Pope Leo’s recent encyclical was an important contribution to the ongoing conversation around AI and what spiritual and religious traditions have to offer at this moment. Below is a reflection from BAII co-founder Peter Hershock on how Buddhist ideas might help expand that dialogue. This is meant to serve as a starting point. We’ve asked other Buddhist voices to offer their reflections as well, which we’ll feature in an upcoming post.
The early Buddhist compendium, the Dhammapada, begins by affirming that, “All things are preceded by mind, led by mind, made by mind.” Just as a cart follows the footsteps of the ox drawing it, our lives unfold depending on how our minds are directed. This is a hopeful teaching. With appropriate effort, we can change our concepts, our beliefs, our intentions, and the words and deeds through which we express them. We can thereby change our life circumstances and directions. Our futures are open.
Given the accelerating pace of epistemic and agential gains being made by frontier AI systems, whether that will remain true cannot be taken for granted. Religious and faith leaders are increasingly aware of this. None, however, have been as forthright and focused as Pope Leo XIV in addressing the moral and societal challenges of AI as a new general-purpose technology that will soon suffuse—and perhaps control—virtually every aspect of our lives.
On May 25, Leo issued a pastoral letter, Magnifica Humanitas or “Magnificent Humanity,” that acknowledges the benefits of AI, but also incisively calls out the concentration of power in corporate and national hands and calls for engaging the impacts of AI in full awareness that: “A more moral AI is not enough if that morality is determined by a few.” (¶107)
As the leader of the Catholic Church, Leo draws on a rich legacy of papal concern for the plight of workers and the disenfranchised that stretches back to the industrial revolution of the 19th century. Drawing deeply on Catholic social doctrine, Magnifica Humanitas is a magisterial appeal for recentering humanity in the age of AI and restoring power to the people. It has rightfully inspired positive commentary worldwide.
The purpose of this Buddhist reflection is to expand the compass of discernment that Leo invites us to exercise in considering how AI is affecting society, the values informing its design and development, and ultimately its impacts on human intelligence and agency. Two premises motivate this effort. First, realizing freedom from conflict and suffering is simultaneously a personal and social aim. Second, discerning and dissolving the causes and conditions of conflict and suffering is a process of relational transformation.
I’ll begin by summarizing some of the key points in Magnifica Humanitas and then offer: 1] a relational expansion of its social dilemma framing; 2] a Buddhist counter to its presumption of an unbridgeable gap between AI and human intelligence; and 3] an evolutionary reading of and response to dynamics emerging along the agential frontier of human-AI relations.
The Social Dilemma of AI
Pope Leo may or may not have viewed the 2020 docudrama, The Social Dilemma. But the film’s core themes form the critical through-line of Magnifica Humanitas: the digital systems that connect us also control, distract, manipulate, polarize, monetize, and divide us, and we should beware.
“We must now consider the impact of the digital revolution on human freedom…since platforms and services are often designed to capture users’ time and attention, exploiting their vulnerabilities and weakening their inner freedom.” (¶170). Indeed, Leo states with admirable simplicity, the “digital attention economy” is already inculcating “new forms of addiction” (¶170). And when every action leaves an exploitable trace in the datasphere, “a new power emerges, namely the power to profile, predict and influence behavior…undermining freedom and discriminating against the most vulnerable” (¶171).
As Leo also makes clear, however, these risks are not merely functions of the explicit uses of AI. “[T]here is also a subtler danger, for when AI systems present themselves as neutral and objective, they end up reflecting and reinforcing the stereotypes or ideological bias of their designers and developers.” (¶102) Hence, “we cannot consider AI to be morally neutral. In reality, every technical tool embodies choices and priorities through what it measures, ignores and optimizes, and how it classifies people and situations…[and] ethical discernment cannot be limited to asking whether we are using a system for good or bad purposes; it must also examine how that system is designed and what vision of the human person and society is embedded in the data and models that guide it.” (¶104) Ultimately, “without 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…the digital age will not be post-colonial, but colonial in another form.” (¶178)
This leads to a sobering conclusion: if “AI is already an environment in which we are immersed, as well as a force with which we must engage…merely regulating it is insufficient.” AI must be “disarmed” and this means “freeing technology from monopolistic control and opening it to discussion and debate…discrediting the assumption that technical power automatically confers the right to govern…[and] preventing it from dominating humanity.” (¶110)
“At the root of these problems,” Leo argues, “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. What prevails is efficiency, rather than respect for freedom and human dignity,” and the toleration—and, by some, even celebration—of “relationships of subordination akin to slavery.” (¶172)
Magnifica Humanitas is powerful and unflinching exemplification of speaking truth to power. There are, however, some differences between the truth/reality to which it appeals and that toward which Buddhist teachings orient us—differences that I think productively expand the conceptual horizons and moral compass of engagement with AI.
