Will the Sun Rise Tomorrow? Leveraging What You Know When Investigating Existing Conditions.

Existing conditions of a site are common and stubborn constraints and sources of challenges for construction projects. Unlike new construction, existing conditions cannot be specified. Existing conditions are often difficult to observe and are variable, creating significant uncertainty and risk. Exploration and testing are the standard means of mitigating the uncertainty associated with existing conditions. This may include visual observation, probing, material sampling and in-situ and laboratory testing. However, investigations are expensive and never fully eliminate uncertainty. As a result, a lot of design professionals have trouble managing the risk associated with existing conditions and resort to excessively conservative design, which increases construction costs and often creates added risks. A more rational approach to the uncertainties of existing conditions can result in more cost-effective investigation programs and reduce construction costs and risk.

In his book, The Signal and the Noise, data journalist and statistician Nate Silver advocates for the application of Bayes theorem, developed by eighteenth-century minister Thomas Bayes in England. Mathematically, Bayes theorem expresses the probability that a hypothesis is true given that some related event has happened. Philosophically, it is the idea that humanity learns about the universe through experience and progressively more accurate models of the world. As more information is available, our models can be revised and improved, becoming closer to the truth with each iteration.

Bayes’s work was published posthumously by his friend, Richard Price. In describing Bayes’s work, Price constructed a thought experiment involving the first man in the world seeing the sun rise for the first time. Initially, the first man would not know whether or not a sunrise is a rare occurrence, but after a few days, he would infer that the sun rises every day. His prediction that the sun would also rise the next day would increase a little each day, converging on, but never reaching 100 percent.

Silver calls the approach used by the first man in the world to infer that the sun rises every day “Bayesian reasoning”. The process begins with a hypothesis and an initial estimate of the probability that the hypothesis is true. These are sometimes referred to as “priors”. Priors can be quantitative, but this terminology is also applied in a qualitative sense by data journalists and other non-mathematicians to describe political science and economics hypotheses. The occurrence of a conditioning event provides an additional piece of evidence pertaining to the hypothesis. If the conditioning event contributes evidence in favor of the hypothesis, then the probability that the hypothesis is true increases from the prior probability. While Bayes theorem relates probabilities, it can be applied to estimators of random values or even qualitative events.

Bayesian reasoning can be a useful tool in evaluating existing conditions. Any available information can be used to derive hypotheses about unseen conditions and properties that must be investigated. Surficial geology maps, visible site features and local experience will provide clues as to the conditions to be encountered in borings and test pits. Similarly, preliminary superficial observations along with knowledge of historical and local practices can yield an idea of the likely structural systems. The hypotheses developed from available information together comprise a working model of the system of interest and its response to the proposed construction. The model and a measure of confidence are the priors in this context. The task of the exploration is then to confirm the working model and fill in details.

In the old days, geotechnical engineers and “soils engineers” before them did something like this somewhat unconsciously. This is the “engineering judgment” aspect of geotechnical practice, in which the individual engineer’s experience is the basis of priors for use in understanding what they observe in the field. Some geotechnical engineers still operate this way. Good structural engineers with experience with existing structures also use a sort of Bayesian reasoning. They can apply available documentation, preliminary analysis and observation to determine what questions they need to have answered in the field by means of probes and testing. Structural engineers have an advantage in that they are typically the end users of the data collected, strengthening their confidence in their priors.

In any discipline, it takes experience to apply Bayesian reasoning well. The need for experience and the cost of experience is probably why too often investigations are approached without it. Instead, a common approach is to largely ignore information apart from the data obtained from explorations and testing. For example, many geotechnical firms do not group soil data by geologic origin and as a result, exaggerate the variability of the data. The result of this kind of approach is familiar to many design professionals: explorations and testing scoped to win a low-bid contract, or to utilize the testing agency’s preferred techniques, or to provide some arbitrary idea about spatial coverage. These investigations are often inconclusive or make conclusions that are little better than a good set of priors. This leads to conservative or arbitrary recommendations rather than a project-specific interpretation of the data and risk-targeted insights.

Silver compares Price’s sunrise example of Bayesian reasoning to the counterargument by contemporary Scottish philosopher David Hume that you could not make a rational prediction about something such as whether the sun will rise because you could not be absolutely certain of the outcome. Engineers have to work in the physical world, however, not the philosophical world. The fact is you always know something about a structure or site and this available information should be exploited in planning an investigation and used as context for subsequent observations and data. Properties for most materials will usually be distributed in a particular range. For example, sound structural concrete usually has a compression strength greater than 2000 psi. Certain structural systems are more common in a given application than others, such as most single-family homes being conventional wood frame construction. Conditioned on hypotheses such as these, you can determine what you need to find out to answer the questions at hand. For example, based on priors, a qualitative assessment – an element is damaged, distressed or not – may be more useful and less expensive than performing a representative number of tests for a particular parameter. But you have to have some concept of the existing conditions to make this judgment.

Investigations of existing conditions are expensive and should not be planned arbitrarily. It always makes sense to make use of any available information to develop priors when planning an exploration and testing program. Since it is typically not practicable to obtain a representative sample of the condition of interest, the purpose of an investigation should be to confirm or overturn hypotheses derived from the priors. If you assume you know nothing, you have to obtain more data in the field and interpret the results more conservatively because the data obtained is inherently less certain. Few investigators bother to collect additional data to overcome their unwillingness to think about the information they have. Instead, arbitrary conservatism is the tool of choice. We can do better than willful ignorance and conservatism. Experience and thoughtful approaches like Bayesian reasoning can be used to increase the cost-effectiveness of our accounting for existing conditions.

The information and statements in this document are for information purposes only and do not comprise the professional advice of the author or create a professional relationship between reader and author.

Do you need help planning a structural or foundation investigation? Read more or get in touch!