With big data (and its three 3Vs of volume, velocity, and variety), we have the opportunity to create a completely different form of government, a government based on data. This form of government derives from a different perspective of time as having an origin in the immediate future (the source of new observations) instead of the distant past. In this perspective, the time interval of experiencing the world defines us and constrains us. In particular, we operate on observations that are stories told by intelligence operating on shorter time intervals (such as those experienced by molecules) and we are inherently ignorant of stories told by intelligence operating on long time intervals (such as those experienced by planets). Our reality of current experience is fully dependent on the future and disconnected from the past.
With government by big data, we have an opportunity to make quick decisions based on the current data without a need for human cognitive justification. The volume and variety of data allows us to make a larger number of decisions spanning many if not all aspects of government. The velocity of data allows us quickly or even automatically to replace old decisions with new decisions when new data arrives. This breadth and frequency of decision-making diversifies our portfolio of government decisions to tolerate occasional losses or failures because they will be offset by the greater benefits of our successes. We can permit ourselves to make decisions based on correlations alone without demanding justifications from some underlying truth. Simple or spurious correlations with no causal justification often will persist in time long enough to gain a benefit or to restrain the losses.
In an earlier post, I described government by data with the word dedodemocracy that implies continued public participation in government. This form of government has a authoritarian form where the public must accept and follow decisions based on data alone. However, the public still has a role by having access to the data and the analytics. In this form of government, the citizens will be expected to have data science skills in the sense of being able to construct queries, to perform analytics, and to interpret data for themselves. Although machine analytics will determine rules for present government that must be obeyed, the source of information for those analytics will be available to the public, allowing them to plan their lives around the current trends that will influence future decisions.
The change in perspective of time, placing the origin of time in the immediate future instead of the distant past, justifies a new approach to government that we can describe as just-in-time. The decisions are made just in time, not only for government but for how we plan our lives.
Just in time governing is very different from modern governments.
Modern governments place high importance on careful deliberation and continuity with past decisions. Modern governments move slow and are very reluctant to dismiss the decisions of ancestors. Modern governments honor a contract with the past. This connection with the past is consistent with the world view that the origin of time is in the distant past so that the past determines or constrains the present.
In contrast, just in time government by data recognizes that the origin of time is in the immediate future. There is no contract with the past, and there is no meaningful reason to abide by such a contract. Decisions by data based on high volume, velocity, and variety means that these decisions can and often will contradict past decisions. The emphasis of data-based decision making is choosing the most beneficial course of action in response to current data. There is no obligation to be consistent with the past decisions.
This approach to government appears disrespectful. Respect for ancestors, elders, or even our own prior decisions derives from the perspective of the origin of time in the distant past. We justify that respect based on notions of causal connection with the past. While we can respect how the past dealt with problems in their time, we are not disrespecting them when we confront modern problems. We need flexibility to find the best solutions for the current issues with current priorities and sensibilities.
Changing the perspective of time to place the origin of time in the immediate future respects the past by recognizing that we can not understand it. As I attempted to describe in my last post, the future is understood through a chain of story-telling that progresses up through ever longer time intervals. The longer time intervals (such as the one we live in) only have access to stories told by shorter time-intervals closer to the future origin of time. In comparison to our own or our ancestral past, we live or immediate lives in a shorter time interval. The ancestors are operating on a longer time interval. As result we have no access to any stories they are constructing.
In government by data, any connection with the past is completely irrelevant. This irrelevance is a consequence of the recognition that we are ignorant of what the past thinks about us. In my last post, I gave the example of the cancer patient trying to reason with the cancerous activities within cells. From the perspective of those faster-paced cancerous activities, the slower-paced experiences of the patient are irrelevant. The patient is finally complaining about what the cancerous activities did in the distant past. Currently, the cancerous activity is busy trying to deal with the current challenges at their time-scale and these current activities will affect the patient later. In this metaphor, the cancer patient living in a longer time interval is irrelevant to the current challenges of the activities within the cell. While we have technologies to eliminate the cancerous cells, we have no means to negotiate with the cancerous cells to convince them to reform their ways on their own. The cells live in the short time intervals that reside inaccessibly in patient’s future.
Similarly, our past now has experiences in longer time intervals and they are presumably constructing new stories about what they meant in context of what happened afterwards. It is possible the beings that reside in longer time intervals may have agency to affect our lives analogous to the cancer patient eliminating the cancerous cells. We may indeed be at the risk of some similar kinds of treatment by our ancestors. I argue that we have no way to know what will motivate them to take such action (assuming such actions were possible).
