and past behaviors can all be quantified and inputted into the algorithm to generate a deterministic probability, or single course of action or emotion, that the individual in question is most likely to be conducting or experiencing (3). Granted, the actions that the robot is able to predict are extremely limited in complexity, with the most sophisticated of these actions being one’s movement patterns (1). Despite this, the development of such an algorithm presents a revolution in the field of behavioral mathematics, as previous models had yielded only probabilistic outcomes, as opposed to the deterministic nature of this contemporary algorithm which provides a singular, definitive course of action that an individual is likely to make (4). The creation of such an algorithm has thus reinvigorated debate on the topic of the predictability of human behavior, especially in the context of emergentism.
Emergentism, simply put, is the emergence of complex systems from an aggregation of simpler components and their interactions (7). One particular example of emergentism is in the interactions of elementary particles to form atoms. For instance, protons, neutrons, and electrons — the fundamental building blocks of matter — may interact with one another to form more complex structures such as atoms. These atoms, though recognized as individual entities, are, in actuality, aggregations of simpler particles. Likewise, these atoms themselves may interact further, yielding more complex structures such as molecules. Such molecules may then undergo additional reactions, yielding intricate chemical or even biological systems. In this sense, increasingly sophisticated systems are formed from the interactions of simpler, individual forces or particles. The formation of these systems can be effectively predicted provided the appropriate information concerning the interactions of the constituents of the system (7). For instance, we know that a proton, neutron, and electron will react to form hydrogen, simply due to predetermined interactions amongst these elementary particles. Although emergentism is more clearly understood in the context of particles as opposed to humans, the same principles hold true in various social settings.
The fundamental principles of emergentism are, as it turns out, in effect throughout a variety of human interactions, as is the case with the flow of large crowds, the formation of opinions on socio-economic dynamics, and migration fluxes. Although these examples seem to be quite different, they all share two fundamental properties: First, individuals operate almost always on the basis of a one-to-one relationship. For instance, people attempt to avoid collisions with one another in large crowds, or they discuss political matters with acquaintances and radicalize their opinions. Second, the result of such interactions is the spontaneous emergence of group effects visible at a larger scale; pedestrians walking in opposite directions on a crowded sidewalk tend to organize into lanes and the population of a country changes its political inclination over time. Predicting the emergence of such events is contingent not upon our capability to deconstruct the individual interactions themselves, but rather in our ability to understand how each interaction contributes to the development of the group effect. In this sense, we humans are very predictable; the seemingly unrelated interactions that we have with others all collectively contribute to the establishment of a group wide effect, often times without our explicit desire to do so. And yes, while it may be true that we can never fully predict what any individual human being will do, the outcome of the interactions of an aggregation of humans can be, and already is, predicted to an extraordinary level of accuracy.
The ability to quantify and predict human behavior on a group, and to a limited extent, individual scale has and indubitably will continue to have revolutionary effects on many aspects of our society, particularly with respect to advertising and criminology. Many firms today use internet cookies — information storing digital repositories — to target consumers with advertisements specific to their demonstrated interests (6). An individual’s browsing history and patterns of interaction on the internet may be quantified and inputted into a predictive algorithm to generate a digital “portfolio” of what the individual in question may be enticed to purchase (6). This “portfolio” then enables businesses to target consumers with individualized advertisements, increasing the effectiveness of the advertising process as a whole (6). Such a process is referred to as behavioral advertising, and much of its effectiveness rests in the accuracy of the predictive algorithm. Ultimately, this advertising not only helps these firms to increase their revenues, sustain profits, and ultimately expand their business, but it also helps consumer to find products which greatly appeal to their interests, leading to the most economically efficient allocation of resources.
Likewise, algorithms capable of predicting human behavior have and will most likely continue to be utilized in the field of criminology to assist in the apprehension of criminals and the profiling of individuals likely to commit crimes. Such an algorithm exists today, and is incredibly useful in predicting recidivism in the criminal justice system. This algorithm functions by analyzing various factors related to a criminal’s incarceration such as the age at which they were arrested, the severity of the crime, and the type of offense committed (8). This information is then utilized, in conjunction with a psychological analysis, to determine the probability of criminal recidivism (8). Individuals likely to relapse into criminal behavior may then be monitored and supported, helping these individuals to avoid a life of crime and reducing the criminal element overall.
While these predictive algorithms may be of immense utility in advertising and criminology, they also pose several unsettling questions regarding the nature of our existence and the future of predictive technologies. For one, if our behaviors can be predicted to such uncanny accuracy, do we really have free will? or, do we, as all other matter does, follow predetermined patterns of interactions which dictate our actions? Can such technologies be utilized in more intrusive applications such as predicting and ultimately dictating one’s career path, preemptively incarcerating individuals likely to commit crimes, and profiling political candidates to assess their aptitude? Given the explosive growth that has occurred in the field of behavioral mathematics, there is no doubt that we will be able to speak about these questions in realistic terms in the near future. Yet, there is no guarantee that we will come any closer to unveiling the true mechanisms of human nature. For, as predictable as we humans are, the fundamental forces driving our interactions with one another are, and most likely will remain, as much a mystery now as it will be when we perfect our predictive algorithms. After all, knowing that protons, neutrons, and electrons interact to form atoms doesn’t mean that we fully understand how such interactions are driven. Likewise, we may be able to predict an individual’s behavior, but we will never truly understand what drives such actions in the first place. Such a lack of understanding as to the driving forces of human behavior may frustrate researchers, but it is also a unique refreshment — something infinitely precious in this otherwise discovered world.
- “An Algorithm That Can Predict Human Behavior Better Than Humans — Slashdot.” An Algorithm That Can Predict Human Behavior Better Than Humans — Slashdot. SlashDot, 08 Nov. 2014. Web. 01 Mar. 2016.
2. Goldhill, Olivia. “Mathematically Predicting Human Behavior.” Quartz. Quartz Publishing Co, 18 Oct. 2015. Web. 1 Mar. 2016.
3. “Human Behavior Is 93% Predictable.” Physics. PHYS, 10 Feb. 2012. Web. 1 Mar. 2016.
4. Pentland, Alex, and Andrew Liu. “Modeling and Prediction of Human Behavior.” Neural Computation 11.1 (2013): 229–42. MIT EDU. Mit Edu, 15 Dec. 2013. Web. 1 Mar. 2016.
5. “Predicting the Unpredictable.” Mathematics of Planet Earth. MPE, 23 Apr. 2014. Web. 1 Mar. 2016.
6. FTC Staff Report: Self-regulatory Principles for Online Behavioral Advertising. Washington, D.C.: Federal Trade Commission, 2009. Federal Trade Commission. FTC, 1 Feb. 2014. Web. 24 Feb. 2016.
7. O’Connor, Timothy. “Emergent Properties.” Stanford University. Stanford University, 24 Sept. 2012. Web. 04 Mar. 2016.
8. “Algorithm Predicts Criminal Behavior.” Algorithm Predicts Criminal Behavior. Center for Science and Law, June-July 2015. Web. 08 Mar. 2016.