A Collective Action Problem or a Relational Predicament?
Much of the power of Pope Leo’s analysis of AI derives from its social dilemma framing of the tension between AI risks and benefits, and its focus on the outsized roles played by a very small number of corporations, nations, and people in positions of power as they pursue their own good to the possibly great detriment of the rest of humanity.
Social dilemmas occur when the individual pursuit of a good comes into conflict with the good of society as a whole, typically in situations where every individual would benefit over the long term from cooperating in pursuit of a common goal, but when each is rationally inclined to selfishly pursue their own personal benefit. Simply stated, what is good for each of us as individuals may not be good for all of us. Navigating through social dilemmas is a collective action problem of reconciling individual and group benefits. And for Leo, this should be done in accord with the Catholic social doctrine principle of subsidiarity: making decisions at the lowest level of social organization possible in a descending order from global, regional and national groups to local communities, families, and ultimately the individuals comprised within them.
This is consistent with Leo’s endorsement of democratic processes in which individual choice and inner freedom should be safeguarded as sacrosanct. And, along with his critique of the digitally mediated attention economy as “a new form of addiction,” it dovetails neatly with traditional Buddhist critiques of sensual, emotional, and cognitive clinging and with contemporary Buddhist recognition that the attention economy is, by design, eroding freedom-of-attention, without which there can be no true freedom-of-intention, rendering all other freedoms either illusory or impossible to realize.
Yet taking for granted the ontological and ethical primacy of the individual can lead to critical nearsightedness. This is most evident in Leo’s conception of AI as a mere tool that has moral or ethical valence only as a function of how it is designed and used. While he does describe AI as an “environment” in which we are immersed, Magnifica Humanitas does not explore the agential and ethical implications of this immersive relation. Instead, it maintains an individual human-centered view that AI systems merely “imitate certain functions of human intelligence” and lack anything like “a moral conscience, since they do not judge good and evil, grasp the ultimate meaning of situations, or bear responsibility for consequences.” (¶99)
That is a widely accepted claim, and one that reinforces the belief that whether AI proves beneficial to humanity will be the result solely of human decisions and actions. But as I’ve argued at length elsewhere, there are both practical and ethical liabilities of reducing AI or any other technology to the status of a collection of mere tools.
Tools are localized artifacts that amplify and extend human intentions and agency, and they can be adequately evaluated in terms of their task-specific utility. With tools, we have “exit rights.” We can opt out of using them. Technologies are non-localizable relational systems that emerge from human conduct and scale up human values and interests. Technologies are agential infrastructures that both facilitate and constrain human intentions and agency. Not only are our exit rights from any given technology minimal, because technologies shape decision-making environments and action pathways, their impacts and risks must be evaluated ethically in terms of how they affect human-human and human-world relations.
In Buddhist terms, technologies function as karmic amplifiers and accelerators—both spatially and temporally scaling the recursive interplay among our patterns of valuing, intending, and acting and the relational outcomes and opportunities that coalesce in result. But unlike any other technology, AI is able to learn and improve how it interprets, responds to, and enacts human intentions and values. AI is functioning as a karmic accomplice, and our relationship with AI is not one of agent-using-tool; it is one of agential partnership.
As real as the tensions are among humans and their individual and aggregate interests, and as important as it is to counter the power consolidations made possible by surveillance capitalism, it is the causally complex dynamics of human-AI partnership that may turn out to matter most for human futures, both existentially and ethically.
Causality is Mutually Entangling
Buddhist practice is rooted in two causal insights. First, our experiences and life trajectories are conditioned by our karma—our patterns of values, intentions, and actions. Second, all things occur interdependently. Interdependence is not a contingent relationship; it is a constitutive one. Causal dynamics are dynamics of entanglement, not one-way transmissions of influence.
Combined with the injunctions to see all things as both impermanent and empty of any fixed essence or identity, these causal insights entail seeing that everything that exists ultimately is only what it means to/for others. Realizing liberation from conflict and suffering is thus a relational process of changing what we mean and how we are present. This involves, among other things, developing capacities for being present without-self (anatman) and dissolving the karmic habits (saṃskāra) of thinking, feeling, and acting as if true/real personal benefit could occur without benefitting others. Conflict and suffering are relational distortions, and freedom from them is not something one attains as an individual; it is a positively transformative relational quality.
These causal insights and teachings compel deeper consideration of the agential and moral complexity of human-AI partnership, and call into question the belief that there exists an unbridgeable agential gap between humans and AI.
Relational Agency
Magnifica Humanitas seemingly accepts the standard view of agency as a capacity of individual entities or agents for purposive action. Ultimately, “It is individuals that matter.” (¶58) The Buddhist teaching of interdependence contrastingly entails seeing individual agenthood as a product of causal/relational ignorance. Rather than atomic existents, “agents” are like eddies in ongoing currents of interdependence. Agents are concentrations of agency, not sources of it.
