In the first post in this series I argued that:
This led me to articulate a basic mechanical reason to accelerate international AI governance research now: governance bodies at rest tend to stay at rest.
In this post, I’d like to examine why this is the case. In brief, my answer is that: governance moves at the speed of law, which, according to present-day legal theory, can only strongly interact with certain “sources” of legal authority.
I argue from the point of view of historical and legal scholarship, not from some vaguely cynical position of realist geopolitics. All too often, I believe, we let ourselves be persuaded by pundits that international entities are ineffective because they simply formalize existing unequal distributions of power. Of course, this is often at least partly true–international governance is obviously political. But there are more capable explanations of this phenomenon outside of raw cynicism. And by understanding these explanations, we can also see why many international governance instruments have been increasingly successful. I think we need to understand these mechanics in order to identify where, precisely, funding- and thought- effort can be most profitably applied. As I hope is clear by now: I think that it can be applied most effectively by accelerating international AI governance research.
Governance in well-ordered communities (such as those that have developed AI systems) is typically thought to rely upon the rule of law. More formally, well-ordered communities have developed a “habit of obeying” laws that in turn allows them to experience substantial freedoms and relative social harmony. Theoretical appreciations of these points also apply to the international community of sovereign states. But where does law come from?
Usually when we call something “law” we think of “statute”, or the laws made by certain bodies of our government after following some political process (debate, voting, deliberation, et.c.). In some states (e.g. parliamentary democracies), these processes are well defined and legislation (statute-making) is the preserve of a specific part of the government. In any case, it seems clear that, most of the time, we think of governments as the origin of laws.
But international society has no legislative body. The U.N. does not make laws. Neither does any other global governance entity. So, where does international law come from?
This might seem like an exceedingly academic question. But, as we’ll see, the answer to this question is crucial to an understanding of why international governance is a high-viscosity fluid, and thus what can be done to reduce that viscosity. We might continue by discussing six discernable modes of international law. In discussing each, we can also start to see potential answers to our question about the origin of law.
(You may of course already feel as if you have a solid grounding of the development of important international legal doctrines, in which case I invite you to skim or skip to the next section: “Sources come from discrete authorities.”)
For hundreds of years, European thinkers argued that all law came from the actions of sovereigns, who many characterized as the figurative embodiment of the state themselves. An act by the sovereign as the “head of state”, therefore, was attributable to the whole “body of state” they represented. And just as subjects were completely subject to sovereigns, sovereigns themselves were accountable only to divine or natural justice–hence the inherent “independence” of sovereigns (idiomatically, “they above whom no other reigns”). In the famous words of Louis XIV, “l'état? C’est moi!”
Hence the shock when, from time to time, people took matters into their own hands by exiling, deposing, or dispensing with their monarchs.
Early international law was little more than an attempt to solve this problem of unaccountability. Its sole purpose was to regulate relations between sovereign states by establishing norms by which sovereigns interacted with one another–norms that came to be called the “law of nations.” But how could any kind of system for regulating fundamentally unregulatable actors be achieved?
Around the late 15th century, early international lawyers suggested that relations between sovereigns were analogous to relations between equal subjects. Each was free to contract, due to their similar sovereign capacities. And each could actually perform the obligations entailed in pretty much any agreement, as each was sovereign in their jurisdiction. Thus the birth of the modern treaty, as something that closely resembled a private contract of the day.
Importing analogies from the private law to govern relations between states has been a dominant mode of international governance ever since. However, this understanding of treaty relations also has revealed several substantial shortcomings. For instance, just as in private law, contracts are not always honored. Disagreements between individuals can be adjudicated in civil court. But what did the law of nations say about disagreements between states?
Interestingly, disagreements between states were also largely handled by importing private law analogies. For instance, the principle of arbitration. In private law, any disagreements about contracts or contracting parties’ performance of obligations imposed by contracts can be resolved through arbitration. Sometimes that arbitration is carried out by courts, but sometimes that is carried out by independent mediators. In any case, the important factor is an external party.
The appeal of this solution was that it was already a familiar instrument to many European sovereigns. Many of Europe’s political leaders had previously thought of their relationships as mediated by divine authorities. Deference to the pope was still common among many Catholic sovereigns as late as the 16th century, a relic of a time when Canon Law served to unite much of Europe’s legal practices. Papal oversight of European relationships with non-Christian peoples (i.e. Indigenous Americans) was often (but not always) respected by Spanish authorities.
