Legitimacy and Non-Domination in Solar Radiation Management Research

Introduction

The environmental impacts of anthropogenic climate change, from an increase in global temperatures melting polar ice caps to the generation of extreme weather events, appear to be happening even more quickly than anticipated (Graester et al, 2018). Yet, many are worried that global action towards emissions mitigation has stalled or will be inadequate. As a consequence, there has been growing interest in geoengineering strategies that may slow or reduce the negative consequences of climate change despite their apparent greater risk than responses that focus on mitigation. Research budgets are growing, extensive experimentation in the form of computer modeling is commonplace, and outdoor experimentation is under way (Dykema, James, Anderson, & Weisenstein, 2014). In light of this new interest, many groups – some activist, some governmental, and some scientific – have proposed a variety of governance schemes for the research and possible deployment of risky geoengineering strategies. The purpose of this paper is to explore what I take to be a significant limitation of the proposed regimes: their failure to grapple with the significant power imbalances between the scientific and policy-making communities of the developed world doing the research and those of the developing world that will suffer the worst consequences of both climate change and climate engineering. This paper will argue that the value of non-domination, including and especially how it relates to global inequality, should guide our thinking about how to legitimately engage in research into solar radiation management.

It would be unfair to expect a contextually limited governance regime to resolve all issues of global inequality and domination. Yet, I argue that three common features of proposed geoengineering regimes exacerbate inequalities in vulnerability, power, and capability in problematic and avoidable ways. These features are a reliance on the indoor/outdoor distinction, scientific self-governance, and information-sharing. In each case, the regimes underestimate the effects of SRM research on the political relations between the developed and developing world. These regimes treat research and deployment as fundamentally distinct activities that ought to be regulated according to distinct normative logics. However, when we focus on the way that research affects power relations, we can see that this picture is false and that underlying political values that govern deployment should be applied to research.

Towards a Better Taxonomy of Geoengineering Strategies

In what follows, I argue that some geoengineering strategies offer specific governance challenges due to the greater power they offer potential deployers over the environment and, subsequently, over the choices of others. An implication of this idea is that various taxonomies of geoengineering strategies – those based on risk or on mechanism – will fail to capture all of the governance issues related to these strategies.

Climate (or ‘geo-’) engineering is typically defined (The Royal Society, 2009) as ‘intentional, large-scale intervention into the climate system.’ Despite broad agreement on the definition, geoengineering is a vague concept, composed of strategies and interventions that vary considerably in their normative and public policy import. Thus, there is no singularly appropriate governance regime for geoengineering deployment or research. Some (The Royal Society, 2009) have suggested that a good proxy of normative relevance is the distinction between carbon dioxide removal (CDR) and solar radiation management (SRM). The former refers to strategies that purport to scrub greenhouse gases from the atmosphere, including and especially carbon capture and storage, while the latter refers to interventions to increase the reflectivity, or ‘albedo,’ of the planet. If more of the sun’s energy is reflected back into space, then the planet will cool even if we assume a constant – or, within limits, even an increasing – amount of greenhouse gas in the atmosphere. The Royal Society argued that CDR broadly corresponded to the ‘safe’ interventions and SRM corresponded to ‘dangerous’ interventions. Yet, this distinction is misleading. First, even CDR strategies that are themselves safe will nonetheless require land-use decisions that, when compared to agricultural production, generate risks themselves.¹ What’s more, there are CDR strategies that intrinsically generate significant environmental risks. Some of these CDR strategies – such as ocean iron fertilization – have been the subject of controversial experimentation (Strong, Chisholm, Miller, & Cullen, 2009). Second, some SRM strategies – such as whitening roads or roofs – are perfectly harmless if not especially effective.

The CDR/SRM distinction is, at best, an imperfect proxy for determining which geoengineering strategies require significant governance in order to be researched or deployed (Gardiner, 2011, 344ff). A significantly better distinction would rely upon risk. We could separate geoengineering strategies into those that exceed a certain likelihood of generating negative environmental impacts and, through those impacts, undermining human interests. On this view, the most controversial yet potentially effective solar radiation management strategy – the injection of reflective aerosols particles (RAIs) into the atmosphere² – would need to be governed because, for example, it risks disruption of welfare-bearing precipitation systems.³ Yet, another intervention that relied upon a similar mechanism, such as making roofs more reflective, would not be a governance priority because it would not generate substantial environmental risks. It is important that these judgments do not assume that any particular intervention is unjustified. An intervention may be a governance priority in virtue of generating significant risks and yet be justified all things considered by the benefits its produces. So, this distinction would suggest that some interventions are more difficult to justify because they likely impose substantial probabilities of negative effects. On this view, governance plays an instrumental role on broadly consequentialist grounds since the less risky strategies need not be governed since things cannot go particularly wrong even if they are used unwisely. A consequence of this instrumentalist conception is that research governance is only relevant insofar as it allows us to manage the risks of deployment. Research into a risky strategy – such as ocean fertilization – might need to be governed in order to prevent ‘technological lock-in’ where the institutional incentives towards deployment will become more difficult to resist as greater resources are devoted to research and development (Cairns, 2014). Outside of the effects on deployment, the only relevant governance considerations for research are the potential environmental consequences of the experiments themselves. This will be relevant when we discuss the current governance assumptions about the ‘harmless’ nature of ‘indoor’ experimentation and modelling.

There are some important limitations of the risk conception. First, focusing on risk can lead to a problem of distribution. That is, if we aggregate risks across populations, then a large group of people suffering a minimal risk might outweigh a small group of people being subject to a catastrophic risk. This is especially true if we understand risks in economic terms, where the economic value of richer populations can create even more severe distributional inequalities. Second, risk evaluations of RAIs must implicitly make reference to a baseline. RAIs may not be especially risky when compared to a business as usual scenario of a greater than 3 degree increase in global average temperatures over the preindustrial baseline, but this is primarily because the consequences of such warming would be extremely severe. Yet, in comparison to a baseline where the developed world – those who are considering RAI deployment – satisfy their moral obligations through mitigation and adaptation, RAIs look very risky. Finally, and most importantly, the governance regime will be required to make decisions that generate winners and losers, distributing benefits and burdens. Yet, the risk-based regime is not inherently concerned with the procedural legitimacy or the relational equality of the agents and stakeholders involved in the regime. It seems clear that, at the least, the deployment of RAIs involves or constitutes an exercise of power, and there are normative considerations that ought to inform our evaluation of that exercise beyond whether it distributes risk in an optimal fashion.

