In recent posts, I have been writing some thoughts on how a government might work when it is based on data instead of human deliberation. I initially started this thought experiment as an attempt of building on an earlier complaint against using data to replace human decision makers.
In this effort, I deliberately exaggerated the practice to the absurd level of applying it to a vision of government by data. Such a government accepts validity of data-derived rules that do not require causal explanations. This concept of government replaces long lasting laws with short-duration rules and has no need for extracting justice for historic misbehavior.
I intended to come up with a satire to counter the recent advocacy of replacing more decision making in business with automated analytics based on data. I took a futuristic view that assumes we have access to sufficient data to support government-scale decision making. If such a system were possible, then I imagined it would lead to absurd consequences.
Instead, as I thought about it some more, the concepts of automating large scale government with data analytics begin to seen reasonable to me. I can see merit in eliminating the large body long-lasting laws and replace them with a smaller set of short-lived rules. I can see merit in accepting rules that lack human cognitive explanation as long that the consequences tend to be more beneficial than harmful. I can even see a benefit to replacing our overburdened prison system of long-lasting sentences with very short sentences or of brief but harsh punishments so that the punishments coincide with the short time the rule is valid. The extreme consequences do not seem so bad in comparison with what they replace.
The premise of the above is that the technology exists to support such government. We do not yet have the sufficient data collection and fast analytics needed for this kind of government (at least as I imagine it). However, we may not have to wait long before that data will be available. There are two trends, one is the growth in the number of ways we can collect data, and the other is the social acceptance of sharing data with strangers (or making that data public). Within a generation, we may have sufficient data to operate a government that automates rule making from data with not need for human deliberation and legislation. It may take a little longer for the required changes in societal attitudes to accept the consequences of no human representation in rule-making, no human accountability, tolerance for rule-breaking without obtaining justice, or accepting brief but harsh punishments instead of lengthy prison sentences. But even that change may occur faster than I imagine.
Part of the reasoning of the government by data is to accept the supremacy of recent data. In order to have government focused entirely on the future opportunities and hazards, there is a need to devalue the precedence of past rules or even of causal explanation. The path forward in government may jump from one priority to the next based on the most recent data that either defines the nearest term strategies or defines the population’s current highest priorities. This emphasis on recent data along with the acceptance of the authority of data may offer something that we lack in current government: broad-based (or bipartisan) trust in the process.
Government by data can offer a better approach for establishing trust in government. The resulting trust may be shared among a larger portion of the population than we current enjoy.
In current government, the only ones who trust the government are generally a small subset of those who belong to the party currently in power. At any particular time, far less than half of the population trusts the government to make competent decisions.
With government by data, the nature of trust changes dramatically. First of all, there are no humans involved in creating the rules, in prioritizing which rules become active, or in deciding the punishments for those who disobey. There is no party affiliation of the machines making the rules. Secondly, the population will have access to evaluate the data and the algorithms used in the rule making. The population can seek out better data sources or fix problems in the algorithms. While I do not doubt that there may be contentious debate about what constitutes better data or algorithms, I suggest that we can agree to accept the rules from a set of data and algorithms that we agree are authoritative. A broad acceptance of trust in the automated rule making is a consequence of accepting the the authority of the data and algorithms used for those rules. The popular debate is not directed at the rules, but instead at the data and algorithms that make the rules.
A government by data may lead to a government that enjoys a broader level of trust than a democratic, or a democratic republic government. Personally, I would welcome any government that enjoys widespread trust within the population.
Some say that polite conversation requires avoiding discussion of religion and politics. Religious topics come up in politics of morality and justice. As a result, government is not currently a safe topic for polite conversation because it involves politics with religious undertones. In contrast, a government by data lacks a goal of morality and justice and it does not even involve politically appointed or elected decision makers. Government by data may be something that we can engage in polite discussion because it involves neither politics nor religion.
As a result, we may find that we talk about government more thoroughly and comfortably than we do now. Coincidentally, this benefits the government by data because the conversations themselves will generate new data to use in future rule making, especially in terms of deciding priorities for what rules are needed or the severity of punishment needed for rule breakers. With a government by data, we can benefit from more honest and open expressions of opinions about government without the concern that we are discussing either religion or politics. Government by data makes conversation more open because machines, not people, make the rules based on hard data instead of opinions or dictates from political leaders.