Our best strategy is to have a nimble government that responds quickly to new data that will inform us of some change in our reality. The best government is one that considers all of the available data to make the best decision for the moment. This government can not be constrained by respect for the past because we need to place priority on the most recent information. The origin of time is in the future, not the past.
This thinking is also consistent with my recurring complaints about substituting model-generated data for missing observations, a phenomena I label as “dark data” (my term is not consistent with data science practice). I can restate my earlier arguments against dark data as the problem of contaminating our data of fresh observations with prejudices from our past. I chose the term dark data as analogy to astronomy’s dark matter and dark energy: model-generated concepts to take the place of missing observations. I am very suspicious that such substitutions are misleading us from recognizing the reality that may be that the nature of empty space may be fundamentally different at galaxy and intergalactic scales when compared to human scales. It would be better to leave the observations as missing than to substitute our prior biases.
Government by data can be a nimble government. Decisions can occur quickly and frequently. We will replace old decisions with newer ones when new data arrives. The duration of validity for a decision will be very short: at a national level, the duration will be a few months or a couple years. Such nimble and fleeting decisions are contrary to our notion of laws that has a connotation of having a lasting influence. I prefer not using the word law to describe the data-analytic derived decisions. Although we are obligated to follow decisions during the duration the decision is valid, we expect this obligation to be short-lived because a new decision will soon replace it.
Eliminating the concept of law having a lasting consequence on our lives has implications about our concepts of accountability, justice, and punishments. For example, long-term incarceration as a penalty for violating a decision makes no sense when the duration of punishment exceeds the duration that the decision is valid. Even with modern government, it is difficult to defend the idea of someone still in prison for a crime committed decades earlier when current sensibilities would not have imposed a similar penalty. Prison sentences that outlive the relevance of a laws interpretation are unjust.
The frequent renewal of decisions in government by data means that the decisions do not represent some long-lasting law. Consequently, there is no meaning for a long-lasting punishment. We either need shorter duration punishments (perhaps prison terms that expire when new data causes an update to the governing decision) or find other means of addressing disobedience to the data-driven decisions.
Enabling government to make rapid and short-lived decisions based on the most current data may require dismissing our notions of justice or accountability such as those that derive from religious-inspired traditions.
I can see some justification for incarcerating someone who disobeys a current decision, but he should be released when the decision is updated. In my view, the only valid goal of the incarceration is to assure that the active population is obeying the decision so that we can obtain data about the consequences of the decisions when everyone obeys. The goal of incarcerating a disobedient member is not to punish the individual or to attempt to reform him. Instead the goal is to remove him from the data collection so we can observe what happens when the decision is faithfully followed.
This approach does not mean that we ignore the earlier transgression when the decision expires. On the contrary, we can deal with repeat offenders by being more proactive in demanding his obedience to newer decisions. The past evidence becomes part of the individual’s record as someone who is prone to disobey a decision. Alternatively, the record may identify demographic characteristics of such disobedience: the individual may be more prone to obey as he gets older but younger people in his condition may take his prior role.
Part of the data to inform new decisions is the observation of compliance. While we may temporarily imprison a large number of people for failing to follow a decision, the data on this population will provide demographic information about the population that found the decision unacceptable. A newer decision based on this data will hopefully account for these objections to have a larger rate of acceptance. After making this new decision, we will release everyone from previous sentencing and renew the cycle that may result in new arrests for those new groups who fail to obey the new decisions.
This makes sense to me. I do not like the idea that people are in prison for crimes committed decades ago where if they were tried today for the same crime they would not have received the same sentence, or may not have received a sentence at all. Government by data disrupts the notion of ethics. Given a new focus on the immediate future, we have less interest in past transgressions. This focus is less interested in concepts of justice or appropriate punishments. Instead, the past transgressions gives us data about the nature of the individual member of society, and this data can inform better future decisions that takes account of this valuable information.
The modern concept of a criminal also implies some permanent state of a person’s identity. Government by rules instead of laws would instead recognize rule breakers and this status would only have meaning while the rule is in force. We may add to the public record the fact that an individual had broken a rule. This data can be useful to build future rules that are less likely to be broken. Any penalty on a rule breaker, such as imprisonment, should not exceed the duration of the rule, but may be justified to last for the duration in order to allow the collection of data when the entire population cooperates.
A conversion to government by data can provide an opportunity to address the problem of the excessive accumulation of laws. In our current system, each law may have been passed to address some immediate urgency, but the law remains on the books long after the issue is urgent. Prosecution continues to use old laws try new criminals when those laws probably would not be passed today if they had not already been on the books. Modern prosecution is unjustly creative in reinterpreting old laws to new situations.