As a result of these worries, I want to propose a different geoengineering distinction that emphasizes relational elements and human agency. There is a difference between capability-enhancing geoengineering strategies and passive, cost-reducing interventions. That is, some geoengineering strategies provide particular agents with more power, a greater ability to shape the environment and choices of others, and some do not. This feature of an intervention can be separated from its environmental risks. A geoengineering intervention can be risky yet do little to improve an agent’s ability to manipulate nature to suit their interests while another geoengineering intervention can be less risky in terms of its environmental consequences while grounding an enhanced ability to manipulate nature. Another way to put this is to distinguish between two sets of risks: environmental consequences of deployment and the possibility of human abuse. So, a geoengineering strategy that merely blocked or prevented some negative consequence of climate change, was widely available, subject to veto from stakeholders, did little to develop follow-on capacities to change the environment but nonetheless generated negative environmental impacts would represent a different governance challenge from a less immediately risky intervention that could be deployed by an elite few in ways that disproportionately served their interests and served as a further foundation for manipulations of the environment.⁴ So, let’s call interventions that exacerbate existing or generate new capabilities to exercise greater control over others ‘dominating’ while interventions that have sufficiently high likelihood of generating bad environmental consequences ‘risky.’ In ‘pure risk’ cases, the instrumental governance paradigm seems much more plausible: if those risks do not exacerbate current governance inequalities, then it seems correct to suggest that the governance objective is to ensure that the intervention is deployed wisely. Yet, the instrumental paradigm is inadequate when faced with dominating interventions that look relatively safe, or at least present themselves as clearly passing any cost-benefit analysis, but nonetheless exacerbate asymmetries in power and capacity between privileged and non-privileged agents.

This, I submit, is a reason that many are more skeptical or wary of geoengineering interventions than a strict understanding of the risks would entail.⁵ For example, the possibility of routine weather manipulation represented by cloud seeding or brightening generates concerns about who will be able to control these fundamental preconditions of our day-to-day lives. Yet, it is clear that reflective aerosol injections are both risky and dominating, sharing the features that demand robust governance. First, RAIs have the potential to cause negative climate effects, especially with regards to precipitation (Tilmes et al., 2013). Second, research and deployment of RAIs are currently justified – if they are justified – on maximizing, consequentialist grounds in ways that ignore other considerations (Barrett, 2008; Crutzen, 2006). RAIs will be bad, but climate change is worse, or so the argument goes (Gardiner, 2010). This generates strong pressure to adopt an instrumentalist conception of governance. Third, RAIs also have the greatest dominating potential of existing geoengineering strategies. RAIs can be deployed by a small group of people, require considerable engineering skill, and lead to improved capabilities to manipulate the environment. While RAIs, perhaps, can be used carefully in order to moderately shave off the worst aspects of climate change (Keith, 2013), many people could potentially benefit – while failing to pay the full costs – from more extensive RAI interventions. While it is true for the moment, that SRM interventions into the atmosphere are fairly crude such that it is not currently possible to carefully control temperature and precipitation, the very ideas of ‘peak-shaving’ and engaging in RAI safely demand that scientists develop more fine-grained control over RAI and its effects.⁶ Thus, scientists are aiming to generate RAI and other SRM techniques that will increasingly give us control over important components of the atmospheric system which could be potentially be used for well or ill. So, the governance imperatives for RAIs are strong due to both their riskiness and their potential for domination. Unfortunately, we shall see that most governance initiatives – especially those that govern geoengineering research – are much more interested in former, failing to take the latter sufficiently seriously.

It is worth pausing here to discuss the role of power and domination in the context of legitimacy.⁷ I argue that some geoengineering strategies look especially problematic, in part, because they exacerbate or lead to new opportunities for domination. I want to suggest, following the republican tradition, that domination represents an intrinsic relational wrong and a serious injustice. That is, if I dominate someone, our relationship is a source of political injustice (e.g. is wrongful) even if I benefit them or treat them well. Domination occurs when a person is subject to the superior arbitrary power of another. This definition involves two key components: superior power and arbitrariness. One agent has superior power over another when the superior is in a position to asymmetrically alter the choice situation – what choices are available to the agent and their relative costs – of the inferior position. In other words, superior power gives its possessor a degree of asymmetric control over its subjects. Superior power can be generated in a variety of ways: legal entitlements, physical strength, or technological capability. These different elements are all ways of providing control over the choices of another. It is important to see that one can dominate even if one never exercises the superior power; the mere fact that one’s choices are subject to the decisions of the superior is sufficient. However, superior power, by itself, does not always generate domination because superior power can be controlled and made accountable to those subject to it. If superior power is not contestable, accountable, or controlled, then superior power can be exercised unilaterally and without the permission of those in the inferior position. A legitimate state, on this view, does an adequate job of maintaining the freedom of everyone by instantiating a political order that either prevents people from having superior power or ensuring that all exercises of superior power are, in some sense, structured such that they are accountable to those who are less powerful. That is, preventing domination should be understood as a criteria of legitimacy for a political system and dominated people have a serious complaint against political systems that permit it.

With this in mind, we can understand the idea that some geoengineering strategies grant those who develop them greater direct control over the choice situation of others while some do not. This focus on ‘dominating’ geoengineering responses also explains the intuition that geoengineering is especially in need of justification when compared to other strategies such as mitigation or adaptation. The reason is that most mitigation strategies – such as renewable energy – do not imbue their users with additional influence over the environment such that they can exercise greater power over others. Of course, the possession of the intellectual property, resources, and manufacturing capacities for renewable energy production can give individuals social or economic power and thus indirectly lead to domination. But that is a feature of a social system that distributes benefits and entitlements in certain ways. Renewable energy, painting roads and roofs white, or carbon capture do not grant the possessor the ability to directly intervene into the choice situations of others, but some forms of geoengineering – including and especially RAIs – allow the possessors of that technology to asymmetrically structure the options that are available to other agents. Those technologies require more stringent governance.

Three Features of Geoengineering Research Governance

The problematic nature of geoengineering research governance is not a consequence of malice. Rather, since most research governance is by scientists and for scientists, the governance community not implausibly assumes that most of the people subject to its potential rules are acting in good faith. Those who will be subject to the power the research community generates are not especially salient and the relevant policy choices will likely be made by others. Yet, this lack of focus on power dynamics has led to three gaps in the research governance proposals. In what follows, I rely upon a general paradigm of research governance that brings together common features of the proposed regimes. Not every regime contains every feature but all contain at least one and most more than one.⁸ These features are: dramatically increased regulatory requirements for outdoor research as compared to indoor research, information-sharing as the primary mechanism of generating research equity, and dependence on scientific self-governance.

