Toward a Science of Policy

Results from the project "Equitable Water Policy Decisions Using Big Data," funded by the Mellon Foundation.

Introduction

From a biological perspective, all human action is aimed at self-preservation. While there may be differences in the proximate goals of individuals, all their actions (whether of mice or men) can be explained by the pursuit of predictability in the conditions of their existence. The proximate goals appear to be different simply because different individuals calculate the consequences of their actions over different lengths of time, i.e., some calculate further into the future than others. In the context of scarce resources, where desires exceed means, this leads to a contest for them. While technology does extend the means significantly from generation to generation, our desires have also continued to grow as we seek ever more of predictability.

A question naturally arises: does adoption of some simple guidelines to action ensure better chances of survival in the future? Does liberalism, or socialism, or conservatism, or populism guarantee a better future? Do corporate policies, or design guidelines developed by professional associations for engineering guarantee better profits or engineering performance? What is the framework in which such questions can be answered definitively? Do there exist definitive simple strategies for individual or societal success, as claimed by motivational speakers, including politicians and academics?

For comparison, let us look at a game such as chess or go, extremely simple compared to the contest for resources in complex modern society. Potential successful strategies in these games are exponential in the length of their descriptions. A two player game with N moves and M kinds of moves can have strategies that require M to the power of N steps to detail so every possible move of the opponent is considered. These games are relatively simple as the moves of the other players and their situation are entirely known unlike human actions in society where each player has little knowledge of the conditions, knowledge, motivations and intentions of other players.

In society, the number of players and moves is obviously large, and in modern networked society, the number of moves available to everyone of them is increasing with the advance of technology. By contrast, in the pre-technological societies in which the ideas of the humanities and social sciences were developed, the number of moves available to individuals was small. Women were severely restricted in their movements, everyone was subject to religious, class and caste constraints, free discussion of ideas often resulted in injury, imprisonment, or death, besides limited travel and communication for most everyone.

The great danger we face as a society is that most of what is considered education is training in the humanities and social sciences, which are firmly anchored in pre-technological literature even if they have begun to use modern statistical methods. These ideas were developed in pre-technological non-networked societies with severe constraints. Hence we have lawyers, journalists, politicians and academics supplying simplistic solutions to complex modern problems inspired by pre-industrial guidelines with many still believing them because of ideological affinity and what they have been taught to believe in school and college.

Moreover, most schoolteachers worldwide have a background in the humanities and social sciences, and indeed education credentials and education research are grounded in philosophies long invalidated by discoveries in modern mathematics and physics. However, political and religious beliefs only undergo slight modification when invalidated by evidence unlike theories in the physical sciences. This is because of the hypothetical rather than empirical basis of the current practice of the humanities, the social sciences and consequently, all education.

As society has become more networked, we have increasing failure in the predictive power of the humanities and the social sciences. We have seen in recent times that economists and sophisticated mathematical finance have failed to predict the markets. Similarly, many of the modernized data driven theories of the humanities and social sciences are being disproved by various recent political campaigns. For example, the media using political science or the ‘science’ of policy, or the social sciences or even big data has completely failed to predict the rise of Donald Trump or even to analyze his candidacy [1]. Established business has also failed to destroy his candidacy even though it makes them nervous, in spite of spending millions of dollars. All of this calls into question whether we have any science of policy or even social science. By definition, a science has predictive power that gives us a means to control events, not just a capacity to explain events in hindsight.

Common Misconceptions of Pre-Technological Society

The following hypotheses, which are accepted without question by different groups of policymakers or the public based on their ideological affiliations are examples of such misconceptions.

