The technical Pope: practical and human responses to artificial intelligence

Matthew Morris casts a computer scientist’s eye over Leo XIV’s new document on AI and thinks the Pope is right to stress the priority of the human over the technological.

On 15 May a math major from Villanova University signed a document in Rome. This document would go on to be publicized in almost every country, major news network, magazine, and even within tech companies. The world had just received its first papal teaching on artificial intelligence (AI), in the encyclical Magnifica humanitas (MH).

But the document covers far more than just AI: it presents a sweeping history of the Catholic Church’s teaching on social questions, lays out fundamental principles for how one should view the human person, identifies modern narratives driving technological progress, discusses how truth might be safeguarded in this age, develops existing teachings on the dignity of work, warns how technology can endanger our freedom, and teaches strongly against the normalization of war in the modern world.

In a recent Adamah Media article, Edward A. David explained the essence of the encyclical, which is intimately tied to its name, ‘Magnificent humanity’ – in this new technological age, we cannot lose sight of the greatness of our human nature.

AI cannot dictate what we become, and any technology must be at the service of humanity.

In this article, I would like to focus on three ideas.  Firstly, change is not inevitable simply because we have developed some technology which can be applied – we need to adopt technology wisely. Next, I will offer some practical takeaways from the encyclical that everyone can incorporate into their lives. Finally, I will consider how we can design systems that match the ethical and technical requirements the Pope is calling for.

The illusion of inevitability

It is easy to fall prey to the idea that the development of technology is an unstoppable process, that we must all go along for the ride, that we must adapt or perish. This transforms humanity into what, in 1954, the philosopher Martin Heidegger termed the ‘standing reserve’ – we are reduced to a resource, stockpiled, and put on standby to be used by the technology. The encyclical raises the same concern:

At the root of these problems lies a technocratic and post-humanist mentality that tends to regard the human person as an object to be manipulated or a resource to be optimized, removing all safeguards against the unchecked pursuit of profit. (MH 172)

To give an example of hope against this message of inevitability, we first cloned a sheep in 1996 and have been able to clone humans since at least 2001. However, over 70 countries have explicitly banned human cloning, and society has largely rallied against it. When it comes to developing AI, the Pope calls for prudence, rigorous evaluation, and sometimes a genuine slow-down in adoption.

It is not enough to invoke ethics in the abstract; robust legal frameworks, independent oversight, informed users and a political system that does not abdicate its responsibility are required. Otherwise, change will be governed only by technocratic thinking and presented as necessary and inevitable, ultimately imposing rules shaped by those who control data, infrastructure and computing power. (MH 106)

Many today are realising the dangers of technology that has already been developed and is common among us: parents are restricting their children from accessing technology at a young age, adults are abandoning smart phones in favour of ‘dumb phones’ that can only text or call, and some workers are rejecting the integration of AI into their jobs. 

Artists have been slamming AI-generated art, and the gaming community has pushed back against using AI in game design to the extent that studios have been forced to scrap entire projects and walk back their use of the technology.

We do not need to cram AI into every corner of society simply because we can: rather, we should think carefully and prudently about where it can genuinely advance the common good and respect human dignity.

Practical takeaways

Whilst much of the encyclical is aimed at lawmakers, people involved in the development of AI, and those in management of technology companies, the Pope also provides a lot of concrete advice that applies to everyone. He emphasizes that the task of building towards the common good is one to which we are all called: everyone has their own responsibilities and opportunities.

Certainly, not everyone has the same power to make a difference. There are those who govern, make investment decisions, lead institutions, conduct research, educate, produce or provide information, and then there are those who only seem to live their daily lives. Yet, no one is without responsibility. We all have our own areas for action, and it is precisely there — and nowhere else — that we must choose whether to fuel the mentality of force (even if only through indifference, cynicism, lies or hatred), or to preserve the mindset of peace (with truth, moderation, closeness and care). (MH 212)

First, he calls on us to ‘disarm our words’. Both in general and in our use of the internet, we should eliminate any of our unjust prejudices, speak truth, give wise advice, support those who need comfort, and call out injustice when we see it. 

In a world ever more filled with fake information easily created and distributed by automated systems, we need to be careful with what we share, believe, and consume.

When using AI tools, we are also warned against becoming excessively reliant on them, searching for ready-made answers rather than grappling with the truth, thereby weakening our personal creativity and judgement. 

Numerous scientific studies have demonstrated how using AI tends to make you less creative, worse at problem solving and critical thinking, more likely to give up on a task if you encounter difficulty, and less able to respond to complex situations. 

Since the invention of the calculator, of course we have become worse at adding large numbers together. But how central to our human flourishing is the task of adding large numbers, compared with the fundamental abilities discussed above that are at risk of being eroded?

The encyclical argues that the essence of our humanity is our capacity for relationship and love. To safeguard this, the Pope calls for us to be ‘martyrs of everyday life’, in caring for others, educating them, and comforting them. Such a way of life requires perseverance, memory, and interior conversion, so as to continually begin anew.

The quality of a civilization is measured not by the power of its means, but by the care it is able to offer, by its ability to recognize the other as a face not merely as a function. The ability to care for one another is a fundamental dimension of our humanity, one that is learned and mastered through lived experience. Reading stories to a child, offering company to an elderly person and arranging a home so that it is welcoming are simple gestures often rooted in family life. They teach us to value care at a societal level and train us to recognize others as persons worthy of attention. (MH 114)

We must also be careful not to view our human limitations and suffering as defects to be overcome. When it comes to AI in particular, I suggest we need to be ready to put work into things which AI might well do better than we can. One’s inability to write a letter to a friend that is as eloquent as the one from ChatGPT is not a reason to simply delegate it to the AI. Work through the struggle, write the letter, and know that the end product has more value simply by virtue of it being written by you.