The Salience of the Indoor/Outdoor Distinction

One of the core research distinctions in the context of RAIs is between that which occurs ‘indoors’ and that which occurs ‘outdoors.’ More specifically, indoor research is a hodgepodge of experimental and scientific techniques that can occur entirely within a lab and generate no intrinsic environmental risks. These ‘experiments’ can be physically robust, such as testing the effect of various geoengineering interventions within a sealed chamber designed to mimic outdoor environmental conditions or occur entirely within a computer. It is telling that even the most stringent scientists who argue that there must be governance in place before experimentation engage in computer modelling and other indoor experimentation; indoor experimentation is viewed as harmless and thus need not be governed. Outdoor investigations, on the other hand, are deeply controversial and it is routine among the scientific community to suggest that there should be a moratorium on such experiments until a proper governance regime can be put in place. Popular condemnation has forced outdoor experiments in ocean iron fertilization to cease and planned experiments in RAIs to be cancelled. The justification for this distinction falls squarely within the risk paradigm: outdoor experiments risk negative environmental or health consequences and also create lock-in incentives towards deployment.⁹ So, drawing a sharp distinction between outdoor and indoor experiments is about managing risk.

Yet, it is hard to sustain this distinction. First, it is unclear why small outdoor experiments carry meaningful environmental risks, especially in comparison to other human activities that are considered routine. Of course, large-scale experiments that are de facto deployments carry environmental risks, but that is to recapitulate the distinction between research and deployment, not between indoor and outdoor research. Second, the logic behind lock-in is the idea that committing resources towards the development of a technology motivates people to deploy the technology so that the resources are not ‘wasted.’ Yet, indoor experiments also require resources – labs, work-hours, and the like – so it cannot be the case that outdoor experiments uniquely activate lock-in (McKinnon, 2018). It is perhaps true that moving to outdoor experiments makes lock-in reasons for governance stronger, but it seems that if indoor experimentation makes outdoor experimentation more likely, then lock-in considerations would also demand governance for indoor experimentation.

But the strongest reason to reject the idea that there is a sharp distinction between and indoor and out experimentation and to govern indoor experiments – including computer modelling – is that the indoor/outdoor distinction is not especially relevant to domination. To see this, consider the following scenario:

ISLAND: Imagine that two groups find themselves washed ashore on a desert island. After a period of time, they manage to build sustainable, if temporary societies, and eventually come into contact. Each group has equal amounts of technical skill and engineering resources. However, one group – the Morlocks – has discovered the plans for materials and know-how to construct a dam, which can with difficulty be built with their existing technology base, while the other group – the Eloi – has not and will be vulnerable to flooding if the dam is built. However, the Eloi could manage the flooding if they could build a similar dam or levee.

The groups are equal in terms of material resources, population, and technological skill. Yet, one group has discovered information that would allow them to create a technology that would lead a significant improvement in their capabilities to structure the choice situation of the other. Do the Morlocks have power over the Eloi

Does the asymmetric availability of information on how to create and use powerful technologies grant power to agents even if they have not yet created them? The answer seems to be, ‘yes.’ If so, this means that coming to possess additional information represents a potential threat of domination. Imagine a debate between the Morlocks and Eloi over the distribution of a valuable resource. The Morlocks could, of course, threaten to develop the dam unless the resources are distributed in ways favorable to them. The Eloi would need to take that threat very seriously and would likely need to make concessions in order to prevent it from happening. So, the mere discovery of information, even without any production or deployment, can result in a power imbalance if that information changes the ability of one agent to control what happens to another.

Let’s adapt the scenario somewhat. Imagine that the Morlocks do not merely discover the plans for the dam and the advanced materials to build it, but that they have some unique set of resources or skills that, through great effort, allow them to create both the plans and a workable design. That research program, aimed at simply creating more knowledge, changes the power dynamics between the two communities. Again, a good indication of the consequences, in terms of power dynamics, of this research program is that Eloi would likely be willing to take steps – through negotiation or otherwise – to prevent the research program from disrupting the equality between the communities. In other words, the research program is, from the Eloi perspective – a threat that must be contained and that reflects greater possibilities for domination. What’s more, the example nicely illustrates another feature of domination: the irrelevance of intention. It does not matter whether the scientists are developing the dam – or intend to actually build it from the discovered plans in the first scenario – in order to change the inter-group power dynamics. Perhaps the Morlocks are developing the dam for peaceful purposes – hydroelectric power or flood management – and plan to scrupulously avoid even the appearance of an explicit threat. This might make the change in power dynamics more palatable, but the peaceful program would still make the Morlocks more powerful because acquiring that knowledge allows them to develop the dam and then use it as they see fit. The Eloi are subject to the whims of the Morlocks, and one might think that that fact by itself changes the relationship between the two groups.

This is salient for RAI research. Like the dam, RAI can give one side an outsized ability to control the environment: a more sophisticated dam can determine water levels and a more sophisticated RAI program will give individuals the power to ratchet temperature up and down. RAI programs will research how to effectively deploy the reflective particles, build particles that are more durable and less costly, and model the consequences of various deployment suites. In other words, these programs – regardless of intention – allow those who control the relevant decision-making nexus to develop new and better capabilities to manipulate the climate in the way they prefer. Again, perhaps the people developing RAIs prefer to use them in ways that benefit mankind, but this is irrelevant to the question of domination. What is relevant is that even indoor experiments develop these capabilities, decrease the cost of deployment, or help policy-makers deploy RAIs more effectively. That is, they change the power relations, exacerbating the inequalities between people who have and can use the information to manipulate the environment. So, as a consequence, the distinction between indoor and outdoor begins to dissolve; it may be true that outdoor experiments lead to greater marginal increases in capability, but it is clear that even environmentally harmless computer modelling can increase one’s potential ability to use RAIs effectively, assuming the modelling tracks some relevant feature of deployment and one has some capability to translate the.¹⁰ I hasten to note that this does not mean that we should refrain from RAI research. I am only suggesting that insofar as we are concerned with domination, we should be interested in governing both indoor and outdoor research.

Information-Sharing as the Primary Mechanism of Research Equity

Geoengineering researchers are aware of the sharp inequalities between researchers in the developed and developing world. They correctly perceive that these inequalities can generate worries about the legitimacy of the decision-making process. Many have subsequently argued¹¹ that some inclusive political process needs to govern eventual deployment decisions, but the primary mechanism for maintaining the legitimacy of the research process is to ensure that scientists around the world have access to its results. This sort of transparency, again, is consistent with a risk paradigm. Having an open and transparent discussion where individuals can test and replicate various results helps ensure that people of good will can avoid serious mistakes that would lead to imprudent interventions. It will surface potential problems and decrease the chances of groupthink. Yet, transparency is an inadequate compensation for existing inequalities in resources and capability. In the face of severe inequality, any objections generated through transparency still depend upon the good will of the powerful for uptake.

