Calculated Knowledge vs. Direct Knowledge

I’m working on a new lens for interpreting the world that I’d like to share with you. Everything you know about the universe can be divided into direct knowledge and calculated knowledge. Direct knowledge is what has been directly observed by humanity, and calculated knowledge is only based on formulas, models, studies, estimates, extrapolations, etc.

For example, if a study finds an extremely high correlation between smoking and lung cancer, an association between the two would be calculated knowledge. If a scientist discovered and described the biological processes that directly turn cigarette use into deadly cancer cells in most people, but not all people, and why, that would be direct knowledge.

Present, Past, and Future

Science contains both types of knowledge. Calculated knowledge is usually built on some direct knowledge, and sometimes advances in science allow it to become direct knowledge as well. The distances to Mars and other planets was originally calculated knowledge built on direct observations about sizes and distances.

Now that we have successfully sent objects to these planets, which travelled at known speeds and took exactly as long to get there as we calculated, I think it’s safe to say those distances now fall under direct knowledge. However, as far as I understand, the distances to most other stars and distant galaxies would still be considered calculated knowledge – although based on the direct verification of previously calculated space distances, you may be very confident that this knowledge is correct. But it has not yet been directly observed. (If you really want to get into calculated space knowledge, check out the Planck mission, or the “invisible” dark matter and energy that have to exist to make our formulas make sense.)

History also is made of both types of knowledge. I would consider the following statement as direct knowledge: “Abraham Lincoln was the 16th President of the United States.” While there is no one alive today who can directly observe this, we have thousands of unbroken chains of people who did observe this and passed this knowledge to people alive today, and absolutely no one who disputed the recording of this knowledge from its beginning or afterward.

However, once we step beyond the bounds of reliably recorded history – and “reliably” and “recorded” are not always easy to define – we eventually cross from direct history into calculated history. All the ages of fossils and rock layers are based on formulas about the way elements appear to consistently change over time.

The future, of course, is almost all calculated knowledge, whether we’re talking about short-term weather forecasts or long-term climate models or any-term political budgets. The only exceptions would be specific scheduled events, such as, “The Supreme Court will hear cases on gay marriage next week,” or “The second Peter Jackson Hobbit movie will release December 13, 2013.”

So What Does All This Mean?

What is the purpose of trying to divide knowledge into these two categories? It reveals some patterns about human behavior. I think calculated knowledge is generally less reliable than direct knowledge. Certainly, direct knowledge can be overridden as mistaken observations are replaced by better ones. We may discover a new document that changes our perception of a historical event. Movie release dates may be postponed.

But as a whole, I think direct knowledge faces less upheaval than calculated knowledge. In this Economist interview about the “half-life” of facts, Samuel Arbesman says “the social sciences have a much faster rate of decay than the physical sciences,” because studies about human behavior are messier than observations about the arc of a parabolic cannonball.

Or, as I would put it, we can directly observe the path of a cannonball through the air and derive formulas that describe it. But social science studies may, for example, only be able to ask people questions and calculate the results; they are not directly observing the neurons in the brains behind those answers. So it makes sense that there is more turnover in those fields.

To overturn direct knowledge, you have to make a direct observation that contradicts an equivalent previous direct observation. To overturn calculated knowledge, you simply have to add some factors that weren’t considered before, or take new measurements and recalculate. Ideally, your updated calculated knowledge will be more accurate – see Einstein’s physics replacing Newton’s, for example – but it may not be any more direct until someone proves it much later.

This also leads me to suggest that calculated knowledge is generally more controversial than direct knowledge. I think the vast majority of disagreements between people have to do with calculated knowledge. Sure, direct knowledge has its disagreements – Who killed JFK? Was the government behind 9/11? – but these tend to be limited in scope and attraction.

You can find many more people – not experts or academic elites, mind you, but people in general – disagreeing about whether or not animals evolved from a common ancestor, or whether or not the earth is rapidly getting warmer, or whether or not minimum wage increases unemployment, or whether or not various gun control measures increase safety. All of these questions rely on calculated knowledge.

To put it another way, it is must easier to dispute calculated knowledge that conflicts with your existing beliefs than it is to dispute direct knowledge. Thus, I do not find it surprising that most opposing worldviews today tend to agree about direct knowledge while disagreeing about calculated knowledge.

Economics, of course, is a field with much disagreement because it is almost entirely built on calculated knowledge. Microeconomics is privileged with some direct observations about transactions, but macroeconomics is pretty much forced to make calculations about complicated results over time to confirm its theories.

So it is not at all surprising that most economists agree on the microeconomics of rent control while sustaining major disagreements about the complex macroeconomic fields of government involvement in the economy and control of the money supply. How can we directly prove something changed the unemployment rate when the unemployment rate itself is a questionably estimated calculation?

So How Do I Use This?

The calculated/direct knowledge divide is a useful tool for interpreting scientific news. For example, there was a study this week claiming that swallows are “evolving” (i.e. naturally selecting) shorter wings around highways as the increased agility avoids cars better. This was trumpeted all over the news, the front page of Reddit, etc. It appears that the researchers 1) found fewer roadkill in an area over time, 2) found roadkill birds a few millimeters longer than non-roadkill birds, and 3) concluded that shorter-winged birds were eluding cars better and that natural selection was favoring them.

Looking at this through my new lens, this screams CALCULATED KNOWLEDGE! I haven’t read the paper to see what they accounted for – so I may be presenting an unfairly non-robust summary from a quick reading of poor news reporting – but I can think of plenty of reasons they might have found fewer roadkill or that their millimeters of measured difference might be irrelevant.

