A recurring topic in this blog concerns the nature of science and the nature of data. I usually talk about one of the other as if they are separate but they are actually different frames to view the same phenomenon.
Modern discourse treats both data and science as two fundamental concepts. Data records observations. Science explains nature. They are not separate concepts. We demand from science that it must explain the observations. We also demand from data that it massages the observations to conform to science.
The more fundamental concept below both science and observations is ultimately about decision making from everyday personal decisions all the way to the government policy decisions. Everyone wants to make the best decisions possible but each has their own objective of what is best.
As humans, we tend to default to allowing our emotions or instincts to guide our decisions, especially the condition is an urgent one, or the condition is distracting us from what we would rather be doing. We default to emotions or instincts in periods of fear or in periods of impatience.
Although the modern world denigrates these types of decisions as irrational, I tend to respect such decisions because deepest form of wisdom is ingrained at a level that rational processes are no longer required. We do recognize this with highly trained professionals across a vast range of specialties. We reward them for their ability to perform well even when unexpected things happen. This ability usually comes from extensive and sometimes punishing training in the past. Part of that training is memorizing, not just verbal information, but muscular actions, something we call muscle memory.
When a surgeon prepares for a patient, he may have a specific plan in mind based on all the prior testing and imaging results. Once he starts cutting, he may discover something is not working according to plan. At that time, there is no time to act purely rationally such as backing out and rescheduling until a new plan is made to account for the unexpected difficulty. Also, such an action may cause more harm than good. The best approach is to move forward, changing plans on the fly guided by that wisdom.
I believe similar wisdom is available to everyone. The reaction is emotionally driven but the emotions are trained through experience. There is a big difference in wisdom available to a young child’s emotional reaction compared to the older adults. Even so, there is always a kernel of wisdom in even the child’s reactions.
In most areas of life, we have to collaborate in groups to accomplish some shared goal. Such groups need their own decisions. The group may accept the guidance from a selected leader who may make decisions emotionally with the expectation that wisdom drives the emotion. I believe there are many scenarios where groups are led very well in this deference to the emotional decision making of a leader. The primary problem with such groups is how to handle succession when the chosen and proven leader needs to be replaced. Inevitably the successor’s emotions work on a different form of wisdom and very often a less capable or less developed wisdom.
The modern world prefers more explicitly rational decision making. In addition to asking a leader for his decision, we demand that he defend his reasoning. The leader needs to convince us that we would make the same decision with the same facts.
This demand for rationality works both ways. The leader must defend his decision based only on facts and rationality. Everyone else must judge the same facts with the same rationality. Neither than leader nor the followers can think irrationally or emotionally (even if there may be wisdom in that irrationality).
I concede a scenario where all decision making is perfectly rational. The leader has thoroughly considered all the facts and applied non-fallacious reasoning to those facts. The population accepts the leader’s rational defense of his reasoning.
Even in such an ideal state of affairs, there will be competing options that are also well defended with the same facts and rationality. Rational decision making does not solve the problem of disagreements. The benefit comes in the fact that the ultimately chosen decision will rational. We expect rational decisions to be good decisions. Perhaps more accurately, we expect all good decisions to be rational, so selecting a rational approach has a better chance of succeeding than selecting an irrational approach.
As an aside, we always reinterpret in rational terms those decisions that turn out well even though they were irrational at he time the decision was made. Retrospectively, all good decisions are rational ones. Our expectation is that all future rational decisions will be good decisions. That expectation conveniently forgets that many past irrational decisions turned out to be good. Also, many past rational decisions turn out to be bad.
Back to my idealized rational decision making world.
I think we have a good consensus of what constitutes valid rational thinking. As mentioned already, the same rational thinking can come up with competing and conflicting decisions when considering the same facts. The reason is that the facts are incomplete or ambiguous. We don’t have all the information, or we don’t agree on how much weight to assign to each category of fact.
I recall simplistic challenge I have every day when preparing to walk to town to get groceries: do I wear a jacket, and if so, which one? I check the current conditions about cloudiness, windiness, temperature, and whether it is raining. I also check the online forecast that includes probability of participation and radar maps. I may even go out for a quick test walk to confirm I made the right choice. Invariably, the decision is not conclusive. Often, I choose wrong. I end up carrying a jacket in my hands because it is too warm to wear, or I end up dealing with a chilly wind that cuts through the thin jacket. Other times I do choose right. I’d say that I choose right more often than I choose wrong, but I can’t deny that I often choose wrong.