1. The Delusion of Expertise. In pre-technological society, where universal cause-effect relationships such as Newton’s laws of motion were not understood, all human effort was empirical craftsmanship. Because of the constraints on human expression, there was little chance for capable individuals to excel, except perhaps as soldiers in turbulent times. However, the notion that expert advice arising from memorization of various books or crafts can help solve complex problems still persists. An obvious example is the large number of think tanks and expert advisers on all variety of topics in both corporations and government. Even the structure of modern academia populated with disciplinary experts or specialists reflects this pre-technological belief. In practice, expertise from multiple disciplines cannot be integrated by the experts whose knowledge is narrow and contextual, but only by true generalists with depth of knowledge in all those multiple disciplines.
2. Research must be specialized. This is a corollary to the delusion of expertise. All research by definition is interdisciplinary, as what is disciplinary is already documented in the textbooks, by definition. Second, the difficulty of decomposing an interdisciplinary problem into disciplinary problems, which is that of the decomposition of an optimization problem into lower dimensional or disciplinary problems is itself a very difficult problem. Hence, interdisciplinary problems, or research problems are likely to be solved by individuals or teams of individuals with depth in multiple disciplines, if at all. It took perhaps 40 years before IBM learned how to modularize computing, and produce the personal computer architecture, with many of the most accomplished engineers and scientists on its payroll. Hence it unrealistic to expect that a group of lawyers and politicians will solve much more profound social problems by sitting together for a while. It is also incorrect for government funding agencies to assume that having individuals from multiple disciplines in a proposal ensures integration of those disciplines in the research performed by the group.
3. Free trade is universally beneficial. Unilateral free trade for a given country is only beneficial when all countries have a commodity based currency, and similar access to information. This automatically forces exchanges based on the natural competitive advantages of different regions and groups to the benefit of all parties, as a certain quantity of the commodity has to be exchanged for goods or services. Second, if one side runs out of the commodity such as gold, the exchange cannot continue. Currently, currencies are based on debt and non-market barriers to trade and information exchange can significantly distort the international division of labor through various kinds of debt purchases.
4. All professions of religious belief have equal merit. All religions claim that practice of their principles helps individuals harness more fundamental natural laws in a manner that violates known physical constraints. Of course, if they do claim exceptions to natural laws, they contradict themselves, as our intuition of a consistent existence is the basis of all our actions, and certainly all our interactions and communication, besides being the basis for all scientific endeavor including unified physical theories. Similarly, if a religion is interpreted so as to produce logical contradictions, the interpretation must certainly be invalid, whether or not the religious belief is valid. Indeed, many wars have been fought because of such beliefs. The South fought the North in the United States in part because they believed they were on God’s side [2] and therefore invincible in spite of being completely dominated by the north’s industrial infrastructure; the Germans in the First World War held similar beliefs [4]. Many Native Americans fought the vastly superior US Army at Wounded Knee believing themselves to be invulnerable to bullets as did the Sudanese under the Mahdi against the British. Hence, when someone claims to interpret a religion, they should first prove that they have practiced the religion, for example, by walking on water or raising the dead, or moving mountains, or living without food. Otherwise, their interpretation does not have any greater standing than that of anyone else. Mere memorization of various books in schools does not give anyone standing, as an inanimate computer can do that far faster, and besides, with modern search engines, give answers from books far faster.
5. Socialism produces just societies. Whether it is China or the Soviet Union, we have seen that the central planning of economies produces a concentration of power and large scale corruption, with little scope for justice. Those pointing to the Scandinavian countries forget that they grew for a long time with relatively unregulated economies, little war and violence, before they instituted their generous welfare states. A major reason for the failure of central planning is that there are few individuals with both great intelligence and great integrity that can handle the great concentrations of power inherent to socialist systems. It is not generally possible for a group of individuals to solve extremely complex problems in unison as most optimization problems are not obviously distributable if at all. That is, the problem of delegation of responsibility can be almost as difficult as the problem itself as mentioned above in the problem of research.
6. Markets are inherently efficient in processing information. Most economic analysis does not yet recognize that the value of money today is a random variable, given that we have debt based currencies worldwide, and the value of that debt fluctuates significantly as asset bubbles form and get liquidated. Hence, those who have the earliest information on the creation of money or its liquidation have an informational advantage over those who don’t, i.e., an arbitrage opportunity which is not supposed to exist. The efficient markets hypothesis [3] may still be valid when the fluctuation of currency values (e.g., its standard deviation) is a factor of 10 less than the average productivity of the economy (GDP growth rate), but this is certainly not the case anywhere today.
