Early on in this COVID19 situation, I wrote about a particular approach that differed from the approach we actually took. I was writing from the perspective of using this situation for a thought experiment about how being guided by bright data is different from being guided by dark data. In my blog, I describe bright data as direct observations from reliable sensors, while dark data is computed information using models from previously tested science. I use the word dark to reflect a sinister dimension to this practice. Dark data fills in the gaps in what is possible to observe with our beliefs about what is going on.
There are many concepts of how computer-driven government would operate. My fantasy government of data and urgency has some specific constraints.
- The default condition is that there are no enforced rules. I mean this in the extreme, people can get away with anything by default.
- The government (of algorithms working on data) will impose rules with absolute authority beyond any person’s or group of people’s ability to over ride. Again, this is meant in the extreme sense of machine tyranny over the people.
- The government’s rules will only last temporarily with automatic expiration dates ideally as soon as possible to return to the default condition.
- The government will only impose rules when the population will demand, a democratic majority expression of urgency.
- Despite the tyranny of machine’s authority, there is abundant roles for the people in a sense that replaces democracy: People select and verify data, People select algorithms and the measures to optimize, and People determine the urgency trigger to introduce a new rule.
- The algorithms place priority on bright data of most recent sensor observations and these are assumed to be both diverse and exhaustive. The algorithms uses dark data (from prior scientific discoveries) sparingly and only when the dark data does not contradict the bright data. For my fantasy government, I mean this in the extreme sense of making decisions based on correlation instead of causation.
The point of this fantasy government is to be a thought experiment to imagine how would things work differently compared to what we are actually doing. There is a lot of variations in government approaches, but it appears unanimous that governments are following rational decision making relying heavily on previous scientific discoveries. I recognize there are controversies about whether one government’s choice of science is more accurate than another, but in general decisions do appear to come with some reference to a experimentally tested hypothesis.
My fantasy government is irrational. When it does make some rule, it will act on correlations, even spurious ones that contradict already established causalities.
In the extreme example, it may observe the fire-fighters are frequently present when a house is on fire and conclude that fire-fighters started the fire. While that is possible, I believe it is unlikely when given sufficient quantity and variety of data. The algorithms will have access to abundant and diverse information about what led up to the ignition of the fire.
The causation versus correlation argument usually involves a very small number of factors (often just two) that are approachable through scientific testing. The algorithms have access to observations to a vast number of factors related to a vast number of interconnected processes. Correlations at this scale resist a proof of causality, or at least the urgency of the matter leaves no time to rationally establish the causality.
Urgency is a key part of my fantasy government. By default, the government is inert to the population, but it continues to collect data of the people, by the people, and for the people. When sufficiently abundant and diverse, these observations will capture the knowledge about the world either directly from sensors or indirectly by observing what people are doing and distinguishing what is working or not.
Correlations of this scale of data can produce helpful guidance even if lacks a rational explanation that can convince any human. We see this happen with machine learning that work well enough to give the machine autonomy to replace what we once trusted only to other humans. We are content with the demonstration of being right more often than wrong, and the benefits of the right results outweigh the harms of the wrong results. When that happens, we do not demand a rational explanation.
In the context of the current COVID19 situation, I imagined my fantasy government would prefer observations over scientific models. In specific terms, from the start the observations were that this disease will spread, people will need medical attention, and many will die. The fantasy government could immediately act to find the best policy with the assumption that these consequences are inevitable.
- It could do this because it diminishes the relevance of past science that gave hope to our actual governments.
- It would do this when it became clear that the science was not living up to its promises in preventing the spread, in treating the illnesses, or even in predicting the needed medical resources.
The issue in front of the fantasy government is what can we do to have the best future for the survivors of this calamity. Here, I’m presuming that the prior democratic selection of algorithms and measures of merit would place the most emphasis on the future, the next generation. In the absence of an immediate threat, I like to believe we would prefer to plan for the future.
In the current crisis, it is self-evident that there will be losses and there will be survivors. When this is over, even if there is a hideous number of casualties, there will be survivors. A planning algorithm would consider the perspective of those survivors: what opportunities are available to them, and what burdens and resentments are placed upon them.
Recall that this fantasy government is under no obligation to prove rationality. It could observe the current measures of rates of transmission, complications, and death without needing to project a numeric total count for these. Those rates may very well imply that hundreds of millions of lives may be lost. Urgent quickly-expiring decision making needs only the information of the inevitability of casualties and survivors.
The fantasy government would optimize the conditions available to the survivors. Given the complex nature of our economy, it is important to maximize the number of survivors, but primarily for economically productive survivors. The fantasy government does include data about human condition to weigh in the economic benefits of moral support from elderly retirees, so it is not automatically dismissive of those populations. It will, however, place emphasis on best preparing the survivors for the future when this crisis is overcome.
The fantasy government would not obsess over raw or per capita numbers. The important observation is that there will be some number, and that number is likely to be significant.
The fantasy government only enacts temporary, quickly expiring, rules. The rules would expire long before the numbers are fully counted. If urgency remains after the first rule expires, there will be a new assessment with updated metrics. Things could look worse, or they could look better. The new policies would reflect that new assessment accordingly.
In a much earlier post, I suggested the need to mobilize and to incentivize the less vulnerable to be as fully productive as possible in order to allow the more vulnerable to be protected and isolated. The economic activity of the less vulnerable could facilitate the spread of the disease and thus the risk to the vulnerable. In contrast, not allowing this economy would impede our ability to care for the vulnerable. I believe we are seeing exactly this right now with the results that our actions are themselves causing casualties.
The fantasy government has the ability to make decisions that would be impossible under our current governments. The fantasy government readily accepts the inevitability that bad things will happen, while our democracies act on the belief that our science can prevent those bad things from happening. In fact, bad outcomes are inevitable. Even small numbers of these outcomes can have dramatic effects: we will lose people who are important in various ways.
The current government’s obsession over the absolute numbers ignores the fact that when this is over, these casualties will only be numbers. In the future, our government will have to contend with the remaining survivors. At that time, the survivors will judge the government’s wisdom in prioritizing dark data over the observations available to it.