But when mediation failed, international lawyers deemed it accepted practice to “appeal to heaven” for final arbitration. That is–to go to war. Until the early 20th century, war was regarded by most legal theorists as a legitimate method of arbitration. The winner’s claim was judged (by god, or by natural law) to have been valid and the loser’s rejected, in this understanding.
It is perhaps not surprising to learn that European states found themselves at war nearly as often as they remained at peace during the centuries in which these new legal theories were being developed. Viewing war as a legitimate method of resolving disputes tended to yield more war, and war often created new disputes.
This was one of the primary reasons why sovereigns themselves were often replaced by representative assemblies, entities based on the principle that the collective people of a state were themselves the source of sovereign power. Many peoples came to believe that governments should be more directly accountable to peoples, especially once they began taxing economic activity to raise revenue to fight wars. Furthermore, as European power spread across the globe, European doctrines of sovereignty were challenged by colonial conditions.
Facing a different world than early modern legal theory had anticipated, a new group of lawyers and social scientists articulated a “scientific” basis for law, including international law. Trying to move as far away from a universal “natural law” as possible, which seemed, at best, an unenforceable ideal, and, at worst, an enabling mechanism for state violence, they instead claimed that the only sources of law were to be found in the economic, social, and cultural formations that already existed within a given populace.
In other words, “scientific” approaches to law were not efforts to create “the state of things that ought to be,” as natural law posited, but rather “to render clearly the state of the state of things” as they already were. Of course, in order for that approach to be actionable, much had to be known about various peoples subject to law and their interests. Hence the origin of many of the social sciences in general–ethnography, anthropology, population statistics, modern political science, criminology, modern history, and more.
And how was this applied to international law? Instead of relying on natural law, which prescribed universal values that all states should adhere to, “scientific” efforts to refine international legal instruments had to be based on clearly documentable examples of times when states actually achieved a unity of legal principles. In other words, they returned to the maxim of early Roman international law–that any international law as such had to be based on laws already “common to all mankind.” In this view, treaties were viewed as most effective when based on already existing legal, political, or economic realities.
While many diplomatic historians have rightly observed that political leaders, imbued with a kind of liberal international idealism, were responsible for forming the world’s first formal international governance entity–the League of Nations–its operations heavily relied on precisely this “scientific” approach to law and governance.
“Scientific” approaches to law are still very much dominant today. Many of these approaches laid the groundwork for concepts of development and the formation of other important entities like the International Court of Arbitration or instruments like the Hague Conventions and the Geneva Conventions. Each is widely considered to be some of the most effective instruments of international governance humanity has yet created (and more on them in later posts).
However, an argument against the exclusive reliance on “scientific” legal instruments to govern the world and the behavior of its states was made persuasively by the horrors of the Second World War. All major states had previously agreed to abandon war as a method of advancing national interests. Yet, Axis aggression and the perpetration of state-sponsored mass killing programs laid bare the limits of international governance instruments that relied solely on state compliance with humanity’s common values, especially in cases where they ran afoul of state policy.
Enter the doctrine of human rights. The short version is that, during the post-war International Military Tribunals at Nuremberg and Tokyo, allied prosecutors charged Axis political leaders (i.e. those exercising sovereign powers on behalf of their state) with three broad international crimes, all resulting from criminal negligence of the state’s sovereign duty to respect the “Inalienable rights of man” as expressed in natural law doctrines. That, in itself, was revolutionary. Let that sink in, given the historical inviolability of state sovereignty up to that point.
The postwar settlement also established the clear utility of governance promoting ideals, and specifically through international governance instruments and bodies. Funny enough, post-war international lawyers enlisted the pope again in their efforts to equip international law with concepts to govern the relationships established between states and their subjects (“the dignity of the individual” was terminology suggested by him, after all).
This principle is the main point of the Universal Declaration of Human Rights issued as one of the U.N.’s foundational documents. In other words, by virtue of being a human on earth, you were supposed to be able access international legal protections that may not exist in the statutes of your state. When people engage in “human rights work,” an important aspect of their endeavor, understood from a legal perspective, is doing what the states in which they are working ought to have done in the first place.
Present-day practices of international law have been clearly shaped by all of these preceding eras. “Scientific” approaches are still influential, though these have evolved into a distinct domain of “customary” international law. By that, lawyers mean laws or norms that have developed over time, according to what can be documented. Importantly, customary international laws need not necessarily be formalized as treaties or declared by municipal statutes. Rather, they mean there is an observable habit of obedience to a certain norm, or perception of an obligation, such that it has the effect of law–a law “common to all mankind.”
