Constructing geological models requires a mixture of deductive and inductive reasoning. There may not be an absolute solution to every level of geological question, but there is an absolute solution to a properly framed geological hypothesis at the resolution the data allows. Answer what you can deterministically and implement probability when you need it.
Deductive thinking dominates the foundation building of a testable hypothesis: X equates to Y, Y equates to Z, therefore we predict X equates to Z.
The pitfalls here lie in the validity of the inputs that are used to fix ‘proven’ points in your geological model. These deductive assumptions become the trunk of a hypothesis from which branching probabilistic ranges will be leveraged.
Inductive reasoning takes over when significant uncertainty enters the problem: X usually equates to Y, Y usually equates to Z, therefore it is likely that X equates to Z. Even though the input or fixed points in the hypothesis may be sound and the logical argument for the predicted outcome is valid, the result may still not agree with the prediction. An educated range of solutions is required. Multiple working hypotheses fit in this space.
Bias in geological predictions can come in many forms, and will be the reason the geological model is not useful, fails to identify the critical risk in the system, or fails to capture the range of uncertainty in outcomes.
Deductive failures include misuse of geological principles underpinning the basic assumptions of the predictive model, not recognizing the full range of uncertainty inherent in imperfect geological datasets, and failure to utilize all deterministic geological constraints in the given datasets. The balance lies between recognizing that there are often dead-ends to our thinking that need to be culled, while not pruning the wandering but promising branches of the hypothesis back too far.
Sometimes the interpreter is biased by relying only upon information that is readily available or most easily accessible. For a subsurface geologist, this pitfall could include not being able or willing to interpret seismic reflection data. The lowest risk predictive geological model can only be achieved by the integration of all available datasets.
There is a tendency to try to confirm rather than disprove a preferred geological model. After creating an elegant, novel solution that fits a preponderance of the data there is a desire to see it proven true, nevermind that nagging piece of contrary information. Geological prediction is non-unique. Don’t hide the weakest pieces of your story, highlight them and explain the counter-arguments. Given the many potential outcomes, quite often a geological hypothesis can be harder to prove than to disprove.
Another common bias is the inclination to perceive order where it has not been proven to exist. We seek to find the rules that explain complex spatial and temporal processes that allow us some understanding of the depth of time and earth processes. As geologists we must seek to find these rules, but we must not overreach.
Some ways to get better
• Some will seek subsurface solutions and certainty primarily through application of the latest technology. Spend as much time as possible on ideas — building, testing, iterating — and spend only enough time on software as needed to help you create and execute better ideas.
• Balance the benefits of collaboration with the dilution of groupthink. Curate all available ideas.
• By employing critical examination techniques to facets of your geological interpretation, such as structurally balanced cross sections and chronostratigraphic charts, you will be able to assess your model’s feasibility and critical weaknesses.
• No prediction is complete without some measure of its related uncertainty.