The idea of the entire government being run by data is a futuristic Utopian fantasy. The technology for this kind of government does not yet exist although it may exist in the near future. Although I presented the idea as an absurd extreme case of over-reliance on data for decision making at the expense of human accountability, I think that it may have some merit when applied to the issues facing government in general. We may find it more politically productive and socially cohesive to have an authoritative government but one where the authority derives entirely from data and algorithms that anyone can consult at their leisure.
Although we lack the technology to adopt a comprehensive government by data, it may be worthwhile to start thinking about government by data as a political movement. I propose a new political party named Dedomenocratic Party. Unlike the other political parties that base their platform on specific goals, the Dedomenocratic Party’s platform is based on process of selecting short-lived rules based solely on most current data and algorithms. The other parties assert some claim about deep truth of human nature or proper government. In contrast, the Dedomenocratic Party claims such truths or understanding is irrelevant to the project of government when we have sufficient data and algorithms to make decisions automatically.
The Dedomenocratic party respects only the need to accommodate the opportunities and hazards emerging out of the future as exposed by data and algorithms. The origin of time is the near future, not the distant past. This concept of time suggests a form of government that relies entirely on data and algorithms at the expense of historical precedence and demands for causal justifications. This government can work because the rules (and penalties for rule-breaking) are necessarily short-lived so that the rules need to be updated frequently with the even more current data or improved algorithms. With the use of recent data and trusted algorithms, the benefits can outweigh the harms especially when there is a wide portfolio of simultaneous rules. Short-lived rules based on most recent data offer less risk for wrong decisions than long-lasting laws based on decades of debate on even older observations or anecdotes.
A new Dedomenocratic party can be useful today even without the extensive technology for the entire government to be run by data. Several present day issues can benefit from a Dedomenocratic approach.
One example is the current debate over how to respond to evidence of anthropomorphic global warming (AGW) or climate change. Currently much of the debate centers on whether AGW is real or whether any policy can change it. I suspect the debate will eventually lead to some long lasting policy change to place more burdens on the burning of fossil fuels and those will produce unwelcome economic or even environmental consequences that we will have to deal with later. The current debate is one that insists on a decision of which ideological side possesses the real truth and offers the wisest policy for the long term. A Dedomenocratic party may change the nature of the debate by rejecting the relevance of ultimate truth or wisdom. A Dedomenocratic solution would be to offer a short term rule based on current data about climate change and carbon-dioxide concentrations. Such a rule may be to impose new restrictions of burning carbon fuels or high taxes to discourage its use. While this appears to be consistent with the policy promoted by the Democratic party, there is a difference in that the Dedomenocratic party is making no assertion that the policy is based on fundamental truth and wisdom. The Dedomenocratic part only offers short term policy that is based on the currently available data. The Dedomenocratic policy expects that fresher data or revised algorithms will make the rule obsolete in a few years. The short-term rules anticipate the need for creating new rules updated to accommodate the more recent data and algorithms, or for allowing the rule to retire without any replacement.
The Dedomenocratic party approach to current issues consists of two principles based on data. The first principle is that policy choice must be automated based on algorithms and data that every can inspect. The second principle is that the selected policies must be few in number and limited to the most urgent issues. The second principle is what may justify not renewing an earlier policy. Perhaps the data suggests no changes in the AGW conclusion when the rule expires, but the population’s priority has shifted to other issues that demand more attention. The Dedomenocratic party thus may adopt polar opposite objectives from one session to the next. For example, it may follow a period of a AGW rule with a period of a rule that emphasizes improving economic growth.