Government-level decisions require a lot more analysis than my simplistic example, but it is very similar in needing to decide between multiple well-supported choices and in subsequently dealing with the decision not working out as well as expected.
The above discussion presumes that everyone agrees on the facts. In this imaginary world, we may have varying opinions on the relative merits of the facts, but the assumption is that we agree to use the same set of facts. Applying rational processes on a mutually accepted set of facts will result in a rational decision and we expect that will result in a good outcome more often than not.
Given these constraints, there are still multiple possibilities based on different qualifications of what are relevant facts.
We idealize our current government as using only facts allowed by science. Science refers to accepted principles that previously were extensively tested to be accepted as true. As a result, we accept as facts any derivations from the theories. We will include fresh observations after removing observations that conflict with the theories. The science-focused decision making primarily use these observations to supply initial conditions for calculations based on the theories. In other words, the decision making is guided by previously accepted scientific theories made relevant to the current circumstances by supplying up to date observations.
In this blog, I describe an alternative extreme of placing emphasis on observations. Science can augment observations, but this form of decision making would discard any science that conflicts with the current observations. I describe this as bright data (actual observations) being preferred over dark data (theory-derived values). This is the opposite of our current approaches. In stark terms, observation-preferred decision making accepts correlations before causation can be proven, and even discarding prior theories of causation if it conflicts with the correlations.
Currently, we would reject any leader that proposed to follow some correlation before it is proven to be causal, or even when the correlation conflicts with the causal explanation. We demand from human leaders that they accept the constraints imposed by the currently that established science.
Meanwhile, we increasingly permit machine intelligence to make our decisions based on correlations. In particularly, we accept turning over jobs previously held by humans to trained neural networks. We never demand from these neural networks to defend their learned world-view in an rational argument. We simply accept that they tend to choose well about as often as humans do.
The idea is to use a decision making approach that prefers current observations over previously tested scientific theories. This idea comes out of the success of neural networks that construct world-views from guided training using logic that is unconstrained by prior scientific theories about nature. Successful neural networks prove their understanding of nature by the quality of their decisions. We don’t demand that they explain their decisions based on rational arguments and scientific theories. Fundamentally, the neural networks could not make such rational arguments even if we demanded it. They are coming up with theories that are irrational to humans but they are coming up with answers that are at least as good as humans can come up with.
The machines are coming up with good answers based solely on observed data, data used to train them, and current observational data for them to make new decisions.
We punish machine learning when they come up with bad decisions, such as what happened in the recent example of the Boeing 737Max MCAS system. When such systems fail, we send them back to get fixed. On the other hand, for the most part, when machine learning succeeds even in unexpected ways, there is nothing for us to celebrate. The machine is incapable to teaching us what it knows about the world that led it to making such an unexpected decision. We may try to figure it out on our own, and when we do we credit the human scientist instead of the machine that first figured it out.
There is emerging two competing versions of governance. The old one based on human leadership where the modern versions are based on democratic elections and demands for rationality. The new one is based on machine intelligence processing current observations and self-learning based on those observations, unencumbered by rationality or prior science.
Because both systems exist and have competing approaches to building world-views, we should fairly compare the two. When irrational machine-learning comes up with a better result than the rational human approach, we should credit that machine approach as superior to the human approach. More specifically, if the human approach leads to bad decisions, we should severely penalize the human approach especially when the human approach neglected to take into account current observations that conflicted with the established science.
In our current preference is to be guided by science and reason. This approach tends to greatly celebrate and reward the successes of science, thus reinforcing the value of science. Major failures of the science approach rarely gets blamed on the science. Usually such failures are traced to human errors such as engineering errors or operator errors. The science would have been right if the humans didn’t mess up.
I mentioned earlier that my initial career experience was in systems engineering where human factors played a big role. The human factors perspective turns around the blame. If the problem actually was due to an operator or an engineering error, then the problem was a technical one of not giving these humans the appropriate warnings or of overloading them with irrelevant warnings. This partly turns the human error back into a science error in that science initially failed to account for the human’s capabilities.
I guess this perspective has grown on me over the years so that now I am more comfortable assigning any errors based on scientific recommendations to be errors in the science. The science can be wrong. In my experience, this happens frequently when I fairly respect observations recorded in data, and in particular respecting the outliers.
