Making Better Decisions: Discriminating Between Observation and Inference

Returning to our discussion of strategies to eliminate cognitive biases and improve strategic decision making, we arrive at Brewer’s third strategy:  discriminate between observation and inference, between established fact and subsequent conjecture.  The last post in this series touched on this issue, but it is worth revisiting in greater detail.  One of the things that plagues strategic decision making is our frequent tendency not to discriminate between observation and inference and between established fact and subsequent conjecture.  This tendency is normal and virtually everyone exhibits it to some degree.  However, when making strategic decisions, we want our judgments to be based on observation and fact to the maximum extent possible.  When making inferences, we want observation and established fact to support our inferences.  We want our inferences to be likely, not merely conjecture or possibility.  But how do we do that?

The first step is to train oneself to identify when an inference or conjecture is being made.  One way to do this (among many) is to ask whether the information is the product of a sense impression.  Do we have the information because we saw it, heard it, felt it, touched it, smelled it? To return to a first report of injury, the existence of a first report with writing that states the employee reported the injury on Y date is an observation because we saw the report.  When we see the report and hold the report and examine the report, it becomes and established fact.  Whether the employee actually reported the injury on Y date is not a fact.  Instead, if we posit that the employee actually reported the injury on Y date, we are making an inference based on a variety of facts and assumptions (such as the employer is reliable in reporting injuries, has never had an employee dispute the date the injury was reported, etc.).  It is important to recognize that the fact of the first report of injury is different from the state of affairs it purports to represent, which is an inference, however strong.

This distinction even arises in diagnostic imaging studies, which we typically think of as “objective” evidence of injury or the lack thereof, conflating “objective” with “fact.”  The image is a fact, what it signifies is an inference that an interpreting physician makes.  For example, a person complains of a knee injury that suggests a meniscus tear to a treating orthopedist.  The treating orthopedist orders an MRI which does not appear to demonstrate a meniscus tear.  When we evaluate the medical records in the claim, we frequently conclude that if an MRI (or more properly the radiologist’s report interpreting the images) does not show pathology then none exists.  This is an assumption.  The only fact is the images the MRI scan generates.  The simple fact that a radiologist concluded that the images do not show the presence of a meniscus tear does not mean that a meniscus tear is not present.  We know for a fact that MRIs do not demonstrate every meniscus tear.  However, we assume that an MRI is accurate because we know or have been told that MRIs accurately demonstrate the presence of most meniscus tears.  Again, this is an assumption, not a fact.  In our example, the treating orthopedist may perform a diagnostic arthroscopy and find that a meniscus tear is present.  A physician in an IME report recently summed up the problem of conflating what an MRI scan actually demonstrates (observation) with the inference of pathology or lack thereof:

I would stress to the reader that diagnostically the arthroscopic evaluation of the knee is far more likely to be the gold standard of accuracy versus that of an MRI scan…  I would note that there are, of course, instances wherein it can indeed be difficult to differentiate a recurrent tear from a picture of a meniscus that has been previously operated on.  Furthermore, this case is a stellar example of how MRI scans can in fact be inaccurate despite expert interpretation.

In our MRI example, another assumption is being made:  if an MRI reveals pathology, the pathology must be causing dysfunction.  We know this is a questionable assumption based on numerous studies showing that large portions of the population have conditions ranging from rotator cuff tears to “herniated” discs that are present on MRI scans but asymptomatic.  Whether the presence of pathology causes dysfunction is a separate question that the physician makes based on many factors including physical examination, history/mechanism of injury, medical records, and diagnostic imaging studies.  The strength of the inference that a particular pathology is causing dysfunction is determined by reviewing all factors.  The imaging study alone may be enough to make a strong inference, but often more support is needed before an inference can or should be made.

Another example that arises frequently in both the worker’s compensation and liability settings is the conflicting report of injury.  For example, let’s assume that an employee reported to the employer that she did not remember a specific event but had been lifting heavy pipes all day and noticed shoulder was getting sore.  The employee seeks treatment with her primary care physician who refers her to an orthopedic specialist several weeks after the date of injury because the shoulder condition did not improve.  In the initial notes from the orthopedist, the employee is reported to have stated that she was lifting a heavy pipe and noted the immediate onset of shoulder pain.  Obviously there is a discrepancy between the records, but what does the discrepancy mean?  Does the discrepancy mean that the employee is untruthful or that the condition is less likely to have occurred at work?

The established facts in this scenario are that the first report of injury states the condition arose gradually during the course of a work day and did not follow a specific traumatic event while the orthopedist’s notes state that condition arose acutely, following a specific lifting event.  These are the only facts we know.  Any statement about what the facts mean is an inference and is not a fact.  Before drawing any conclusions, I would want to obtain more information.  For example, did the doctor’s office press the employee to identify a specific event?  It would not be unheard of for a member of a physician’s staff to ask the injured worker something along the lines of, “Well, if you had to guess, what incident would have caused your shoulder pain?”  I would also want to know how the injury was reported.  Perhaps the employee said something along the lines of, “I lifted a pipe and felt something in my shoulder.  I kept lifting heavy pipes all day and it just got worse and worse.”  Either piece of information would make the discrepancy in reporting appear less significant.  On the other hand, if there is no indication that the first report is inaccurate or that the orthopedist’s office asked the employee to identify a specific traumatic event, then the inferences that A) the employee appears to be unreliable or dishonest and B) the condition may not have arisen out of the employment are stronger.  The point is that the discrepancy in the records only reflects a discrepancy in the records.  This is our observation and the only established facts.  To the extent that we infer that the employee is dishonest or that the work-relatedness is questionable from the discrepancy, we are making an inference that is not fact.  When making such an inference, must be mindful that other information is necessary before we can decide whether the inference is strong or weak. 

When evaluating claims, it is critical that we distinguish between observation and inference, between established fact and conjecture.  Failing to do so will cause us to estimate the strength and weaknesses of our arguments inaccurately.  If we do not accurately estimate our arguments, we cannot effectively administer our claims.  One way to help ensure that we are accurate in our assessments is to discriminate between observation and inference, to ensure that our conjecture is supported by established fact and to recognize when we lack support for our inferences and conjectures.


log in to comment

Back to Blog
Recent Posts