Inference then Observation

Scientific knowledge advances most assuredly through direct observation; through the introduction of advanced technologies and techniques for observation. New observations overturn earlier more simplistic explanations for causality. The historic example is that of miasma, bad gases believed to cause diseases in people who lived next to swamps, where it turned out disease-bearing mosquitoes also took up residence. Screens on windows that should do nothing in the face of these gases, turned out to be quite effective in eliminating mosquito-borne diseases. A Nobel Prize was awarded for measuring ion channels in cells, a process that previously had to be inferred. With direct observation, it is often possible to "see" causation, especially if one can manipulate the processes under observation.

Almost all of the bright objects in this Hubble Space Telescope image are galaxies in the cluster known as Abell 2218. The cluster is so massive and so compact that its gravity bends and focuses the light from galaxies that lie behind it. As a result, multiple images of these background galaxies are distorted into long faint arcs – a simple lensing effect analogous to viewing distant street lamps through a glass of wine.
However, inference, the imagination, must outrun direct observation in the hunt for blockbuster+ products. Innovative R&D must always be one or two steps ahead of observational technologies. We infer causality and in doing so identify the need for more advanced observational technologies and techniques. But this means, of course, that future observational advances may unmask our stubborn persistence with today’s miasmas.

All scientific work is incomplete – whether it be observational or experimental. All scientific work is liable to be upset or modified by advancing knowledge. That does not confer upon us a freedom to ignore the knowledge we already have, or to postpone the action that it appears to demand at a given time. Hill, Austin Bradford (1965). The Environment and Disease: Association or Causation? p. 300

To infer causality one needs to combine many independent kinds of empirical evidence. Since there are assumptions built into each piece of evidence, researchers must gather varied and more ample evidence, each with different underlying sets of assumptions. Great care is taken to account for alternative explanations for individual findings. Inferences require a complex interplay among many lines of evidence, each of which can have its own layers of interpretation. Regression models are peripheral to the exercise.

We step back from cheap heuristics+ in determining causality and make this a personal determination for the investigators. We set up the research environment in recognition of the oft-quoted A.B. Hill’s citation above. Industrial researchers need to know when to act in the face of incomplete scientific understanding. Act too soon and you may get it wrong. Act too late and you may lose the race.

Below are some ‘fun’ ways researchers can get tripped up during inference of causality, taken from Schoenbach.

The Hidden Side of Causality

Temporality There are many cases in medical history of opportunistic diseases, where the pathogen was always present but an external trigger causes it to become harmful. Simply correlating the presence of pathogens before the onset of a disease leads investigators to draw wrong conclusions.
Latency Pellagra typically develops four months after the onset of a niacin-deficient diet. This lag made it difficult to associate cause with effect, confounded by seasonality in that cases of pellagra were higher in spring and summer (when food rich in niacin are plentiful) than in winter (when the disease was incipient).
Infrequency Deficiency in Glucose-6-phosphate dehydrogenase (G6PD), an enzyme, can increase the risk of hemolytic anemia or methemoglobinemia when taking certain types of drugs. G6PD deficiency is found in approximately 10% of African American males, and in 1-2% of males of Mediterranean, Indian, and Asian descent.
Necessity vs. Sufficiency Toxoplasmosis is considered to be the third leading cause of death attributed to foodborn illness in the United States. More than 60 million men, women, and children in the U.S. carry the Toxoplasma parasite, but very few have symptoms because the immune system usually keeps the parasite in check.
Confounding Variables

Investigators attributed the cause of cervical cancer to tissue irritation; others pointed to syphilis, or herpes, or chlamydia; still others found circumcision of the husband to be protective. Today it is believed that cervical cancer is in large part a sexually transmitted disease, caused by certain types of human papillomavirus, or HPV

the_environment_and_disease_-_association_or_causation_a._b._hill.pdf994.01 KB