To illustrate, let us return to ISLAND. In the first scenario, transparency would solve the potential domination of the dam construction. This is because we assumed no differences in capability or resources such that the only difference is one of knowledge. So, if the Morlocks wish to develop the dam for peaceful purposes and avoid domination, they can give the Eloi a copy of the plans, making their plans completely transparent and then the Eloi could counter potential flooding by building a dam of their own. Yet, in the second scenario, the reason the Morlocks were able to develop the weapon is that they had a unique set of resources or capabilities. Similarly, imagine that the Morlocks were uniquely positioned to benefit in an outsized fashion from the technology – or to suffer considerably lower costs than the Eloi for production – in ways that were not fully reducible to their knowledge or skill; perhaps the production of material for the dam required access to raw materials that were far more accessible to the Morlocks or the Eloi were positioned on the high ground such that they enjoyed a massive advantage in dealing with the subsequent flooding. In each of these cases, the mere availability of the plans would not equalize the position between the two groups. Even if the Eloi were to come upon the plans for the dam, they would not be in a position to exploit that knowledge as effectively as the Morlocks.

Yet, even if we assume a less extreme version of the example where the Morlocks are only in a more favorable position to exploit the relevant knowledge, information-sharing will at best decrease the power differential between the groups without eliminating it. The fact that the Morlocks can research or deploy the technology more effectively or efficiently will grant them superior power over the Eloi even if the availability of the relevant information makes it possible for the Eloi to deploy the relevant technology. This is an unsurprising result: if two agents begin with different capabilities, those capabilities can intersect with new information such that the combination can result in greater, equivalent, or only marginally reduced asymmetries in power. This need not only apply to technology. Without the resources and institutional support to actually participate in the international legal system, mere knowledge of, for example, human rights law will not necessarily lead to an improvement in the political position of a marginalized group. Or, at the very least, the introduction of a new legal regime will not sufficiently benefit the marginalized group such they improve their power relative to the other agents in the system.

The relevance to RAIs is two-fold. First, while many regimes argue for widespread availability of scientific research in the name of open-source transparency, they fail to require researchers to give up their intellectual property on the underlying invention. Yet, scientific transparency is quite different from legal entitlement. Much of the future work on geoengineering will involve the design of reflective particles, delivery systems, or monitoring mechanisms that can nonetheless be proprietary even if information about the consequences of those mechanisms is readily available. If researchers can use domestic and international legal systems in order to place significant obstacles on the ability of marginalized people to generate new capabilities, then information-sharing will have greatly diminished effects. The second, and more important, connection is that RAI research is and will be disproportionately performed by agents with asymmetric and superior resources and capabilities. Developed nations, where the vast majority of RAI research is done, have superior resources, engineering skill and educational infrastructure. RAI research will concern a variety of issues that will increase effectiveness, decrease costs, and grant greater ability to direct or minimize negative impacts. RAI research, therefore, could lead to the development of usable particles and delivery methods, increasing the power of the deploying agent to manipulate the environment. Yet, those advantages will accrue, or accrue at a greater level, to those who have the relevant institutional support. The developing world insofar as they are less advantageously positioned will be less capable of deploying RAIs. Thus, research into a new technology (where only one party is positioned to deploy optimally) threatens to exacerbate existing power inequalities by providing one party with capabilities that cannot be matched by the agents with antecedently inferior resources and power. In a context where the capacity to exploit information is differentiated, open and transparent information-sharing may not prevent power asymmetries from becoming more severe.

Scientific Self-Governance

The third problematic element of RAI governance is that political and legal interventions occur too late in the process. The relevant scientific research regimes are almost universally committed to scientific self-regulation. So, they believe that fellow scientists and researchers should be the primary agents when it comes to creating and then applying rules to regulate RAI research. These regimes suggest some limits to this scientific self-governance. In the event of large-scale outdoor research, the various political regimes designed to regulate deployment – should they exist – or legal regimes governing pollution could potentially be invoked in order to include other agents. And there are obvious advantages to having scientists play the primary role in collectively determining what research is appropriate. The greatest of these advantages is epistemic. Scientists have essential knowledge to evaluate the potential risks and rewards of any particular research regime; others agents will almost certainly need to rely – to at least some extent – upon the expert testimony of scientists.¹² Furthermore, the agents who will be most directly constrained by the regime will be the scientists themselves, a structure that invites all the objections of self-regulation.¹³ After all, it will be scientists that will be named and shamed, sanctioned, or otherwise constrained in their research pursuits and experiments. Careers and lives will be affected by these regimes, and researchers have a legitimate claim that their interests be represented substantively and procedurally. What’s more, excessive political interventions into the scientific process might undermine its epistemological status and one might expect scientists to be more sensitive to those worries than outside agents.

There are significant worries about the effectiveness of scientific self-governance. Scientists are strongly incentivized to continue with their research and experimentation, so it is difficult to shut down promising research avenues. Furthermore, scientific institutions lack effective and timely enforcement mechanisms. What’s more, what enforcement mechanisms that do exist may not be applicable to researchers in the private sectors.

However, even if these issues could be resolved, there would be a deeper problem. Self-constraint is an inadequate response to the possibilities of domination. Imagine a climate governance regime made up of rich and powerful high-emitting individuals who have fairly intense economic relations with poor low-emitting individuals who will experience negative impacts from carbon emissions. A set of social norms and incentives amongst high-emitting individuals causes them to check each other and to take steps to mitigate their emissions and serve the interests of the low-emitting class but without any participation or contribution by the low-emitters themselves. In other words, high-emitters are constrained but only by other high-emitters. Does this reflect a set of sufficient political protections for low-emitters such that the exercise of power is legitimately non-arbitrary? Obviously, we should be skeptical that a scheme that relied upon the good will and interests of the powerful agents to serve the interests of the powerless is going to fail. Deploying the mechanisms of social sanction against a fellow member of your class is unpleasant, expensive, difficult, and unsure. And of course, one would not wish to establish precedents that would undermine the economic usefulness of low-emitters and so the protection of their interests would always be, in this sense, indirect and limited. Some high-emitters may feel genuinely altruistic towards low-emitters, but any system that that relied upon those feelings and routinely set the economic and social interests of the powerful against the interests of the powerless would likely be unstable.¹⁴

But we might think that the system was problematic even if it was relatively successful and a stable equilibrium that served the interests of the serfs existed. There seems to be something unjust about the fact that these serfs would need to rely upon the charity of powerful aristocrats for the furtherance of their interests beyond their economic usefulness. In other words, the serfs seem dependent upon the will of the aristocrats for the protections of their rights; they cannot participate in their own protection or contest the decisions of their masters. Perhaps those masters are benevolent, but it is unjust to be so subject to others no matter how benevolent they might be. In the literature on domination,¹⁵ this is understood as the idea that true political equality requires that no person be subject to the arbitrary will of another. Arbitrariness is then understood as being subject to the decisions of another where there are no effective avenues of control or accountability requiring that the powerful take the interests of weaker stakeholders into account.