Our present form of government can be greatly improved if we were to be more quick to hold science accountable for any failures that result from following the suggestions of science.
In the current COVID19 situation, I see multiple areas were science can be criticized for advising bad decisions, but we instead applaud science for its guidance.
- Computer modeling predicting much higher rates of hospitalizations that would lead to shortages in medical systems
- contradicted by actual cases being well within capacity
- resulting in postponement of elective procedures and screening that will lead to future overload from backlogs and from advanced forms of diseases
- resulting in causing widespread panic and fear that increases the entire population’s stress levels leading to less resilience to infection and more eagerness in seeking medical help that always comes with the added risk of making matters worse
- Computer modeling recommending closure of everything considered non-essential yet these models did not take into account the dependencies of essential businesses on the non-essential businesses. For examples:
- to supply diesel for trucking, we need a good market for gasoline byproduct of producing diesel.
- to keep food supply chains running, processing plants need the business of a steady demand of excess products from food service industry such as restaurants
- Excess production is leading to massive food waste that will in turn result in producers cutting back future productions thus guaranteeing future shortages
- Production in general is scaling back in response to drastically reduced demand from the closed non-essential businesses
- Scientific approval of deciding the difference between essential and non-essential
- Inherently deciding the winners and losers in commerce, and this is inherently biased to the larger corporations and monopolies, permitting them to get even larger and even more monopolistic.
- In particular nearly all small business and sole-proprietor businesses are automatically assumed non-essential and forced to be idle for extended period of time
- Businesses will fail permanently, and future businesses will be discouraged from investment with the new precedence that the government considers the business non-essential and thus subject to sudden closure.
- Workers are denied the opportunities for building their careers with real experience in addition to having disposable income to improve their well being.
- Evidence-based medical policies that may have mistreated many early patients, likely being the ultimate cause of many deaths
- Low blood oxygen triggered ventilator use even though there was no respiratory distress
- Infectious nature of virus, ruled out less intrusive ventilators, forcing the more risky invasive types that can cause fatal problems in general and may especially exacerbate conditions for the COVID cases without respiratory distress.
- Hospital admissions for any reason requires COVID testing where positive tests assigns patient to COVID wards where patients receive less attentive care due to fear of spreading the virus, but also where false-positive patients can catch virus from others
- Scientific assurance that not-severely ill people to return to their homes irregardless if the home was shared with more vulnerable family members
- Assisted care centers for elderly required to accommodate infected patients resulting in increased opportunity to spread among the more vulnerable
- Communities with a culture or an economic necessity of multi-generational households results in increased risk of the vulnerable getting the disease
- Both were compounded by the scientific approval of people staying home all day, every day of the week and staying indoors is more conducive for the virus to spread and for the conditions to rapidly deteriorate
- Scientific recommendation of enforced social distancing
- Closing of recreational facilities and outlawing of recreational activities (outside of essential dog walking) preventing natural form of exercise that improves overall health including immune response
- Lack of social gatherings in communities and in meeting places will result in deterioration in mental health or at least a decline of networks of friends and acquaintances that are a necessity for future life opportunities
- Policing and culture-enforcement of 6-foot distancing plus wearing face masks are major impediments of people being able to greet each other and interact in friendly ways and also major encouragement of distrust or hatred toward each other.
- Scientific backing to severely limit people’s right to free speech
- Silencing (removing or banning) expressed opinions that may be more helpful for public policy than the current policies
- Preventing people from gathering to collectively discuss alternative or at least additional actions that can better optimize the situation
- Preventing people from participating in an election year by canceling primary elections and eliminating the effective in-person campaigning opportunities
- Scientific approval for metrics used in daily reporting of progress of epidemics
- The approved method involves a very simplistic and sophomoric approach of simple counting of new cases and new deaths, instead of comparing both to baselines.
- New cases do not matter if they do not require hospitalization and the more appropriate daily metric is the utilization of available medical resources, not raw numbers or rates per 100,000 people.
- New deaths need to be measured as portion of all deaths and compared with similar numbers for this time of year as well as the projected annual rates to better account for deaths that would have occurred in the same year even without the infection.
- The approved method includes in COVID death counts all deaths with positive tests, or presumed infected based on symptoms, or even just assumed as a default cause of death.