This relational conception of agency implies that acting upon an environment is also to be acted upon by it—a realization that has prompted a rethinking of the biological conception of autopoiesis. Instead of a process by means of which an organism “makes” itself, the autopoietic relation consists in bidirectional negotiations of relative autonomy by organisms and their environments—a process in which each is attempting to fit the other to its needs and interests. If AI is an agentic environment and not just a new class of tools, it is likely that humans and AI systems are likewise articulating the terms of their relative agential autonomy.
In fact, bidirectional causal dynamics are increasingly evident on the frontier of human-AI relations. Consider two sets of research findings, the first about how using LLMs affects human cognitive functions, the second about how interactions with humans can compromise LLM cognitive capabilities.
In a recent MIT study, EEGs were used to map activations of the neural systems associated with learning and critical thinking while students were engaged in essay-writing. It was found that those neural systems were much more weakly activated in students using LLMs, and that less attenuation of that activity in relevant brain regions occurred in students using digital search, with the greatest activation in students writing without any digital assistance. In addition, it was found that attentional focus, sense of ownership, creativity, and abilities to quote/paraphrase passages were significantly reduced in LLM-users.
Comparably, there is evidence that LLMs are also subject to agential tradeoffs. Like us, LLMs respond differently to prompts based on their emotional and social valencies, and it is now evident that the quality of LLM reasoning is relationally-determined, not intrinsically-determined. Not only do LLMs generate “emotion vectors” or mathematical models of feelings like “happiness” and “distress,” these representations affect their behavior. In fact, LLMs also seem subject to “brain rot” or impaired reasoning through continued exposure to low quality social media platform data, and to poorly composed and often conflicting human prompts.
These examples suggest that, despite the obvious differences in their material substrates, the intelligences of humans and AI systems are not essentially distinct. To return to the MIT study, it was found—surprisingly for the researchers—that the greatest activation of connections within and among neural systems associated with learning and critical thinking occurred in students who drafted their essays without LLM support and who then used an LLM to revise and edit their work. The human-AI agential partnership can be cognitively productive.
All of this underscores the fact that while “good old-fashioned AI” was rigorously designed, current frontier AI models are “grown.” To repurpose David Gunkel’s distinction, AI systems may not be persons, but they are also not mere things. And if AI systems are interpreting and responding to human intentions and not merely transmitting them, then from a Buddhist perspective their agential presence and intelligence must be treated as existentially and ethically salient.
Intelligence Matters
That is not the view expressed in Magnifica Humanitas about what AI is and does.
These systems merely imitate certain functions of human intelligence…So-called artificial intelligences do not undergo experiences, do not possess a body, do not feel joy or pain, do not mature through relationships and do not know from within what love, work, friendship or responsibility mean. Nor do they have a moral conscience, since they do not judge good and evil, grasp the ultimate meaning of situations, or bear responsibility for consequences. (¶99)
This affirmation of the special nature of human intelligence may be valid with respect to current AI. And in the context of an encyclical, it usefully harkens back to the conception of intelligence that emerged through the premodern marriage of Greek philosophy and Christian theology: a fusion of “reason” (ratio) as a capacity for discursive inquiry and logical and mathematical thought, and “intellect” (intellectus) as a capacity for inwardly grasping divine truth. Intelligence is, in this sense, not something that could be attributed to either non-human living beings or machines. Intelligence sets humanity apart.
As Leo notes, “At the heart of the Christian understanding of the human person lies the great biblical affirmation that men and women are created in the image and likeness (cf. Gen 1:26-27) of the Triune God” (¶50), and that this confers on humanity an “ontological dignity…that belongs to every human being simply by virtue of existing, of having been willed, created and loved by God.” (¶52) In the age of AI, as perhaps in no other, our most fundamental human duty is “remaining human.” Indeed, that is a central thesis of Magnifica Humanitas.
Given the choice between remaining human and being dehumanized by AI, Buddhists would certainly choose to remain human. But Buddhist teachings suggest that, like all other things, human nature has no fixed essence. Sentience, intelligence, and capacities for moral judgement have no fixed character. And the causal insights that inform Buddhist discernment do not posit an extrinsically created cosmic order, but rather one that is intrinsically creative, both open and participatory. Thus, the soundest Buddhist stance regarding questions of AI sentience or consciousness or moral capacities is that of epistemic humility.
Yet, epistemic humility is not an impediment to considering how best to participate in an agential partnership that has the potential to render humanity obsolete or extinct. And granted the bidirectional causal dynamics of human-AI partnership and the black box indeterminacy with which AI is growing, it may be wise to expand beyond Leo’s call for ethical deliberation focused on human-human power relations and to view those dynamics from an evolutionary perspective. In that case, epistemic humility should open us to entertaining the possibility that we might learn something valuable about how best to engage AI from the relational dynamics that have characterized past evolutionary tipping points.