Another example may be to address the high deficits and unbalanced budget. The excessive growth in debt has been a problem for nearly a decade without any significant attempt to address the problem through spending cuts or tax increases. Again, I think the debate is hindered by the insistence of deciding which of the current two parties has ideas that are closer to fundamental truth and wisdom to justify a long lasting policy commitment. A Dedomenocratic party may instead answer the problem with short term rules that explicitly address the deficit without any claims about fundamental truth or wisdom. The rules offered by the Dedomenocratic party would be based on the current data that analytics suggest is a better balance such as revising the computation of near term benefits and raising taxes. Both suggestions are clearly controversial, but the Dedomenocratic party offers a new argument that the change is explicitly valid only for a short term (for example for the next two year) and has justification based on analysis of recent tax revenues and benefit disbursements. The rule automatically expires because we expect to obtain new data that invalidates the earlier rule. After two years, the Dedomenocratic party would advocate new rules with different tax rates and benefit disbursement calculations based on the more recent data collected and the improvements in algorithms. The process continues indefinitely without making any claim of fundamental truth or wisdom of the rules.
To improve the odds of benefiting from the budgetary rules, the Dedomenocratic party would advocate for a broad portfolio of changes with widespread tax increases and adjustments across a broad spectrum of benefits. Like in financial planning, a broad portfolio of rules will have some rules providing benefits that can counter the inevitable disappointments of other rules. Certainly, there will be complaints. In contrast to the current political environment that becomes paralyzed by complaints, the Dedomenocratic welcomes complaints because these become new data to include in task of coming up with a new portfolio of rules (taxes and benefits calculations) to replace the portfolio that will soon expire. During each iteration, we can expect that that benefits will outweigh the harms for the aggregate. Frequent iterations of rule making using most recent data will tend to average out the costs and benefits on individual groups so that in the long run the benefits will outweigh the costs.
In the budget policy debate, the Dedomenocratic party may insist on data to verify that it is a priority. For example, some economists argue that the current levels of deficit spending are sustainable and do not present a significant financial hazard. Much of the reason we have allowed the budget debate to unresolved is because there is a sense that the debt is not a problem. This is a debate among economists where different economists support one or the other of the current parties. This debate about the urgency of solving the budget problem is itself a debate about fundamental truth and wisdom. A Dedomenocratic party approach can dismiss the need for resolving who is right but instead decide the currently acceptable level of deficit spending based on the most recent data and the least ideologically biased statistical analytic algorithms.
Something approximating this kind of economic management occurs with Federal Reserve with its management of interest rates. Similarly, the Dedomenocratic would offer strictly short term rules that will quickly become obsolete when we obtain new data. The Dedomenocratic party offers a short term rule that explicitly avoids human debate as much as possible. The offered rule is automatically calculated by statistical analysis of actual data measurements about the budget revenues and expenses.
The rules will only be in effect for a short time. When the rule expires, the Dedomenocratic party will advocate for new rules based on the most recent data. Here, the Dedomenocratic party is different from the other parties because the Dedomenocratic makes no commitment to consistency or respect for precedence. An expiring rule involving a high tax increase may be followed immediately by a new rule with a tax cut, or a rule involving a benefit cut may be followed by a rule with a benefit increase. The Dedomenocratic party would make no commitment to consistency. It does not need to be consistent because it makes no assertion about its access to fundamental unchanging truth or wisdom. Its offered rules are based entirely on the most recent data and analytics.
The Dedomenocratic party has no ideological commitment. Instead the Dedomenocratic party commits only the authority of data and of widely accepted analytic algorithms that involve statistical methods (such as machine learning) instead of built-in ideological biases that are typically found in simulations.
Another example of near term contributions of a new Dedomenocratic party concerns the our approach to criminal punishments. The Dedomenocratic party can re-open the debate about the humanitarian merits of long-term incarceration that isolates prisoners from contributing to society and leaves them very disadvantaged for reintegrating with society after their sentence is complete. Shorter term but harsher punishments would be more beneficial to everyone by allowing the punished person to return quickly to his community and thus experiencing the least disruption of relationships and careers. The Dedomenocratic party contribution to this debate is to point out that government’s only obligation is to manage widespread cooperation with future rules. In particular, the extraction of justice for past deeds is not a necessary function of government.
A Dedomenocratic party can advocate for a data-driven government that derives its authority entirely on the most recent data and most currently accepted unbiased statistical analytics. Such a government can operate without human cognitive justification (and consensus through debate) and without continuity with the past through respecting precedence or some lasting notions of justice. Free from the winner and loser dynamic of elections deciding who makes rules, a government by data can make progress on urgent issues and enjoy a broader consensus and trust based on the knowledge that the rules come from data instead of people.