My impression is that we accept the convenience of having outdated laws as providing more opportunities to convict someone we want convicted for some transgression that may be hard to prove. With an abundance of laws, we can find something else that will stick and obtain the justice of punishment that the prosecution seeks. I find this convenience to be repulsive especially if the old laws that convict a criminal are no longer a public urgency. The active laws we should have are the ones that reflect current priorities and attract our urgent attention. Also, it is not valid to outlaw some behavior based on popular disapproval of that behavior. We should only govern behaviors that are having a significant measured impact on society with new rules that promise to address this impact. Once we get to the point where normal social pressures can adequately control the behavior, we no longer need rules to govern it.
An example that comes to mind is the 1906 Pure Food and Drugs Act that gave regulatory power to the Food and Drug Administration. At the time it passed, many companies marketed products for human consumption where the labels were misleading or the contents were hazardous. The public felt that the practice became so widespread that the regulatory laws were necessary. Today’s markets of human foods and drugs are far more mature with strong competitive advantages (such as through competitive advertising) to maintain honesty, quality, and safety of products. We are not currently experiencing the abuses we perceived over a century ago. The modern food and drug markets can self-govern without the need for federal oversight, but the regulatory agency remains. Instead of repealing this outdated law, the regulatory agency has creatively redefined its role to find new ways to regulate food and drug industries. While this regulation may be useful, I think it is very likely that the industry can regulate itself through market forces. There is no evidence that we need assistance from a federal level. The food and drug industries have matured to the point where they can police themselves. We could retire the FDA and only restore some counterpart when we observe a new urgency in the unlikely event of new collective malpractice in the industry.
The concept of governing by data is to have a small set of laws constrained by the current urgent issues and defined by current data. The 3V nature of data will support rapid turn-over of laws so that we can confidently place short-term expiration dates on the laws. I mentioned earlier, these rules should not even be called laws because a law has a connotation of having a durable validity. We do not need laws. We only need rules. The rules should apply only to current priorities and current data.
Government by data will provide us the means to escape from the tyranny of accumulated old laws. Data-driven rule making can provide a government that is relevant to the current population because all of the rules are recent, are based on recent information, and are issued according to current priorities.
For example, part of the reason for the recent protests about unpunished perceptions of police misconduct is that police were following accepted procedures that were originally introduced due to an urgency of violence against police from several decades ago. The current situation is different but the old practices remain. During the violent crime waves of the 1980s, we felt an urgency to empower the police to act more aggressively. Those same policies exist today even though the urgency of violent crime has subsided. While obviously we still experience some threat of violent crime and indeed violence directed toward police, the levels of such threats is much lower recently than it was in the 1980s. The concept of government by data is to concentrate only on the current urgent priorities. This means that we would not make or renew rules on something that occurs at a tolerable rate. We do not make rules for situations that the population does not perceive as an urgent requirement. In a data-driven government, we could have decided to substantially disarm the police patrols based on the historically lower rates of violent crimes. Given the fast nature of rule-making by data, we can rapidly reintroduce the rules when we feel an urgency to address the issue again. Even if that were to happen, we probably would not approach the problem the same way we approached it in the 1980s. The old laws and policies were out of date with modern sensibilities.
Our current government suffers from too many laws as described in a recent USA Today commentary:
[…] the accumulation of laws creates a drag on both prosperity and freedom. Jonathan Rauch calls the problem Demosclerosis, in his excellent book of the same name: Special interest laws build up kind of like arterial plaque, eventually choking off freedom. Economist Mancur Olson calls the same phenomenon “the web of special interests.” In his book The Rise and Decline of Nations, he suggests that this web will inevitably lead to economic and political stagnation […]
In addition, this is leading to an excessive prison population that causes harm to individuals and society as described in this Wall Street Journal opinion:
This is not only a serious humanitarian and social issue, but one with profound economic and fiscal consequences. In an increasingly competitive global economy, equipping Americans for the modern workforce is an economic imperative. Excessive incarceration harms productivity. People in prison are people who aren’t working. And without effective rehabilitation, many are ill-equipped to work after release.
Governing by data dispenses with the notions of causality, justice, accountability. In their places, we can have more comprehensible rules based on current data available to everyone. The recent advancement and maturity of big data technologies makes possible this new form of government. Taking the most advantage of the most recent information requires us to dismiss old decisions, including decisions involving punishment. There is good reason to expect that such nimble government can be much more beneficial to society. The frequency of making new decisions on newer information will make the consequences of bad decisions more tolerable. The result will be a society with fewer rules and where all of the rules have immediate relevancy that everyone can verify by checking the data.