Rabins, The Why of Things

Aristotle suggested that “men are never satisfied until they know the ‘why’ of a thing,” where to know the “why of a thing” is to know its cause(s). Centuries later, causation remains an intractable philosophical problem even as we’ve loosened and redefined the bonds between cause and effect in an attempt to deal with it.

Peter V. Rabins, a psychiatrist at the Johns Hopkins School of Medicine, offers a many-model account in The Why of Things: Causality in Science, Medicine, and Life (Columbia University Press, 2013). According to his three-facet schema, there are three conceptual models of causal logic, four levels of analysis, and three logics by which to gain causal knowledge.

Here I will restrict myself to a brief summary of the conceptual facet of causality—that is, to the categorical, probabilistic, and emergent models.

Categorical logic as applied to causality is binary—something either is or is not the cause of something else, and if it is, it acts directly to bring about an event. This model represents the most common view of causation despite the fact that it is beset by a host of philosophical problems.

In the probabilistic model, common in financial valuation and forecasting, “causes are conceptualized as events that affect the likelihood that another event will occur. In this model, causes act as influences, risk factors, predispositions, modifiers, and buffers.” (p. 45) Here the binary is replaced with the continuous, the categorical with gradations of probability.

There is some support for collapsing these two models. For instance, the “dramatic success of the computer and the digital camera … illustrate that complex, graded phenomena can ultimately be coded digitally. Perhaps nature is constructed digitally (categorically), while humans are constructed to perceive it continuously. Or, conversely, perhaps nature functions continuously, but humans have constructed categorical concepts to simplify it.”

Rabins argues, however, that both models should be kept because they have different functions and strengths (as well as limitations) and because “the choice of a specific model is determined or at least strongly influenced by the circumstances or events being considered.” (p. 57)

The third model, the one most applicable to understanding financial markets as complex adaptive systems, takes an emergent, nonlinear approach. The idea here is that systems are interrelated wholes that require models such as chaos theory, complexity theory, self-organizing systems, and network theory.

What are some of the characteristics of nonlinearity that provide a springboard for both defining and characterizing nonlinear causality?

“First, nonlinear change occurs in systems that have a large number of elements. One or two water molecules would not form ice, nor would a system made up of only two tectonic plates generate an earthquake. The presence of a large number of elements increases the number of potential interactions and increases the probability that an uncommon or unanticipated outcome will occur.” (p. 67)

Following from this first characteristic is a second: limited predictability.

A third, all too familiar characteristic is that outliers are more likely to occur.

Fourth, “some changes that precipitate an event appear to be quite small.” For instance, “the formation of ice and the development of superconductor status … seem to occur after small changes in temperature.” (p. 69) We know, of course, that most “sudden” events occur “after a period of gradual and often unrecognized change or accumulation.” (p. 70)

Finally, nonlinear causality combines “top-down” and “bottom-up” approaches. “The top-down approach begins with a systemwide, big-picture view and identifies interactions at that macro level. The bottom-up approach, on the other hand, starts with the smallest elements and builds a causal explanation based on the interactions at the micro level.” (p. 71)

No single model can capture the activity that occurs in financial markets, no single model can describe the nature of financial markets as a whole. We live in a many-model world where, as the saying goes, all models are wrong but some are useful.