In the discussion about Magnifica humanitas, a great deal of concern is expressed about the concentration of power in a small number of companies.
The concentration of economic power in an area such as AI can arise for a number of reasons. There can be natural monopolies caused by very high fixed costs of research and development which make it difficult for new entrants to enter the industry. And there can be network effects or ‘winner takes all’ situations. In such cases, the success of a particular company might rely on there being a large number of users interacting with each other and sharing information. This is the main reason why social media platforms tend to be monopolistic.
The dominance of large firms is not always a problem. Consumers may prefer that situation to the realistic alternative market structures. And, often, waves of innovation mean that today’s monopoly becomes tomorrow’s forgotten giant.
But, when it comes to AI, there are real fears that its abuse by a powerful corporate entity could damage the vital functions of society.
What can we realistically do?
We might hope that the political authorities can regulate ‘big tech’ to more effectively promote the common good. They could try to eliminate existential threats, ensure that monopoly power is not abused and try to promote the use of AI in ways that genuinely lead to integral human development.
In Church documents following the financial crisis, such as Caritas in Veritate and Oeconomicae et pecuniariae quaestiones (‘Considerations for an Ethical Discernment Regarding Some Aspects of the Present Economic-Financial System’), a considerable amount of confidence was placed in states regulating the financial sector to avoid crises, such as the financial crisis of 2008. But how do we make sure that statutory regulators promote the common good?
Statutory Regulation
Regulators can become captured by the regulated entities. Furthermore, they simply do not possess the knowledge to be able to perfect markets. Indeed, their actions may cause more harm than good: before the financial crisis, financial regulation expanded dramatically, but there was nothing that regulators did which made the crisis less likely or its effects less problematic. Securitisation was welcomed by regulators and promoted by regulatory codes, for example. And reports by prominent economists, such as Joseph Stiglitz, seemed to hugely under-estimate risks involved with complex financial instruments that were underwritten by US taxpayers in government-sponsored agencies.
The lessons from these, and many other similar errors by those close to the government regulatory system is not that regulators are especially error prone but that they are not especially omniscient. There are limits to what we can expect from statutory regulation.
But when it comes to AI and big data, there is an additional problem. Just as we might be worried about the abuses of AI and big data by large corporations, might we not similarly be concerned about abuses by the state? Are we comfortable with the current administrations in the world’s largest countries (or, indeed, many of the smaller ones) being the final arbiter of how our data is used and how AI is used? We are hardly in an era of virtuous government. When it comes to AI, there is a real danger of accidents or malign actors creating chaos in economic or social systems or of promoting vicious forms of warfare. AI can develop, to use a phrase which is not technically correct but conveys the challenge, ‘a mind of its own’, outpacing any attempts to control it both in general and in specific situations. Even if we limit ourselves to the G20, how many of those countries might we be concerned about being the ultimate gatekeeper of how AI and big data are used?
Shared Responsibility
In the ‘frequently asked questions’ document, published with Magnifica humanitas, a pertinent question was raised about the state and the regulation of AI. Unfortunately, that document seems no longer to be available online. The answer given was both sensible and consistent with the long-standing teaching of the Church down the ages. It is a messy solution, but we are in a messy world. It is a solution that is mentioned in the encyclical itself.
Specifically, Pope Leo wrote:
[T]he Social Doctrine of the Church calls for a shared responsibility. It asks that these processes be guided with foresight: by institutions capable of regulating without stifling, and protecting without taking over; by businesses that recognize work and dignity as measures of success; by intermediary organizations and educational communities that rebuild trust and relationships; and by citizens who cultivate responsibility, moderation, discernment and a sense of truth. Only in this way can innovation genuinely serve integral human development, rather than becoming a source of exclusion and dominance (181).
These sentiments were echoed elsewhere in the document.
Indeed, we can learn from how we regulated markets in the past. In finance, for example, there was very little role for the state in regulating markets before the mid-1980s. Regulation came from professions, bodies such as exchanges, industry agreements, and so on. Their replacement by state regulatory bureaus has not been an obvious success – they can concentrate power and risk instead of dispersing it, as we saw in the financial crisis. Within a diverse regulatory eco-system, states and international institutions should play roles that are appropriate to their nature. Understanding exactly what international institutions can do and what states can do (which other bodies cannot do) to appropriately limit the dangers of AI is likely to lead to better results than an approach in which committees proliferate without any clear understanding of their role and limitations and in which the state alone is responsible for regulation. Instead, we can work to create a ‘polycentric’ ecosystem, within which there is also an important role for the family and schools in promoting virtuous behaviour in all markets, including in relation to AI, tech and big data.
Civil Law
Another long-forgotten instrument for regulating behaviour indirectly is tort law. Those who misuse AI in such a way that it creates harms should be held accountable through the civil legal code. This is not just an instrument of justice; the requirement to provide redress provides incentives to avoid recklessness. To be effective, tort law requires a clear understanding of citizens’ rights as well as effective protections of the right to property, including intellectual property. This includes the intellectual property of citizens – we often forget that the purpose of tort law is to protect the weak from the strong, not the other way round.
Overall, this approach would apply principles developed by Elinor Ostrom derived from her research on the management of environmental resources. Her work, surprisingly rarely mentioned in discussions about Catholic social teaching, is certainly compatible with that teaching. Polycentricity does not seek perfection, but workable and practical solutions with different entities within society taking on responsibilities appropriate to their nature.
Concentration and Dispersal: Further Advantages of Polycentricity
One final consideration is worth bearing in mind. There is an assumption, which seems to come through Magnifica humanitas, that the problems that arise with AI and ‘big data’ come from a small number of big players: ‘When such power is concentrated in the hands of a few, it tends to become opaque and evade public oversight’ (95). This is not necessarily so, and it has not been the experience in banking. The much more dispersed US banking model has been historically much more fragile than the more concentrated UK system. The UK experienced the crash of 2008, but that was the first in 140 years and many previous crises were averted by industry (and central bank) action – something made possible by the concentration of the sector. A relatively small number of parties had a strong interest in resolving the problems and could overcome collective action problems.
My hunch would be that the most dangerous events in AI will actually come from somebody working on something we don’t know about. It is the hundreds of millions of actors, some of whom will be deliberately malign and who will be extremely innovative, who may well create the fragilities about which we should be most concerned. A large company with a huge research lab might create a deadly bacterium using AI techniques. But there is good reason to be more concerned about the large number of small organisations that may well cause problems about which few people know until it is too late. There are, currently, literally millions of cyberhackers in the world. Whilst we might be worried about Google, Nvidia, Amazon, and so on, it is not clear that they are the biggest dangers in the world of AI. This is not an argument for or against state regulation. However, the more society as a whole is orientated towards promoting the common good as a result of the actions of families, educators, civil society, the civil legal code and tort law, supported by appropriate, but limited, regulatory responsibilities for states and international institutions in a polycentric order, the more exceptional bad actors will be. We cannot perfect the world; we cannot expect everybody to be a paragon of virtue; but we can try to make the promotion of vice less likely and create structures that seek to promote virtue.