Let me offer a personal reflection here: in a free market, the consumer generally has control over where they put their money. The Pope warns that we must not sacrifice our humanity for the sake of efficiency. 

So perhaps this means we ought to consider paying extra for the painting created by a real human rather than for AI generated art, boycotting the companies which lay off thousands of staff members to automate their jobs with AI and robotics, and seeking out products and services from the companies which genuinely prioritize the humans at their centre (insofar as we have the financial means and knowledge to do so).

Technical solutions

Finally, let me offer a few suggestions as to how we can design technical systems which match the ethical requirements the Pope is calling for.

In seeking to address some of the issues around AI, the encyclical calls for transparency, accountability, shared ownership of data, and analysing the biases of models. These are issues which are not easy to solve from a technical perspective. I have just finished three and a half years of a doctorate in Computer Science working on ways to obtain explanations from AI models, and there is a whole community of people working on these kinds of problems.

First, it is worth noting that the Pope’s teachings, whilst given in the context of AI, apply far more generally to any automated system that is used to aggregate data or make decisions about people. For most of human history, we have done statistical analysis on populations that, whilst useful, can pose many of the risks that Pope Leo identifies. 

These risks include the possibility for bias, reducing human beings to a resource, and removing human agency in decision making. They have been amplified in the modern world, from the development of probability theory in the 1600s, to electronic computers in the 1940s, the world wide web in the 1990s, the breakthroughs in deep learning in 2012, and the recent rapid advances in large language models (such as ChatGPT).

The first commercial credit scoring algorithm was built in 1958, and similar algorithms are now used everywhere across the financial sector. The COMPAS software, created in 1998 and used by judges across the US, has been found to be racially biased when given cases of people with otherwise indistinguishable criminal records. 

In 2018, Amazon had to scrap an internal tool that automatically reviewed resumes because it had learned to reject female applicants by default. 

All these cases come from before the proliferation of modern AI but are still a cautionary tale to warn us that AI is never totally neutral: it can carry the biases present in society or of those who develop the systems, as Pope Leo himself points out on various occasions.

Whilst there is nothing inherently wrong with using data and statistical methods to gain insights into populations or individuals, we must be cautious about thinking that mere data can fully capture the essence of a human person, or a group of people. Three times in the document, the Pope warns against people being reduced to data.

When data and algorithms influence credit distribution, personnel selection or access to services and opportunities, it is necessary that decisions be understandable, contestable and subject to oversight, so that individuals are not reduced to mere profiles. (MH 164)

When it comes to transparency, the core technical difficulty is how companies can share details about their processes, without giving up their intellectual property which is so crucial to their financial viability, and without exposing the private data of individuals registered with the company. 

The mathematical framework of differential privacy provides ways in which info about datasets can be released without compromising the privacy of any individuals within the dataset. For large AI models, instead of publicising the entire model which the company spent a fortune to train, it is possible to release smaller ‘distilled’ versions of the model, which can provide some transparency,

For accountability and exposing model biases, the growing field of ‘explainable AI’ has much to offer. An example of a counterfactual explanation is as follows: imagine you were just denied a loan by an automated system. A counterfactual explanation will tell you what you can change about your financial arrangements to make the system change its mind and give you the loan. This can provide ‘the possibility of recourse’ the encyclical calls for.

For accountability, data-attribution explanations can tell you which parts of the data are responsible for an AI making a certain prediction, which can help you know what part of the data is to blame when something goes wrong. 

However, given the reach of modern AI systems, the number of steps in their creation, and the myriad of people involved in developing and deploying them, we need to consider much more complex systems for accountability, involving mathematics and policies, so that we can identify what (and more importantly, who) is responsible for an AI model’s prediction. As Leo writes:

For AI to respect human dignity and truly serve the common good, responsibility must be clearly defined at every stage: from those who design and develop these systems to those who use them and rely on them for concrete decisions. In many cases, however, the internal processes leading to a result remain opaque, making it harder to assign responsibility and correct errors. This is where accountability becomes crucial: the possibility of identifying who must ‘account’ for decisions, justify them, monitor them, and, when necessary, challenge them and remedy any harm caused. (MH 105)

Finally, the Pope calls us to think creatively about how data can be managed as a common good. 

This is especially important in our current context, where large AI models rely on mountains of data to learn, and a company’s edge in the market is determined more by the quality of their data than by their choice of AI method. AI companies have repeatedly been shown to have illegally obtained and used data for their models, and workers around the world are exploited to provide data-labelling services. 

The encyclical also states the following about the data describing individual persons:

Here lies one of the most urgent moral challenges of our time: to ensure that shared knowledge becomes a true common good rather than an instrument of dominance. This requires restoring to individuals not only the data that describes them, but also the ability to decide how it is used, by whom and for whose benefit. (MH 178)

This is very difficult from a technical perspective: it needs to be established where everyone’s personal data has ended up (which, in our interconnected modern world where data is constantly moving both legally and illegally, is a serious challenge).

This should then be enforced with policy and law to ensure that individuals have control over their data (see GDPR legislation in the EU, for example), with systems created to enable the purchasing and sharing of an individual’s data, where the individual remains in control the entire time. Perhaps the domain of cryptography has much offer in this area.

Such goals require the collaboration of policymakers, researchers, and also individuals, in taking responsibility for their data and the data of others who are more vulnerable and prone to exploitation.

A call to action

Pope Leo is one of the first world leaders to provide a strong statement on AI. The responsibility now falls on us: do we want to create a world that is more human and ordered towards our common good, or do we want to become a resource at the service of the machine, as the currents of technology sweep us away?