The consequences of this view for RAI research are profound. If this type of normative claim is correct, then those who are subject to the exacerbated power imbalance created by RAI research have a claim to avenues of accountability, contestation, and control when it comes to the decisions of research scientists in the developed world. In fact, not only is scientific self-governance inadequate if those collective decision-making bodies are dominated – as they must be, given their membership – by the developed world, but even additional political or legal governance by national or regional scientific bodies would be insufficient (Pettit, 2010; Laborde, 2010). After all, those political and legal agents are designed to represent the interests of their constituents. Adding political regulations to RAI research may correct for some of the limitations of self-governance, but the fundamental normative weakness of a scientific regime whereby rich and powerful members of the global and scientific elite decide, for themselves, to what extent they will accept the criticisms and serve the interests of the developing world will remain illegitimate. And this is no criticism of the motivations or sincerity of the agents developing their self-governing regime; the political structures in place simply do not make room for legitimate research decisions if they are done unilaterally regardless of why the decisions are made. The developing world needs to engage with RAI research that offers mechanisms of contestation beyond mere persuasion. Legitimating avenues of contestation and accountability cannot be mere charity.

To sum up, proposals for RAI research governance have attempted to avoid being too political. They accept that wide swathes of RAI research – including essentially all research currently being done – need not be regulated at all. Only research that has generated public controversy or generates significant environmental risks is normatively relevant. Second, insofar as these regimes are concerned with power imbalances, they argue that these imbalances can be made good through transparent information-sharing. Finally, when it comes to RAI research, the primary agents in developing and implementing these regimes will be the scientists themselves. In each case, these proposals do not take the power dynamics or political implications of their research sufficiently seriously. In fact, it seems likely that many participants in these regimes are not even aware of the ways their actions intersect with global power inequalities. As a consequence, these regimes are unlikely to prevent these inequalities from worsening as research creates more and better means for manipulating the climate.

Two Strategies for Non-Domination and SRM Research: Power Accountability and Power Equalization

It is important to see that RAI research is not necessarily impermissible because it exacerbates current power imbalances. The question is, rather, whether we can prevent RAI-domination by adopting different rules, regulations, or institutional structures. In this section, I wish to propose two ways in which RAI research governance might be able to avoid my criticisms. Yet, I want to argue that each of these proposals have significant weaknesses and that they work at cross-purposes. As a result, the decision to engage in either strategy will be controversial, generating costs and benefits. Unfortunately, the political structures to make that initial strategic decision in a non-dominating fashion does not yet exist. After all, the strategies are themselves supposed to be what resolve domination. This raises a pressing question of global justice: how should powerful agents decide amongst options for reducing non-domination when those agents are, at the moment, in a dominating position. Powerful agents must make a choice about how to respond to the potential of domination in RAI research, but even the choice to create accountability mechanisms for the less powerful will be made without those accountability mechanisms in place; how can the decision of which legitimate procedures should be put in place be itself legitimate? In a sense, the question concerns how one should behave when all available actions are, in some important sense, politically illegitimate.

Domination occurs when two conditions are met. An agent dominates another when the former has superior power over the latter and that power can be exercised arbitrarily. This implies two anti-domination strategies (Pettit, 1997, Chapter 6). The first is power equalization while the second is power accountability. When one agent becomes more powerful, non-domination demands that other agents gain equivalent power or that the power be checked and balanced such that the superior power is held accountable to those who are less powerful. We can understand the differences between these strategies by considering the power relations between the United States and the Soviet Union in the beginning of the Cold War. The success of the Manhattan Project and the subsequent development of atomic weapons initially gave the United States a decisive, qualitative superiority over potential rivals. There were two strategies for dealing with this newfound power asymmetry. One possible avenue – advocated by Albert Einstein, Bertrand Russell, and others who distrusted the ability and the political will of competitive nation states to act responsibly with the new weapons – was controlling, directing, and checking the political effects of nuclear weapons though a centralized and democratic global regime. They argued that nuclear weapons ought to be centrally controlled by the United Nations, which was in turn structured by an international rule of law and a global democracy. The United Nations, on this view, would have control over the most powerful weapons ever devised. The power differential between the UN and member states could be extreme, but the exercise of that power could be made non-arbitrary and accountable. We might, following Pettit, call this the modern conception of how to regulate superior power. Developed in a political milieu where states became more powerful domestically, Enlightenment era domination theorists argued that by dividing state power along functional lines, creating independent judiciaries, and introducing democratic politics, the absolutist authority and increasing institutional power of the state could be made consistent with non-domination of its citizenry.

There is, however, an alternative conception that has its origins in the political constitutions of Greece and Rome. In those societies, non-domination was not generated through institutional regulation but by having the relevant groups within society have equivalent power. For example, both the plebeian and patrician classes of ancient Rome had legally equal¹⁶ legislative entities that could issue binding decrees on the entire polity. In the nuclear context, we can see this as the strategy of the Soviet Union in its development of equivalent nuclear weapons and delivery systems as the United States. The Soviet Union never achieved complete parity with the United States; its deterrence triad was never entirely secure. In response, the Soviet Union developed a nuclear deterrent and pursued conventional superiority in Europe using tactics and organizations that were designed for nuclear survivability. So, the Soviet Union pursued a position of power parity with the United States through a combination of developing equivalent military technology and creating asymmetric counter-measures that substitute and compensate for the shrinking differences in capability. It is important to see that the non-dominating effects are contextual and relational. The Soviet Union was able to produce rough parity in geopolitical position viz. the United States in this specific policy domain. Despite the more general economic and diplomatic superiority of the United States, it was unable to exploit its advantage in nuclear weapons. Furthermore, the nuclear parity between the two superpowers did not ‘reset’ the power relations to before the technological development; the new parity created entirely new power dynamics and risks. It is, however, clear that the Soviet development of nuclear weapons changed the power dynamics between the superpowers, substantially equalizing them. Of course, this equalization is relational; Soviet technological developments may very well have led to greater domination elsewhere or may have exacerbated domestic injustice. The claim I am making is quite limited: the dominating potential of nuclear weapons in the relationship between the United States and Soviet Union was dramatically reduced by their mutual possession of the relevant capabilities.

