What Can Statistics Teach Us About Tradition?
It seems that the headnote over my last piece was more judicious than I realized at the time. Several responses—one from Ben Winter and another from Jacob Prahlow (both of whom are authors on this site)—have taken exception to one part or another of my article, with generous asides that they might have interpreted my article incorrectly. In my opinion, this is precisely what happened; I utterly agree with the theological assertions made by both of my critics. Whatever differences we may have concern the interpretation of my essay, or factual errors that they or I might have committed, or small failures in reasoning on either of our parts; these disagreements do not pertain to our root understanding of Scripture or Tradition. Therefore, rather than commit my energies to refuting either response, I will be content with saying that anything I wrote that might have seemed to contradict the theology of either of these men was not intended to do so.
Instead of addressing either piece, I would like to explore a question that has been posed to me several times since I wrote my essay on Scripture and Tradition: “How does one arrive at the correct Tradition?” This question seems to poke a hole in my theology, and to a certain extent to discredit the idea of tradition guiding theology altogether. If tradition interprets all theological evidence, then can we evaluate the veracity of a tradition with any assurance that we have done so correctly? Would not our current tradition (such as it may be) get in the way and falsely color the evidence? For that matter, if the lenses of tradition are colored, how can anyone in any tradition be certain that they are correct, and not merely deceived by the false perspective of their worldview?
I tend to think of the problem as similar to plotting a line based off of a series of points. Raw data can be communicated fairly objectively across worldviews: if you mention that “Irenaeus said ‘x phrase’ in 190 AD,” then that will be understood across a variety of traditions. Of course, how to interpret said phrase is a much more difficult question. However, the fact itself should be uncontroversial. Similarly, data points can be plotted without any disagreement among statisticians. How those points ought to be connected—what the statistical model should be—is a matter subject to debate.
Perhaps tradition is like a statistical model. It helps us put facts in context and understand the facts that we know are true. In that sense, Tradition interprets Scripture (holy, infallible data points) along with other historical facts, such as the teachings of the Fathers (which, though fallible, are useful). At the same time, however, we can only guess at the true Tradition by using the information we already have at our disposal—including, most prominently, the Scriptures.
When selecting a model, we have to avoid several competing dangers. The first danger is over-fitting. This occurs in mathematics when a series of data points can be matched almost perfectly with a very complex, specific, uncommon model. The theological equivalent seems to exist as well: I have, at times, been a prime example. Rather than adopt a well-accepted theological model that generally fits all data points, an over-fitting theologian will fill his or her theology with complicated and novel theories so that every data point on the graph is “fit.” The problem with this approach in theology is the same as its problem in mathematics; a highly unusual and complex model for explaining evidence is more likely to be incorrect (merely by virtue of its novelty and complexity) than a simpler model that less perfectly captures some data points.
There are differences, of course. Tradition does, I would argue, perfectly capture all of Scripture’s data points. However, because of our fallen capacities, even the best model will not perfectly correspond to our perception of the data. If this is a significant, unexplainable difference, perhaps it is time to revisit the model. On the other hand, if the difference is slight, it may very well be due to human error in understanding the data point, not a fault of the model.
It is for this reason that I am skeptical of attempts to construct entirely new models for understanding theology. Early on in my conversion process, I determined that I would not adopt a theological model that fell outside the three main branches of Christianity. The reason for this decision was fairly simple, yet I still think it was well-grounded: if there is no serious Christian that has ever adopted your basic model for Christianity, isn’t it more likely that you have misjudged than it is that God has chosen to make you the first Christian in history to discover the worldview He intends for humankind to use?
At the same time, there is also a danger of under-fitting. Models that are too loose and too vague to offer a helpful interpretation of all (or at least most) of the available data are not worth adoption. The entire reason that statistical models are created is for the purpose of better understanding the facts; if the model does not draw an informative picture from the data, then it serves no purpose. I think we have all encountered theological systems that suffer from similar ailments. These systems are unable to clearly differentiate between the wheat and the tares in a given realm, whether it be one of orthodoxy and heresy, sin and Christian freedom, canonical and apocryphal books, etc. These questions are of such basic and pressing importance that whatever theological model we adopt must deal with them.
Finally, there is the danger of throwing out data that does not fit a pre-determined model. The difference between the convert and the reprobate, I think, lies here. The convert continues to plot facts on the graph, regardless of how well his model can fit them. Eventually, he will realize his model does not match the data and reject it. The reprobate refuses to consider the new data. He probably will say something along the lines of, “I will look into it later,” with no intention of actually following up. His heart is hardened into the shape of a pre-set model, and he will eventually break rather than bend or shift his shape. Pride comes before the fall.
There is a rich interplay between data and models, between Scripture and Tradition, and between facts and our worldview. I cannot give an adequate description of how each influences the other. However, we are aided by a relatively limited number of realistic models, if we remember to avoid over-fitting. For example, a credo-baptist Church that also embraces transubstantiation does not exist; therefore, if that is the state of one’s theology, it probably can be assumed that one has made an error somewhere along with line (even though transubstantiation and credobaptism are not logically contradictory in and of themselves). Therefore, we are not tasked with crafting the perfect worldview to understand the facts. Rather, we look at our realistic options and decide what best fits the data we can see.
I am not very satisfied with this conclusion as a philosophical inquiry. Perhaps it will do as a practical construct, however—at least for the time being. Until I have access to more data points, I suppose that I will have to wait and see.
Image found here.