Studies of technical practices have done much to show the human labor and decision making that is involved in making financial models.  Human beings as sense makers participate in making science and technology.  In so doing, scientific and technological things take on a local flavor – they are made by and adapted to the specific circumstances of their makers. [Human factor A]

There is, however, another ‘Human Factor’ that matters in model making, one that is never far from, but has been less discussed in our field.  This NYTimes article suggest that recent economic events are “an implicit indictment of the arcane field of financial engineering” for not recognizing that human behavior is intrinsically (or so the argument goes) difficult to model. [Human Factor B]

This proposition strikes at the heart of an epistemological question that the social sciences share:  Is it possible to model human activity, and if so, how?  The question is biting.  It means that the plausibility of quantitative finance is the same as that of the sciences of the social.

For all intents and purposes the article concludes that the solution to the failure of models is to make them better.  This has also been the solution of the traditional social sciences, who express their attentiveness to the issue of the Human Factor [B] through a constant concern over methodology.

The Human Factor [B] seems to naturalize the qualities of humans and make them somewhat rigid.  Yet, it is noteworthy that arguments about engineering and performativity would suggest just as the properties of the physical world can change depending on where and how science is done, so can human beings.  Here, we are back to a softer version of the human as a mutable creature that can be designed and changed.

Indeed, as much activity economic or otherwise shows there are clear patterns in the world that emerge through the interlocking of humans and apparatuses – i.e. we may not know the outcome of the election in advance but the winner will be either Obama or McCain; i.e. the markets may be volatile but they only respond in two ways, up or down.  These patterns are not deterministic, yet, even though the outcome of contingent processes, they are nevertheless, real.

The question then is: Is it possible that quantitative financial models of consumer default risk participated in building a marketplace in which human agents with an increased likelihood of default on their loans were created, even if these models were not themselves able to predict this behavior as an outcome…? Hmmm. I wonder.