7. Immigration is economically beneficial to any country. (Increases the predictability of the conditions of existence, i.e., maximizing the sum of the gains to both the exchequer and the public in a country.) This is only true if there is no welfare state—something true more than a hundred years ago in most countries. Otherwise it is only beneficial to have immigration of individuals who are more productive than the average citizen of the country (GDP/capita), so the country does not become less wealthy.
8. Social media is more trustworthy and less censored. The absence of coherent explanations, inability to predict actual events, and blatant censorship and political biases have resulted in an erosion of trust in media. The Balkanization of society into groups sharing similar views or having similar objectives, facilitated by the internet and social media is a result of this mistrust. Many do not trust the media, and their distrust is justifiable, given that media companies are not performing a public service, but trying to maximize profits through getting larger audiences or better ratings. Social media, with its natural Balkanization into groups sharing similar views is believed by many to be uncensored, and reflective of the views of the public, or at least a segment of it. However, we know that this is not true as social media companies also look to their bottom line, apply rules of censorship aimed at preserving and maximizing revenues [10], and even cut deals with authoritarian governments. This trust in social media arises from pre-technological society, where there was never the capacity or an effort to actively shape opinions and alter beliefs at a rapid rate. So we still give credibility to other individuals who give an appearance of sharing some of our basic beliefs.
9. There is a tradeoff between efficiency and equality [9]. When working within the constraints of a single technology, this is generally true. An example is that of giving everyone low quality water at low cost. The rich can get higher quality water at home through better purification, but the rest don’t. However, when new technology makes it possible to deliver clean water to everyone at sufficiently lower cost, everyone can be better off. Technological progress over the past several decades has greatly increased efficiency, forced governments into far greater transparency and is leveling various power structures. As individuals have greater access to energy and information, concentrations of power are more difficult to maintain. It is only with large political interventions in the economy that inequality has been preserved; trillions of dollars in bailouts for large businesses being an obvious example. Large low interest rate loans for established business, and the well-off is another obvious example. The notion that there is a trade-off is again pre-technological, when the pace of technological change was at best glacial.
10. Rules can be followed. This is again a belief that stems from a far less complex society, with very few rules and regulations. Indeed, if the rules are contradictory, or in sufficiently rapid flux, they cannot be followed. If the number of rules is sufficiently large, they will never be followed, and cannot be checked for consistency by most individuals, leading to comparison of behavior with ‘best practice,’ as defined by some of the players. This results in permanent fear and paralysis of decision making in those affected by such rules. What is possible is a deterministic government or legal system so individuals can plan their actions for the longer term to optimize the use of scarce resources.
11. We can have a deterministic or predictable marketplace. In the medieval era of few or no discoveries, insular local economies, and rigid societies, the relationship between the inputs and outputs of any business, or technology, was generally fixed. Thus, the profitability of various ventures was known a priori, and only relatively random phenomena such as weather determined growth or decay of the economy. However, in today's age of far more rapid discovery, of both resources and technologies, it is not even remotely possible to have a deterministic marketplace as the discovery of resources, whether natural or artificial, cannot be predicted.
12. Games and therefore economies are computable. This misconception, common to most economic analysis, is rooted in simpler times where human action was severely constrained. There is plenty of work showing in fact that policy in a complex economy is undecidable under fairly generic assumptions [6,7]. Indeed, world chess champion Max Euwe [8, 11] showed in 1929 that even chess games under the rules of the time could be infinite without repetition of any sequence of moves.
13. Creativity as simple permutations. We still consider such things as music, art, drama, writing and the fine arts as being creative while mathematics and engineering are not considered creative in popular perception. However, just as computers can memorize anything and quote anything better than the most learned scholar, they can also produce the best classical music, or painting, or writing [12, 13, 14]. The same holds for anything done with the memory--drama, novels and stories, and prominently, journalism today, where Facebook considers journalists as training modules for algorithms [15]. They can be easily created by permutations and combinations of existing works, and therefore do not contain anything new. All real creativity consists in using our basic intuition of the consistency of existence in ever more sophisticated ways, i.e., reason mathematically from commonly observable measurements, or as Newton put it, make deductions from phenomena.