For this to be direct knowledge, they would have to directly observe birds avoiding and/or hitting cars and compare the wingspans of each group; they would also have to measure a statistically significant number of the population’s wingspans over time to tell if the survivors are really getting shorter or if it’s just a fluke.

Of course, that may be impossible. The point is not to belittle the research because it cannot be direct knowledge; the point is simply to explain that since it is not direct knowledge, I am inclined to take it with a big grain of salt.

The calculated/direct knowledge divide is also useful for interpreting political news. For example, Obama may “tell all federal agencies for the first time that they should consider the impact on global warming before approving major projects, from pipelines to highways.” Instead of just considering if ships would “foul the water or generate air pollution locally,” they may also have to “account for the greenhouse gases emitted when exported coal is burned in power plants in Asia.”

I find this to be well-meaning from the save-the-earth perspective, but very questionable from my perspective. Local pollution estimates of approved projects are already a form of calculated knowledge. But now we are taking external calculated knowledge (the earth is getting rapidly warmer), and using that to justify generating more calculated knowledge about how that pollution will affect the rest of the planet. Calculated knowledge built on calculated knowledge!

It’s like social researchers who do studies of studies to try to find broader conclusions… If you combine a bunch of faulty sums, the errors may cancel out, but if they all have the same fatal error, they may compound themselves into greater errors, too.

Now maybe this dangerous journey into epistemology is just a poor attempt to rectify some cognitive dissonance in the way I see the world. Maybe I’m just writing too many words to say old things about the differences between theories and facts or between correlation and causation. But I think this could be a path toward understanding differences and disagreements about all the things we are convinced that we know. Or something like that. I’m open to your thoughts.

6 thoughts on “Calculated Knowledge vs. Direct Knowledge”

  1. I’m not sure I understand the distinction. Is statistical knowledge direct or calculated? For example, is “the mean individual income is $45,326” direct or calculated? On the one hand, it seems calculated because it’s derived by “ask[ing] people questions and calculat[ing] the results.” On the other hand, you also say the evolving swallow researchers could get direct knowledge by comparing statistics, notably the mean wing length.

    Perhaps the distinction is in how the data was collected? Would the claim on mean individual income be calculated if it was derived from survey responses, but direct if it came from bank account balances? That would seem more direct, but it would also be less reliable, as it would not count the unbanked, or people who immediately cash out all or some of their paycheck.

    Then I’m confused about the neuron line–what if we did observe neuronal responses to the question “what is your income”, much as neuroeconomists try to do? Would that be direct because we see the neurons firing, or calculated because we had to use complex statistics and studies built on studies to figure out that this neuron means “one” and that neuron means “apple”?

    1. You bring up good points.. it’s possible this concept breaks down too much to be useful, either entirely or in certain ways. I suppose you could say statistical knowledge like averages are calculated by definition, but it could be considered direct, or as reliable as direct, if it is derived completely from direct values.

      For example, “Scott’s income last year was X” and “Josh’s income last year was Y” would both be direct knowledge. “Scott’s and Josh’s average income was (X+Y)/2” could be considered either – perhaps it’s a kind of singularity where the distinction is irrelevant.

      However, when people say “the mean individual US income is Z” that statement is a little more complicated. Are we including all individuals, including babes? Are we only including people with one or more jobs? If someone unemployed got a job halfway through the year is their income pro-rated or not? If we define exactly who is in the set that is used to define mean income and directly observe accurate values for everyone in that set, then I would say we are calculating knowledge at the irrelevant singularity above. If however we are only observing a subset, either through a survey or even direct observations, then our value can only be calculated knowledge – though you may make a case that your sample is robust enough to make this calculated knowledge as or almost as reliable as a single direct observation.

  2. I’m not sure I understand the distinction. Is statistical knowledge direct or calculated? For example, is “the mean individual income is $45,326” direct or calculated? On the one hand, it seems calculated because it’s derived by “ask[ing] people questions and calculat[ing] the results.” On the other hand, you also say the evolving swallow researchers could get direct knowledge by comparing statistics, notably the mean wing length.

    Perhaps the distinction is in how the data was collected? Would the claim on mean individual income be calculated if it was derived from survey responses, but direct if it came from bank account balances? That would seem more direct, but it would also be less reliable, as it would not count the unbanked, or people who immediately cash out all or some of their paycheck.

    Then I’m confused about the neuron line–what if we did observe neuronal responses to the question “what is your income”, much as neuroeconomists try to do? Would that be direct because we see the neurons firing, or calculated because we had to use complex statistics and studies built on studies to figure out that this neuron means “one” and that neuron means “apple”?

    1. You bring up good points.. it’s possible this concept breaks down too much to be useful, either entirely or in certain ways. I suppose you could say statistical knowledge like averages are calculated by definition, but it could be considered direct, or as reliable as direct, if it is derived completely from direct values.

      For example, “Scott’s income last year was X” and “Josh’s income last year was Y” would both be direct knowledge. “Scott’s and Josh’s average income was (X+Y)/2” could be considered either – perhaps it’s a kind of singularity where the distinction is irrelevant.

      However, when people say “the mean individual US income is Z” that statement is a little more complicated. Are we including all individuals, including babes? Are we only including people with one or more jobs? If someone unemployed got a job halfway through the year is their income pro-rated or not? If we define exactly who is in the set that is used to define mean income and directly observe accurate values for everyone in that set, then I would say we are calculating knowledge at the irrelevant singularity above. If however we are only observing a subset, either through a survey or even direct observations, then our value can only be calculated knowledge – though you may make a case that your sample is robust enough to make this calculated knowledge as or almost as reliable as a single direct observation.

Comments are closed.