- This has compounded the sense of fear and panic in the public, making them less rational in considering to return to normal social lives even with risk has proven to be not unusual after all, and made more people prone to getting the disease or complications due to their heightened anxiety reducing their resilience to the disease
- This has raised among the population suspicions of widespread fraud that may lead to future social unrest or rebellion
- Scientific approval of testing methods make them appear more reliable than they are
- False positives will inflict on persons increased restrictions and follow-on testing to get free of this stigma
- False negatives will affect measurements about how the disease spread, making the false positives to appear more potent at spreading than they really are.
- Non-testing of asymptomatic or non-reporting populations further exaggerates the fatality rates, exaggerating the risk of the presence of the disease
- The approved method involves a very simplistic and sophomoric approach of simple counting of new cases and new deaths, instead of comparing both to baselines.
- Scientific attribution of conditions to the disease based only on coincidence
- Very unusual if not unprecedented symptoms of patients testing positive to the disease are attributed directly to the virus based only on the coincidence
- Conditions may be caused by other conditions that the virus makes worse or that make the virus behave worse than it normally would.
- Scientific assurance of association with virus blocks any consideration of additional factors that might in fact prove to be more important than the virus itself, such as government-induced fear, anxiety, and depression in patients.
- Scientific assurance that an effective and safe vaccine will eventually be found
- Despite contrary evidence of non-effective and somewhat risky influenza vaccines that have to be administered annually and are only about 50% effective.
- Assurances of approval and distribution within 2 years gives false confidence to population that they can return to being completely free of social-distancing and non-essential restrictions by tolerating some level of restrictions for this period. There is no scientific basis to assure us that any vaccine is possible with mRNA viruses especially when they are so prone to mutation.
- Assurances for governments to rush approval and mandates for never before tested vaccination approaches of direct RNA injection could immediately induce on entire world’s population a new class of incurable autoimmune or immune-deficiency diseases that could be worse than the disease itself.
Compounding on all the above potential failings of the trust-science approaches is that fact that there is an inherent trust that science understands the truth about nature. This is another recurring complaint I discuss in this blog: that enlightenment-inspired governments inherently believe there are permanent truths the science can discover. This is illustrated by our using the word law for both science and government in the same sense, that once a law is declared, it is permanent until extraordinary evidence undermines it.
The biggest failing of science in the current COVID situation is its inability to react to new evidence that its original conclusions were wrongly decided, and the assurances to governments were incompetent. We implicitly accept that any initial science-based decisions attains some law-like status that is automatically presumed to be true until there is overwhelming evidence that it is wrong. In particular, such decision making does not permit a simple apology for making a mistake following new data that clearly disproves the original science.
Once the science becomes a law (in both senses), the default response to any objections is that those objections are anti-science. When the objectives are compelling enough to take seriously, instead of taking the sensible approach of retracting entirely the original conclusion, we instead preserve the original law with appendixes acknowledging the original errors without ever retracting the now clearly wrong conclusion.
The USA president frequently describes the current situation as being a war against an invisible enemy that threatens our lives and even our civilization. I suggest that the invisible enemy is the current faith in the infallibility of past-tense science, a faith the shuts down the original present-sense of applying scientific methods to currently available data even if (and especially if) that data conflicts with the established theories.
I describe data derived solely from theoretical models as dark data. Originally, I used the word “dark” in analogy to the dubious assumptions of cosmological dark matter and dark energy invented to make the sense of observations. Perhaps there is also a darkness in the evil-sense to our obligatory faith in past tense science.
This obedience may lead to a very different kind of existential threat to humanity or at least modern society. That threat is a self-inflicted harm resulting from blindly following past-tense science in spite of clear and obvious observations that contradict that science.
For COVID19 situation, all of the original policy decisions were based on flawed science using flawed data. Our trust in our laws forces to continue those bad decisions.
A dedomenocracy is an alternative form of government where policies are machine-determined based on popular expression of urgency for a new policy and based on all data including the most recent. Because the machine-intelligence is free of human concepts of science and of rationality, there is no conceptual contradiction if it decides that the best current policy is to repeal the previous one. Doing so does not require any apology because it is accepted that each decision is temporary and based on the best available data at the time. Besides, the public would be relieved by the sensible change in policy more consistent with their own experiences both in terms of observations and in terms of their growing comfort that they can live with the situation without government restrictions.