When discussing RAI research, analogous policy responses suggest themselves. The power accountability strategy would involve the creation of a new or the restructuring of an existing body to create a centralized agency that will make it possible for stakeholders in the developing and developed world to come to together to legislate and administer a set of checks and balances on RAI research and deployment. We could attach these processes to the existing UNFCCC structure or create a separate management regime. The key features of the regime would be the following. First, it would require a legislative component where various stakeholders could come together, in a position of deliberative equality, to formulate research rules. Second, the regime would need a fact-finding and enforcement component where compliance and sanctions could be monitored in the event of a violation. The regime would need to have real teeth, being able to impose significant penalties on those found in violation. Finally, the regime would need a judicial component that could make impartial and independent judgments concerning violations. What I am describing would combine elements of the International Criminal Court (ICC) with its sophisticated checks designed to insure judicial independence, with the World Trade Organization’s Dispute Settlement Board (DSB) with its linkage between rule violation and sanction. From the ICC, we would add include various mechanisms to include broad-bases geographical support – a certain number of judges would need to come from each area – and safeguards, such as multiple appellate levels, to prevent any one judge from acting unilaterally. From the DSB, judges would be empowered to issue sanctions in the event of a breach. Unlike the DSB which only empowers the state subject to unfair trade practices to issue sanctions, we would need omnilateral, rather than bilateral, enforcement such that all members of the regime would be required to enforce the relevant sanctions. Also, a separate ombudsperson whose objective was to maintain the integrity of the regime and could be empowered to bring cases before the executive would be essential, especially if paired with democratic processes for selecting the members of the office (Caney, 2006). In other words, the kind of robust regime that many imagine for the RAI deployment should also be developed for RAI research. Or perhaps more accurately, we should imagine a robust governance regime that treated research and deployment as a continuous whole, understanding that decisions about research and deployment are intertwined. If this regime was created, then stakeholders from the developing world would have a set of mechanisms by which they could ensure that their interests were represented. Non-domination does not require that one always get one’s way, but it does require that one’s basic rights and one’s position in the deliberative process are ensured; this regime would be a significant step in the right direction.

The accountable power path seems to be ideal. It better captures the idea that climate change is a global problem that should be resolved collectively. Furthermore, it seems substantially less risky. There is a reason why relying upon the unilateral judgments of actors to constrain each other is unstable. Actors make mistakes, and the states that control the relevant capacities might come to be ruled by unjust governments. It can be difficult to reliably send or accurately receive the relevant policy signals, and multiplying the actors with the relevant capabilities increases the risk that misperception will lead to escalation (Sagan & Waltz, 1995). What’s more, the legislative, judicial, and executive functions of the regime help – in Lockean terms – to solve problems in the state of nature (Ripstein, 2009, Chapter 6). Nations will overvalue their own interests and undervalue others; collective decision-making amongst stakeholders leads to more impartial judgments. Further, creating a set of public rules that are enforced and administered will make it easier to adapt to new developments in technology or in existing climate risks. What’s more, if the danger of climate change passes, it will be easier for a centralized authority to coordinate the reduction or elimination of these policy tools. Insofar as people finds this strategy attractive, one of the main contributions of this paper, then, is that it shows that the non-instrumental reasons to develop this accountability regime in one context – deployment – are also non-instrumental reasons to develop this regime for research.

With those advantages, it would seem that very little could be said for power equalization. Yet, this is deceptive. The reason for power equalization in the context of RAI research is feasibility. Those who called for a global centralization of nuclear weapons failed in the face of the obstacles presented by geopolitics and state sovereignty. The political architecture that I described in the previous section would be one of the most politically robust global governance regimes ever devised. It is immensely difficult to create regimes that can level sanctions in ways independent of the geopolitical interests of the constituent states and stakeholders. Once we include impartial adjudicative institutions – which are expensive and difficult to maintain – and executive institutions free of regulatory capture, we can see the extent of the task. What’s more, we have consistently failed to see the creation of similarly robust regimes in the context of climate governance, despite the greater urgency and salience of the issue. A key reason for the potential infeasibility of the accountable power avenue is the intersecting number of affected stakeholders. The regime is sufficiently powerful and sufficiently implicates state sovereignty that it would invoke powerful interests in self-determination and security while the eventual benefits are, for the moment, considered distant and tenuous.

While the power accountability strategy is attractive but perhaps unfeasible, the power equalization strategy – at least in the context of solar radiation management – is feasible but less attractive. This can be seen, for example, in discussions of nuclear proliferation, where arguments that a wider distribution of nuclear weapons would be beneficial are met with considerable skepticism (Waltz, 1981). Nonetheless, this strategy is available in the context of RAI research. Let’s return to the desert island. In the first scenario, domination could be avoided simply by giving the Eloi the information needed to build the damn. Yet, in our second scenario, the Morlocks had access to skills or resources that the Eloi did not. As a consequence, mere information sharing was insufficient. So, let us suppose that the Morlocks wish to prevent or resolve the domination inherent to their position. Sharing the plans of the dam would be insufficient. They could provide a sufficient amount of material to build their own levees to the Eloi, but this would be problematic on a number of levels. First, they would need to provide enough to make the Eloi self-sufficient and that would require complicated and difficult policy judgments. Second, this would be an insufficient response if there was some expectation that the Morlocks would be able to build on the existing plans to develop superior mechanisms of environmental control. In that case, the Eloi would not be able to keep up. So, the clear solution is to provide some of the actual material but to also help the Eloi build their native engineering capacity such that they would be able to manufacture the material and engage in their own research program. Then, if the Morlocks developed new capacities, these now equalized engineering skills or resource access would help the Eloi keep pace. This would be especially true if regular technology transfers were part of Morlock policy.

