Inference vs Deduction: what's the difference, and why does it matter for climate change?
We smartly avoid irreversible surgeries, criminal executions and expensive infrastructure changes based on deductions. Inferential evidence is much more reliable.
To infer something is to measure it and then to regard that measurement as a fact or evidence. You can infer your body weight by stepping on a scale. To deduce something is to suppose an unmeasured or unobserved thing has happened based on surrounding indications or circumstantial evidence. A parent may deduce their children are getting into mischief when their usual playful voices emanating from down the hall fall silent for more than a minute or two. It was common for Sherlock Holmes deduce what had occurred in a particular place after noticing a missing coat, soggy shoes, a worn glove, or a smudge of dirt.
We are encounter both inference and deduction routinely in our daily lives. Consider a medical example: a routine physical exam has detected a lump and there is a family history of cancer. Would you:
Schedule an appointment to see a doctor? Yes, having felt a lump, one can deduce there is need to seek the opinion of a doctor and doctor appointments are easy to make and completely reversible.
Jump straight to radical bilateral mastectomy? No, this is irreversible and consequential. Confirmatory tests are needed: there are specialized biopsy tests for inferring if the lump is malignant or not.
Generally people will accept decisions based on a deduction when the decision is trivial and/or easily reversible. But for consequential and irreversible decisions, inference is preferred or even demanded. For example, in the court of law:
Deciding to lock someone up (consequential, but reversible) may be the court’s decision resulting from the defendant’s lack of alibi, coincidental possession of the plaintiff’s property, and motive or access. But without a video recording of the transgression, guilt is merely being deduced.
Deciding on capital punishment (consequential, irreversible) is often not considered by deduction alone. Here is where inference via DNA testing is crucial: if the victim was allegedly attacked by the defendant and the defendant’s skin DNA was found under the victim’s fingernails, the case is much stronger than circumstance alone.
Deductions can be accurate, but not always. The Innocence Project’s DNA testing service has proven 375 times that convicted felons were, in fact, innocent. Inference of innocence will generally overturn a deduced guilty verdict because inference is stronger than deduction. We give especially high regard to repeat inferences (repeat measurements) by independent laboratories, acknowledging that perhaps it was possible that one laboratory mistakenly tested the wrong sample or was bribed, but it is extraordinarily unlikely that different labs in different countries with different versions of testing machinery came to identical but mistaken conclusions.
Climate Deduction
Consider our need for an inference or deduction prior to overhauling the energy infrastructure for nine billion people. This will require many trillions of dollars, therefore it is:
Major
Consequential
Relatively irreversible
It is reasonable to expect that the global warming effects of CO₂ have been inferred, not deduced. But they haven’t been. Thus far, causality has been deduced by eliminating non-CO₂ causes, leaving CO₂ as the lone remaining culprit. Take a look at the animated video over at CarbonBrief.org (screenshots below):
This is the deductive process in its purest form. A process of elimination.
“Eliminate all other factors, and the one which remains must be the truth.” Sherlock Holmes, speaking about his deductive reasoning
A study1 measuring changing ocean heat content concluded that this additional heat was due to rising CO₂, without actually measuring any radiative changes caused by CO₂. This is pure deduction.
This doesn’t mean that climate science is automatically wrong or can be disregarded. It simply means CO₂ hasn’t been “proven guilty” by inferential means. Some might prefer to see a CO₂ test that is as powerful and reliable as a biopsy result or DNA test. Others see no problem with proceeding to expensive and largely-irreversible infrastructure changes initiated based on deduction alone.
Consider the ranking of evidence quality in Evidence Based Medicine (EBM)2, where A is the strongest and D is the weakest:
To earn an “A” or “B” evidence level, a medical researcher has to perform empirical tests since only a test can yield a measurement (inference). Discussing curious observations is anecdotal evidence and only earns a “D” (the lowest quality). This is akin to eyewitness testimony in a court trial, which is known to be flawed or incorrect a large proportion of the time. The 97% consensus3 that CO₂ is responsible for global warming would earn a “C” rating on an evidence-based medicine scale.
Just to reiterate: the fact that there’s “only” a consensus among climate researcher’s deductions doesn’t mean that the consensus is automatically wrong. It could be right, but evidence is flimsy by medical and judicial standards. It is not unreasonable to want top-quality evidence before radical surgery, capital punishment, or a multi-trillion dollar infrastructure overhaul.
What kind of climate measurement is needed?
The best kinds of inferential measurements are those possessing a “fingerprint” characteristic. That is, some set of highly-detailed features that would be extraordinarily unlikely to repeat by chance alone. Your DNA is your “genetic fingerprint.”
To make an inferential case against CO₂ we need CO₂’s spectroscopic fingerprint. For that, we need a spectrophotometer and a gas cell with a lot of mirrors to bounce a beam of light many times through a gas sample. Shimadzu and Thermofisher make them:
If you fill the shiny cylinder with a gas and turn on the instrument, it measures the spectroscopic fingerprint, characterized by many dozens or hundreds peaks at specific wavenumbers:
With measured spectroscopic fingerprints of various gas species, one can infer (via measurement) how much global warming is caused by CO₂. A spectrometer is placed on a satellite and pointed down at the Earth. After continuously monitoring Earth for a couple of decades, the CO₂ levels will have risen enough to produce a detectable change.
I’m not suggesting a 20 yr pause on climate action in what Michael Mann would refer to as delayism or inactivism until the proper satellite can be built. In fact, a suitable satellite was launched and activated in 2003. We already have the measurements.
I’m also not blithely suggesting “someone should do the study…” because it would be ineffective to make such demands. I spent my own time and money to do the study4. If you jump directly to figure 7 (reproduced below) you will see the satellite measurements of changing infrared over time (upper panel a) directly correspond to the laboratory spectroscopic fingerprint measurement of CO₂ (lower panel b).
This is DNA-quality direct inferential evidence of CO₂’s added greenhouse effect since 2003. This type of spectroscopic study can directly measure the quantity of warming attributable to CO₂.
Von Schuckmann, Karina, et al. "Heat stored in the Earth system: where does the energy go?." Earth System Science Data 12.3 (2020): 2013-2041.
Siwek, J., Gourlay, M. L., Slawson, D. C., & Shaughnessy, A. F. (2002). How to write an evidence-based clinical review article. American family physician, 65(2), 251.
Oreskes, N. (2004). The scientific consensus on climate change. Science, 306(5702), 1686-1686.
Rentsch, Christopher. "Radiative forcing by CO₂ observed at top of atmosphere from 2002-2019." arXiv preprint arXiv:1911.10605 (2019).
Very neat first part. Second left me quite confused.
"With measured spectroscopic fingerprints of various gas species, one can infer (via measurement) how much global warming is caused by CO₂. A spectrometer is placed on a satellite and pointed down at the Earth. After continuously monitoring Earth for a couple of decades, the CO₂ levels will have risen enough to produce a detectable change."
I understand this measure CO₂ levels in the atmosphere - something I though was already tightly monitored. I don't get how this tell us assuredly how much greenhouse is happening in a way that's different from what's in the IPCC reports.
Also the final graph would benefit from being converted to global warming in °c to be actually answering the global question.
I'm not so ready to pat you on the back for the smoking gun. It's expected to see IR response to higher CO2. Could even be done in a lab in a cylinder. So, big deal. The question becomes what is the impact on the climate system, which has lots of positive and negative feedbacks involving mass transport, water vapor, clouds, etc.
Really from a Bayesian perspective, I find the inference charts you showed, to be more interesting and compelling. Seeing a bunch of (almost in essence) IR detection of CO2 is more of a truism.