Are there Any Solutions?

It is not in general possible to derive general solutions or strategies to complex games, as even the length of a strategy will be very long. The same holds even for design optimization problems of various kinds. However, it is possible to verify if a proposed solution indeed satisfies various constraints. There are many problems which are difficult to solve but whose solutions are easy to verify. These are known as NP hard problems in computer science [5]. Similarly, we have the notion of probabilistically checkable proofs [5] that find with high probability if a long solution is correct without having to evaluate the entire solution. Thus solutions developed in a narrower context could be scaled up to cover larger geo-political regions. Potential solutions are more easily developed in smaller problems as they are easier to solve.

A very successful approach to solutions: There is a fundamental trade-off between physical measurement and inference, and solutions to problems that use direct measurement or feedback solutions to various optimization problems are far less complex than open loop or prescriptive solutions, and also far more robust to uncertainties in real world conditions. This is because they use measurement to react to changing conditions. Most of the great inventions of the modern age incorporate these principles besides that of distributed optimization which reduces the complexity of experimentation and therefore the complexity of the associated design problem, making the solution accessible to a larger number of individuals. From the refinery and automobile, to the airplane and the modern electric grid, to the computer, the cell phone, and the internet, these systems have endured because of extreme robustness, indeed resilience to uncertainty in components and operating conditions. However, the questions of finding the correct measurements, and correct distribution of decision making powers, and matching capacity with responsibility are all difficult problems.

Can Big Data solve these problems? Data or Big Data cannot solve these fundamental problems as all data can be distorted unless the decision maker has the ability to fact check all propositions presented before him, and integrate it in a fashion both consistent and correct (consistent with physical laws). Second, the interests of the decision maker have to align with the group she is deciding for. Otherwise the decision maker has to be someone that puts truth above self-preservation which is highly unlikely as we seldom get angels to govern men. The key to equitable policy in any domain in any sphere is to ensure that decision makers are those with great intelligence and integrity and with interests aligned with the group they are making decisions for. The matching of demonstrable decision making ability with responsibility is the great problem of our time if we want to solve all of the fundamental challenges facing mankind such as physical security, water, food and energy availability, besides creating a just society.

Another thing we note is that human beings and indeed most living organisms have been solving computationally difficult problems for a long time—that of preserving their identity as a sequence of memories of events—this is very much a Boolean constraint satisfaction problem which generally requires an exponential number of steps of computation. Moreover, living organisms solve this problem in real time, at least approximately, so the potential exists that humans could be trained to solve extremely difficult problems. Hence we can hope for ever more rapid solutions to the significant challenges of our time and future times. We further know from complexity theory that MIP=NEXP, i.e., even problems solvable by a non-deterministic Turing machine in exponential time (NEXP) can be solved by two or more interactive and independent provers, something applied to modern cryptography. This combined with the fact that humans are NP solvers, gives us hope that honest interaction between human collaborators can solve problems of any level of difficulty. We sum up below  our recommendations in various areas of policy.Their execution still depends upon the much more difficult problem of matching decision making quality with responsibility mentioned above.

Recommendations

The recommendations are based on the principle of moving resources from consumption to investment so the resources available in the future are maximized while allowing for a standard of living better than expected by most in the here and now.

On the one hand this means we have to avoid overconsumption by avoiding the following:

1. The bailing out of too-big-to-fail businesses effectively preserving incompetent management;
2. Subsidizing the well-off thereby removing the connection between wealth and actual service to society;
3. Creating a state protected privileged class in various ways such as outright discrimination, bureaucratic positions without accountability, and occupational licensing to price many professions out the reach of many. All these steps create a class of splurging superhaves who are completely disconnected from current economic reality.

On the other hand, we have to ensure that investment is effective and as long term as possible.

1. The longest term investment is in enhancing human problem solving ability itself, i.e., education—drawing out (educing) the potential inherent in all human beings.This must also include the inculcation of honesty, without which human collaborators cannot solve hard problems. This must also include the inculcation of personal responsibility for personal conditions, so individuals become problem solvers, as without a sense of responsibility, no problem ever gets formulated. 
2. The second most important investment is in creating buffers against various possible natural disasters such as floods, drought, earthquakes, volcanos, and epidemics.These minimize disruptions to the build up of capital, physical, social and intellectual, and enable human beings to operate free from fear.
3. The third is compensation for the consequences of various disasters so that individuals return to the productivity that permits them to invest more rather than engage in subsistence consumption.This is similar to investing in a developing economy where the potential for growth is far greater than investing in a developed economy.
4. Another important step is in the construction of checks and balances both nationally and internationally to avoid violent conflict.While we have many  laws with proven benefits on the books along with the institutions that execute them, their actual implementation is still inconsistent. Several International agreements are more difficult to interpret and implement as similar words are interpreted differently by different cultures. Hence, it is necessary to have both have a common language worldwide (English is the best candidate), and the inculcation of common human objectives--the predictability of the conditions of our existence--in education worldwide.
5. Finally, governments can subsidize the insurance of extremely high risks such as some health problems or give tax breaks for efforts to solve very difficult problems such as going into space or nuclear fusion so that capable individuals are encouraged to take bigger risks. Subsidizing the high risk pool in health insurance is similar to building infrastructure to guard against natural disasters--it minimizes disruptions to the accumulation of all kinds of capital in society. Tax breaks for working on difficult problems ensures on the one hand that science and research are not politicized, and ensures on the other hand that there is a variety of experiments to accelerate discovery.