We could adopt a similar policy with regards to RAI research. Rather than merely making information available to the developing world, the scientific community could make a robust effort to engage in scientific and engineering capability building in the developing world. Not only would this increase global ability to understand and critique existing research, it would also allow the developing world to decide for themselves if their interests would be served by RAI deployment rather than relying on the good will of scientists in the developed world. Capability building, thus, would need to go beyond educational programs but would need to include improvements in the ability of the developing world to engage – on more equal terms – with the actual research and development of reflective particles, delivery mechanisms, and monitoring tools. This, combined with open source transparency and an equitable intellectual property regime, would allow the developing world to be a RAI partner, granting them similar capabilities to manipulate the climate. Like the Soviet development of nuclear weapons, this would balance the power relations within the context of solar geoengineering. Obviously, the broader inequalities that structure our current pursuit of climate justice in particular and global justice in general would still exist, but RAI researchers would be contributing to an increase in power for developing countries.¹⁷

Building these capabilities would also have further knock-on effects that would decrease opportunities for domination. First, developing nations may have different priorities when it comes to RAI research. For example, perhaps they are more interested in developing particles or interventions that would inoculate the atmosphere from risky interventions, preventing RAIs from being deployed effectively. Perhaps there are particles that would bond with proposed reflective aersols, causing them to fall to earth more quickly and dramatically increasing the costs of the intervention. Or more optimistically, perhaps the fact that the negative consequences of RAIs will fall predominantly on the developing world¹⁸ will lead to research that focuses on risk reduction rather than cost reduction. In other words, capability building amongst the developing world may generate research avenues, reduce risk, and generate politically and epistemically useful veto points that will make RAI science better, politically more legitimate, and the engineering more innovative and responsive.

Finally, building these capabilities among the developing world will create interesting game theoretic possibilities in the context of termination shock (Ross & Matthews, 2009).¹⁹ Termination shock is one of the more severe risks of RAIs. The worry is that if we use RAI to reduce global temperatures without mitigating our greenhouse gas emissions, we will need to engage in geoengineering to an increasingly greater extent in order to compensate for the greenhouse effect of our carbon emissions. Yet, if we have to cease RAIs for some reason – a new understanding of the risks, political disruption, and so on – then the climate could undergo rapid warming such that no society could effectively adapt, generating catastrophic consequences. The widespread distribution of the ability to engage in RAI research and eventually, deployment, has two interesting effects on the dynamics of termination shock. First, if many groups are positioned to engage in RAI, it provides redundancy in the event that termination shock is caused by a political or economic disruption amongst the deploying agents. The second effect is based on the important insight that the termination shock scenario, like nuclear war, is unacceptably costly to everyone. Since the developed world does not wish to risk termination shock, the possibility that the developing world may engage in unilateral RAI could become an important bargaining chip in climate negotiations. RAI has the potential to reduce the overall climate risks the developing world faces while, at the same time, creating a future threat where parts of the world that are relatively inoculated from the immediate consequences of climate change would need to take seriously. So, a more equitable distribution of geoengineering capability could prevent disaster while at the same time allowing less powerful agents to use the threat of the disaster to equalize their bargaining position in future negotiations.

The two primary worries concerning power equalization are abuse and escalation (Gardiner, 2013). First, we might worry that the broader dissemination of the technology will encourage bad actors to use it in their own interests. Second, we might worry that multiplying the actors, in a strategic environment, might generate problematic incentives to deploy. In order to generate this sort of collective action problem, several things would need to be true. Agents would need to have preferences as to how RAIs were deployed and there would need to be path dependent dynamics whereby an agent deploying in one way would preclude an agent deploying in the second way. If those two features of RAIs are true, then we could have escalatory dynamics. The first agent could feel the need to sub-optimally deploy in a ‘use it or lose it’ scenario where they anticipate the second agent might use RAIs, thus blocking their preferred deployment scenario. In this way, RAI deployment would exhibit the same pressures as nuclear deterrence, where the possibility that one side develops an asymmetric ‘first strike’ technology undermines the other side’s ability to retaliate forces the threatened side to escalate before their technology becomes obsolete. Thus, we might get deployment that neither side wants because they feel pressured to escalate before their deployment suite becomes obsolete.

Though RAI research is still in its infancy, there is some evidence that these two features obtain but we cannot be certain. It looks like deploying to different regions, with different particles, or to differing extents and altitudes could generate different environmental and social impacts. It might matter to precipitation patterns whether, for example, particles are inserted into the atmosphere at the poles rather than the equator. Different types of particles might also affect the ozone layer in different ways. What’s more, the benefits of any particular strategy could dissipate if another agent deploys differently. Suppose that one agent decreases precipitation risk by a polar insertion and compensates for the additional expense by using reflective particles that are cheaper to produce yet more harmful to the ozone (Tilmes, 2017). A second actor comes along using different particles that are themselves harmless to the ozone but inserts at the equator. If both agents so act, then proposed benefits of ozone safe particles will be erased and the two agents will generate greater risk to the precipitation system than either would accept. And since precipitation risks seem to directly correlate with increased RAI intervention, both agents intervening would likely push the risks beyond what either agent deemed acceptable prior to deployment.

There are two potential responses to the escalation and abuse worries, given these dynamics. First, the ability to unilaterally deploy without input from relevant stakeholders characterizes the status quo. So, it seems like abuse is a severe risk at the moment without any power equalization. Second, it is important to see that the widespread capability to deploy increases the potential costs of deployment to those who do it without an international consensus. Since other agents will be able to disrupt the benefits of first deployment, then optimal RAI usage will demand coordination and those who have traditionally had little say in climate policy will have some degree of influence. Again, the goal is not to resolve all global inequalities but to ensure that particular decisions to research and deploy RAIs are made within a more equal footing. The ability of the developing world to undermine the benefits the developed world may receive from engaging in solar radiation management helps equalize their positions, though it does so in an admittedly risky way. If a developed country wishes to use RAIs in a way that maximally favors its interests at the costs of imposing risk on developing nations, then developing nations are in a position to deploy their own RAI technologies in order to change the cost-benefit calculation of the initial deployer. Thus, in order for the deploying agent to be assured that they will receive a net positive set of benefits and costs, they will need to coordinate with other agents who possess the relevant capabilities. This gives other agents greater influence and power over the eventual policy outcome than they would have had absent their RAI capability. However, since other agents can exercise similar influence, there is little advantage is going first if others can intervene.

These responses only temper but do not eliminate the risks of equalization. However, given how unlikely the accountability strategy is, it is time to think explicitly about how to manage the problems that accompany power equalization. The analogy with nuclear weapons is instructive. Centralization, either through the United Nations or an American monopoly, was either infeasible, unstable, or dominating. This forced the international system to devise ways to manage the proliferation of nuclear weapons. Of course, the features and dynamics of the Cold War and of nuclear weapons are dramatically different from the current context of climate change and solar radiation management, so we should not assume that an effective set of institutions in one case will also be appropriate in the other. Rather, the key insight of that period was that people took the lack of centralization and the relevant failure of power accountability as a given and then designed ways to limit the risks of power equalization. This included alliances structured around ‘nuclear umbrellas,’ emergency redlines for communication, and treaty regimes symmetrically limiting the relevant actors. In the context of RAIs, the sooner we realize that the current structure of unilateral power is untenable and accountability regimes are not forthcoming, the sooner we can begin to create equalization regimes that limit that strategy’s risks and encourage its upsides.