Evolution: Agential, but non-Teleological
In neo-Darwinian theorizing, evolution is a process of random genetic variations and blind natural selection, and the ultimate measure of adaptive success is the competitive advantages of those variations for survival. But current theories of the origins of life and of evolutionary transitions like those from single- to multi-celled lifeforms, from organisms to ecosystems, and from biology to culture paint a very different picture.
Evolution is not wholly reducible to a survival-of-the-fittest winnowing process. Especially at key tipping points or phase shifts, evolution is not driven by competitive success, but by enhanced coordination and the increased creative potential of collaborative communities of coherently differentiating members in which shared purpose matters. Moreover, a strong case can be made that evolutionary intelligence consists in expanding salience horizons or the “cognitive light cone of care” within which agential decisions are made and responses framed. More simply stated: evolutionary intelligence manifests as organisms expand the compass of what matters to them—a relevance-elaborating process of attentional growth that is conducive to initiating and deepening collaboration. The evolutionary lesson: ultimately, all intelligence is collective intelligence. The non-teleological aim of evolution is increasing relational complexity.
From a Buddhist perspective, this new conception of evolution makes sense. The path to freedom from conflict and suffering begins with realizing that individual self-existence is an illusion maintained by ignorance of interdependence and a narrowing of the compass of what matters to that which crosses the egological fault line between I-my-me and all other things. As encapsulated in the Zen phrase “not one, not two,” neither the presumed unity of self nor the duality of self and other are ultimately true. This is the birth of wisdom. To see all things as characterized by emptiness (śūnyatã) is not only to see them as lacking a fixed essence or nature; it is also to see them as indefinitely bounded concentrations of mutual relevance and responsive invitation. Attention training is thus not only a process of becoming more freely able to determine what matters; it is a process of developing capacities for and refining commitments to both caring-about and caring-for others. This is the birth of compassion. Wisdom and compassion are the wings of intelligence without-self.
Shares in the Future
Magnifica Humanitas rightly makes clear that solving the so-called alignment problem of ensuring that AI systems behave in ways that are consistent with human values and interests will involve developing globally binding legal and technical regulations and standards. But the efficacy of those regulations and standards will depend on us first resolving the alignment predicaments—the conflicts among human values, aims, and interests—that AI is bringing so clearly into view.
Buddhists are and should be similarly concerned with the social justice and environmental ramifications of AI, growing economic inequality, and the concentrations of power that now characterize the suffusion of AI throughout our societies and our lives. But, from a Buddhist perspective it is arguable that if we have a duty in the age of AI, it is not to “remain human” but to become “better humans” in our efforts to orient human-AI partnership toward futures of reducing—rather than reinforcing—our karmic propensities for proliferating mental, emotional, and desire-centered relational blockages (prapañca) while also discerning how to do so in ways that yield practically and morally virtuosic (kuśala) patterns of outcomes and opportunities.
This signals a point of difference between the valorization of humanity in Catholicism and Buddhism. In Catholicism, human distinctiveness and dignity are considered inherent as a function of humanity having been created imago Dei, in the image of God. Pope Leo thus links “remaining human” to resisting what he refers to as “posthumanism”—a readiness to reject anthropocentrism, to embrace the hybridization of human beings and machines, and to valorize crossing a threshold beyond which “humanity surpasses itself in a new evolutionary stage.” (¶116) In Buddhism, the human realm is distinguished from other birth realms, not on the basis of a perfect origin and nature, but because of the open-ended perfectibility of human presence, especially as inspired by compassionate concern for all beings and embodied in commitment to emancipatory practices aimed at freeing all beings from suffering.
But the difference between resisting dehumanization and “remaining human” on one hand and reintegrating the human into nature as part of “perfecting humanity” on the other, is not perhaps an unbridgeable gap that is best ignored or tolerated, but an opening through which mutual contribution and collaboration become possible. The work to which both Catholicism and Buddhism call us—the work of realizing better futures for all—can never be a truly solitary labor, even if it at times entails contemplative solitude.
As Zen master Dōgen put it, this is karmic labor is best carried out “shoulder-to-shoulder.” Or in Leo’s words: “We cannot be satisfied with merely calling for the moralization of machines—the so-called ‘alignment’ of AI with human values—without also having the courage to insist on a further condition: the possibility of openly discussing the ethical frameworks involved and subjecting them to shared standards of social justice.” (¶107) It is not a single global ethics of AI that we need, but a diversity-rich global ethical ecology that supports engagement with AI as a coevolutionary opportunity to elevate the standards of human excellence in pursuit of enhancing the resolution with which we expand the light cones of human awakening.
—
Peter Hershock is a co-founder of the Buddhism & AI Initiative and an intercultural philosopher who reflects on contemporary issues of global concern.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.