The details of our recommendations on various policy areas such as water, food, energy, environment, education, healthcare, human rights, and government will be posted as separate policy briefs.

Conclusion: There are no Shortcuts to Salvation or Knowledge

The belief that knowledge can be obtained without effort from hypotheses is akin to medieval notions of random victimhood and random salvation. We cannot hope to obtain knowledge from speculative hypotheses, without physical measurements and logical deductions therefrom. The artificial quantification of human preferences or restaurant ratings do not constitute measurements as they are not physically reproducible in any sense of the word. We should not forget that we are also part of nature and biology even if we are the dominant player. Biology engineers for self-preservation or the reduction in unpredictability of conditions of existence. All engineering is necessarily constrained by physical laws. Physics in turn is a tiny subset of our communicable ideas or mathematics, which are consistent with physical measurements. All knowledge is mathematical, by definition (mathematics is that which is learned), as nothing self-contradictory can be communicated. All that can be communicated reliably is theorems and proofs, i.e., experiments with physical measurements such as mass, length and time and deductions therefrom. While philosophers have speculated logics that avoid the law of the excluded middle, they forget that all of our human communication assumes it. Hence, they contradict themselves in a fashion similar to all those who deny individual choice while using it.

Acknowledgements

The author Kartik B. Ariyur acknowledges discussions with scientists, technologists, futurists, journalists, venture capitalists, financiers, and academics ranging over a wide variety of disciplines. Special thanks to collaborators Valeria Chapman for teaching me the dynamics of checks and balances, Sorin Matei for showing me how to communicate scientifically, Michael Elgersma for pointing me to the fact that non-mathematical subjects are riddled with false premises, Emmanuel Nwadiogbu and Galen King showing the importance of individuals being left alone for any great human accomplishment, and Killu Sanborn for showing the nature of biomedical research and related fundamental questions. The author picked up the term “superhave” from Nicholas Haan of Singularity University who probably coined it.

This work was funded by the Mellon Foundation

References:

1. Why Donald Trump? A quest to figure out what’s happening in America. Clare Malone, retrieved from https://fivethirtyeight.com/features/why-donald-trump/ on 03/23/2016.
2. Religion in the Civil War: The Southern Perspective, Harry S. Stout, https://nationalhumanitiescenter.org/tserve/nineteen/nkeyinfo/cwsouth.htm, on 03/29/2016.
3. "Théorie de la spéculation," Louis Bachelier, Annales Scientifiques de l’école Normale Supérieure 3 (17), pp. 21–86, 1900.
4. The Kaiser and his Court: Wilhelm II and the Government of Germany, John C. G. R?hl (Translated by Terence Cole into English), Cambridge University Press, 1996.
5. Computational Complexity: A Modern Approach, Sanjeev Arora and Boaz Barak, Cambridge University Press, 2009.
6. Computable Foundations for Economics, Vela Velupillai, Routledge, 14 April 2010.
7. Effective Computability of Winning Strategies, Rabin, Michael O, Annals of Mathematics Studies, No. 39: Contributions to the Theory of Games, Vol. III, edited by M. Dresher, A. W. Tucker and P. Wolfe, pp. 147--157, Princeton University Press, Princeton, NJ, 1957.
8. Mengentheoretische Betrachtungen ?ber das Schachspiel, Communicated by Prof. R. Weizenbeck (May, 25, 1929), Proc. Koninklijke Nederlandse Akademie Van Wetenschappen (Amsterdam) 32 (5): pp. 633--642.
9. Equality and Efficiency: The Big Tradeoff, Arthur M. Okun, Brookings Institution Press; Revised edition, 30 April 2015.
10. Facebook guidelines: https://www.theguardian.com/technology/2016/may/12/facebook-trending-news-leaked-documents-editor-guidelines
11. https://en.wikipedia.org/wiki/Max_Euwe
12. https://www.fastcodesign.com/3058708/a-computer-paints-a-rembrandt-and-it-looks-just-like-the-real-thing
13. https://io9.gizmodo.com/5973551/this-classical-music-was-created-by-a-supercomputer-in-less-than-a-second
14. https://www.gizmag.com/creative-artificial-intelligence-computer-algorithmic-music/35764/
15. https://gizmodo.com/want-to-know-what-facebook-really-thinks-of-journalists-1773916117