Thus, we have two fundamental strategies for reducing domination in the context of RAI research. Yet, these strategies invoke opposing dynamics. The accountable power avenue demands centralization and collective decision-making while the equal power avenue demands decentralization as we build up the capabilities in a more disparate set of agents that are not subject to collective authority. We cannot easily do both strategies, just as nuclear proliferation continues to make global regulation of nuclear weapons more difficult. However, while de-centralizing capability and centralizing authority often work at cross-purposes, there are sources of hope. Perhaps the more equitable political contestation the power equalization strategy makes possible will contribute to the generation of the power accountability strategy in the long run. The point of this analysis is not to discourage the pursuit of power accountability but that avoiding some forms of serious injustice require that we think hard about how to manage the power equalization strategy. Ultimately, this is how the world barely avoided catastrophe in the case of nuclear weapons and it should not be surprising if such muddling through is required again.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. For justice-based discussions of land use, see Buck (2016) and Shrader-Frechette (2013).

2. Henceforth RAIs, for reflective aerosol injections. Some (Svoboda, Kellar, Goes, & Tuana, 2011) focus on sulfate aerosols, but sulfates may not be the particles that are actually used. Some (Weisenstein, Keith, & Dykema, 2015) are considering, for example, aluminum-based aerosols. Similarly, the focus on stratospheric aerosol injections assumes a deployment strategy that, while very likely, is not absolutely guaranteed. Thus, I use the broader category in this paper.

3. This is due to interaction effects between the cooling consequences of an albedo increase – reflecting more sunlight into space – interacting with the warming effects of the greenhouse gases that remain in the atmosphere. See Tilmes et al. (2013).

4. As an example of this distinction, consider bio-energy, carbon capture, and storage technologies (BECCS). These technologies generate considerable environmental risks due to their extensive land and water use, but they also have the potential to block many of the worst climate impacts. However, due to how costs are distributed and the relatively small effect of any individual act of BECCS, these technologies rely on broader acceptance by stakeholders and need societal support. It is harder for a group of agents to coordinate in order to control the environment. Thus, managing the risks of BECCS is a different problem from managing the possibilities of human abuse and domination presented by RAI.

5. Surveys of marginalized groups about geoengineering consistently show worries that move beyond the strict risk profile of solar based geoengineering (Carr & Preston, 2017).

6. This kind of control could include a greater ability to manipulate precipitation levels, injection sites, particle types and could include potentially include interventions into extreme weather events. In other words, a suite of geoengineering technologies could, at some point in the future, give agents a significant ability to manipulate the environment. In the far future, we can even imagine swarm nanotechnology that can great localized RAI effects. While we are far from these capabilities, the logic of RAI research is in the direction of greater precision. I thank an anonymous reviewer for pressing me on this point.

7. Pettit (1997), Ripstein (2009), and Iris Marion Young (1990) are all examples of the republican political tradition.

8. Morrow, “International Governance of Geoengineering: A Survey of Reports on Geoengineering Governance, 2009–2015” FCEA Working Paper Series: 001 (2017).

9. A typical but by no means uncommon of an example of ‘research first’ for outdoor experimentation but essentially no restrictions on indoor experimentation due to worries about lock-in is that of Andy Parker (2014).

10. Does this sort of reasoning also apply to merely thinking about new technologies that may generate a power imbalance? After all, if modelling can increase knowledge about how to manipulate the environment, so can merely taking the time to think about, imagine, speculate, and deliberate upon geoengineering. Can this kind of activity contribute to domination I believe the answer is a deeply qualified, ‘Yes.’ That is, it does not strike me as obvious that some thoughts could be, in and of themselves, unjust without the appropriate political order; this shows how much we rely upon good institutions in order to act ways that are consistent with everyone’s freedom. However, two significant qualifiers need to be included. First, most technological development these days is sufficiently complicated that without sophisticated computer modelling, we cannot expect much progress. So, usually, the possibilities for domination made possible by mere thought are de minimis. Second, any attempt to directly restrict freedom of thought – as opposed to incentivizing other research or banning more substantial research down the line – will almost certainly be dominating in its consequences, intended or otherwise. So, we should not infer from the fact that some thoughts could lead to new possibilities for domination that state is permitted to coercively restrict those thoughts.

11. A single but influential example is Rayner et al (2013) that emphasizes the role of information sharing in establishing the legitimacy of geoengineering research.

12. Of course, others have essential knowledge as well, but scientific self-regulation is committed to the idea .

13. See Roy Baumeister and Todd Heatherton (1996) for a summary and, as a specific example, consider the failures of banks to regulate their risk exposure to subprime mortgages in The Financial Crisis Inquiry Report (2011) especially Chapter Four.

14. This is, for example, one reason to think that principles of justice apply to the basic structure; principles that applied directly to individuals would be burdensome and excessively demanding. See Rawls (2001, Section 15 and 16) and Cohen (2001).

15. E.g. Pettit, (1997, pp 116–118) and see Stilz (2009, pp. 67–68).

16. Of course, this was not sufficient for power between the classes to be equal, but it is an example of decreasing domination by granting additional legal powers rather than making superior power accountable.

17. One could argue that this does not help very much because the developed world will simply prevent the developing world from deploying RAIs and so capability building will be irrelevant. First, various positive benefits to the developing world’s research agenda, including the focus on what would best serve their interests and how to block RAI should that be necessary, would remain untouched. Second, the developed world would need to pay economic and reputational costs to block RAI deployment, costs that could play a role in the negotiations I describe below. Third, a developing world coalition that aimed at RAI deployment would be difficult for the developed world to block. Capability building does not change large scale economic and military inequalities, but it would reduce their effects in this particular context. I thank an anonymous reviewer for pressing me on this point.

18. E.g. precipitation disruption will be concentrated in the developing world.

19. Some SRM researchers (Reynolds, Parker, & Irvine, 2016) have argued that termination shock is not necessarily a significant problem for SRM deployment, but I do not find the arguments especially convincing. First, the paper ignores some of the scenarios – such as additional knowledge about SRM effects – that might lead to rapid termination. Second, the paper tends to make overly optimistic assumptions about the motivations of the actors and the capabilities of international civil society in order to safeguard against termination shock.