Customary international law is often entwined with human rights doctrines as well. Consider the norm that has developed of not returning asylum seekers to the country from which they are fleeing. There is no international treaty obligation to avoid doing so, and few states have statutes prohibiting this. Yet, it rises to the level of customary international law because of common adherence to key human rights doctrines that are widely accepted in international society.
As I hope was clear from the above discussion, what ought to be regarded as the sources of law are often subject to as much debate as laws and norms that act as laws themselves. Again, it may sound like an academic question. But, in fact, it is of crucial importance to developing an effective field of international AI governance research. That’s because not all sources of law are equally “authoritative” to those tasked with actually creating international legal instruments or regimes. And not all governance instruments need to be formal laws in order to have a comparable effect.
Legal “authority” is usually thought of as a dichotomy. “Primary” legal authorities are those that anyone considering a law or a judgment must take account of–previous judgments, existing laws, or some sovereign commands, for example. “Secondary” legal authorities are those that may be useful in that process, but are not essential to consider–law review articles, books by experts, amicus curiae briefs, and so on.
To a large extent, the same dichotomy exists in international law as well. Primary authorities, such as prior treaty obligations or a state’s own laws, may have important bearings on a new treaty. Secondary authorities can help parties arrive at further clarity as to questions that may arise, but parties have no obligation to consider them.
That is not to say that secondary authorities are not useful. Because of the heavy reliance on social scientific scholarship that today’s legal theories regard as so important to lawmaking, secondary authorities have often exerted significant influence on the creation of legal instruments. In fact, in these situations, secondary authority has historically played a larger role in clarifying international legal principles that can underpin governance instruments, either by law or by the formation of international governance entities. And it should be apparent that there is a role for both policy and general scholarship in this consideration.
A recent example might be international actions intended to fight climate change. As a highly technical problem, scholarship submitted as secondary authority by academic researchers and U.N. staff scientists played a significant role in shaping the factual basis of climate action treaties. Likewise, policy scholarship provided essential analysis of treaty provisions themselves, ensuring frameworks were feasible enough to succeed (provided sufficient municipal political will).
Enter the problem of AI governance—a highly technical problem and one completely novel to political authorities. Municipal and international governments alike are currently clamoring for this kind of secondary authority. But unlike a municipal government, which can, if needed, order the creation of this authority (e.g. subpoenaing expert witnesses), international entities must rely on voluntary or proprietary knowledge. As proprietary knowledge is not available to international governance entities (most expert knowledge about AI systems is still privately owned), there are precious few sources from which international AI governance efforts can draw. This is why the UN Secretary General’s office has opened a call for papers on important themes relating to AI governance, the best of which will undoubtedly be classified as secondary authorities to any efforts to achieve international understandings on AI governance regimes. Furthermore, it is through such authorities that the existence of norms regarding AI governance across the world’s states can start to be documented.
Will that call be answered? Will responses be high-quality enough to aid the development of international AI governance instruments, principles, or entities? What are the consequences for AI governance and our ability to control rapidly advancing AI systems if the answer to either of those questions is “no?”
I am aware that this is a long, complex post.
But allow me to make some final points that relate what has been discussed to my overall argument that we must accelerate international AI governance research now. (Note: I am working to transition into precisely this kind of role, so this argument may, unavoidably, be self-serving to some degree. I do not think that affects the overall argument, however, for the field needs to expand to accommodate many more than one additional researcher).
I argued previously that international governance takes a long time to develop, perhaps much longer than domestic governance. This post demonstrates several reasons why that has been true. It also argues that focusing on developing a breadth of deep expertise in the target field is a proven pathway to creating bodies of secondary authority that will underpin the formation of international governance instruments and entities.
If this argument is convincing, I think that should lead us to the conclusions that the time to accelerate this research is now if we want to be at all prepared for dramatic advances in AI capabilities in the intermediate future. Governance institutions, especially international governance institutions, need this support now. And it can come in several ways: increased support for individual researchers (funds can support additional positions at leading organizations); increased intersectoral mobility between academic, policy, and industry roles during a single grant period; increasing support for scholars at all levels (early career, mid career, senior).
Each of these suggestions will be explored in greater depth in later posts. However, they are all important considerations when discussing the issue of coordination in international governance. My next post will consider how accelerating AI governance research should be accelerated in order to substantially blunt what are likely to be the sharp edges of such a coordination problem as it pertains to creating initial momentum and sustained attention to the development of AI systems. For, at the end of the day, states and international entities must have sufficient quantities of experts in order to achieve compliance with whatever international governance instruments are developed, especially if those efforts must occur under duress.
(image: the 10th annual meeting of the League of Nations in Geneva, 1930, AP)