Critique of Magnifica Humanitas – Part 2
Toward a Living Doctrine for Complex Adaptive AI Systems
Longstanding Academy member Tom Golway is an American technologist, author, and systems theorist known for his work in emerging technologies such as AI, blockchain, and distributed systems. He writes and speaks on innovation, complex systems, and the ethical and societal impacts of technology, with a strong focus on ethics in STEM and responsible, human-centered development. In this extended reflection on Pope Leo’s recent encyclical, the second in a three-part series, Tom explores the implications of AI on society. Read Part 1.
Published June 9, 2026
By Tom Golway
IV. Authority Without Declaration—The Pantheistic Fallacy
There’s a quieter shift that the encyclical doesn’t quite name. It’s not that people suddenly believe AI systems are authoritative in some explicit sense. It’s that they begin to act as if they are.
Recommendations, summaries, synthesized judgments, over time they start to substitute for slower forms of engagement. Not because anyone declares them final, but because they’re available and efficient. That gradual transfer of judgment matters. It doesn’t look dramatic. But it accumulates.
I’ve described this as the pantheistic fallacy, not the transhumanist claim that AI is conscious or divine, but the more prosaic tendency to treat its outputs with the deference one might give an omniscient source. Trusting a recommendation algorithm’s judgment about what is worth knowing. Accepting a predictive system’s assessment without engaging the underlying evidence. Deferring to synthesized conclusions rather than wrestling with primary sources.
The danger here isn’t philosophical. It doesn’t require any ideological commitment to posthumanism, just convenience, institutional pressure, and the path of least resistance. That’s what makes it harder to name and harder to resist than the more dramatic versions of the same error.
The antidote is what I’ve called gain-of-function intelligence, not replacing human capacities but amplifying them in ways that make the human more curious, more discerning, more capable of holding complexity. AI as a cognitive exoskeleton rather than a cognitive prosthetic. That distinction only holds, though, under specific design conditions—and those conditions are not automatic. Leo XIV’s call to “remain profoundly human” is right in its moral register. What it still needs is an account of how.
This is more dangerous than explicit posthumanism precisely because it requires no philosophical commitment. It just requires convenience, institutional pressure, and the path of least resistance.
V. Work and the Erosion of Capacity
The encyclical’s treatment of work is grounded in familiar and important principles, especially around dignity. But it focuses primarily on employment as an economic issue — job loss, fair wages, the right to participate in society.
There’s another dimension that feels just as important, maybe more so over time: work as a site where cognitive and moral capacities are actually exercised. Analytical work, creative work, judgment-intensive work — these aren’t just economic functions. They’re practices. And practices shape people.
If those activities are steadily offloaded, the loss isn’t only external. It’s internal. This is what I’ve called the epistemic atrophy problem. A society that outsources its reasoning to AI systems doesn’t merely risk unemployment — it risks the gradual diminishment of the polymathic, integrative intelligence that makes human beings capable of moral agency in the first place.
The Church has always understood that virtue is cultivated through practice, not declared by fiat. The same is true of the cognitive virtues — discernment, synthesis, analogical reasoning, epistemic humility. Those capacities depend equally on exercise. And AI systems, deployed without structural intention, are quietly eliminating those exercises from the texture of working life.
The gain-of-function framework responds directly to this: AI should be designed and deployed in ways that increase the demand for human cognitive engagement, not decrease it. That’s not Luddism. It’s design intentionality in service of what the encyclical itself says it wants.
A society that outsources its reasoning to AI systems doesn’t merely risk unemployment—it risks the gradual diminishment of the polymathic, integrative intelligence that makes human beings capable of moral agency in the first place.
VI. Data as a Generative Loop
One of the more interesting moves in the encyclical is its application of subsidiarity to data. But the treatment still tends to assume that data is primarily a resource that can be distributed more or less fairly, like land or capital.
In practice, data behaves less like a static resource and more like part of an ongoing loop. It doesn’t just describe reality. It participates in shaping the reality that future data will reflect. Training data shapes model behavior; model behavior shapes interaction patterns; interaction patterns generate new data that reinforces the original distribution. That recursive structure matters enormously.
Once you see it that way, the key question shifts. It’s no longer just about access to data. It’s about who has influence over the loop itself, who shapes the evaluative frameworks that AI systems internalize, not just the conditions under which those systems are later deployed.
What the principle of the universal destination of goods demands, in this context, isn’t merely equitable access to data as a resource. It demands something closer to what I’d call a generative commons doctrine: institutional structures that distribute the power to shape the learning dynamics of AI systems broadly, across local communities, educational institutions, civil society organizations—rather than leaving that power concentrated in the hands of the few actors with the technical capacity and capital to build these systems in the first place.
Participation in governance processes that have already been shaped by opaque training dynamics is participation downstream of the real decisions. True subsidiarity in the age of generative AI means having influence over what AI systems become, not just over how they’re regulated once they’re already here.
In practice, data behaves less like a static resource and more like part of an ongoing loop. It doesn’t just describe reality. It participates in shaping the reality that future data will reflect. That recursive structure matters enormously.
VII. The Missing Vocabulary of Emergence
What I kept returning to, reading through the document, is that it acknowledges something like a conceptual gap but doesn’t quite fill it. There’s a recognition that AI “challenges the categories of Social Doctrine from within.” That’s a remarkable admission, and it’s correct. But acknowledging the inadequacy of existing categories doesn’t supply new ones.
You can see the strain most clearly when the document tries to talk about responsibility. Who is responsible when an AI system produces a harmful outcome through a chain of interactions no single actor designed or intended? The familiar categories — tool, user, maker, regulator — don’t quite distribute that responsibility in a way that makes sense.
What’s missing is a vocabulary for emergence: for how genuinely novel structures and behaviors arise from complex interaction in ways that can’t be reduced to or predicted from the components. Without it, the analysis tends to fall back on familiar distinctions. Those still matter. They just don’t seem sufficient anymore.
At sufficient scale and integration, AI systems aren’t just powerful tools with value-laden designs. They constitute genuinely novel social and epistemic realities — ones in which the Whiteheadian insight that the whole is more than the sum of its parts isn’t a metaphor but a description of actual causal dynamics.
Moral frameworks that don’t engage with emergence will keep arriving late. The question isn’t only whether AI systems respect human dignity in their current configurations. It’s whether the emergent social and epistemic ecologies they’re producing are ones in which human dignity remains a coherent and actionable concept at all.
The question isn’t only whether AI systems respect human dignity in their current configurations. It’s whether the emergent social and epistemic ecologies they’re producing are ones in which human dignity remains a coherent and actionable concept .
Part 3 will be published on June 16. Read Part 1, and more from Tom Golway on his blog.
References and Further Reading
Tolkien, J.R.R. The Return of the King. London: George Allen & Unwin, 1955.
Leo XIV. (2026). Magnifica humanitas: On safeguarding the human person in the time of artificial intelligence. Vatican City: Libreria Editrice Vaticana.
Golway, T. (2025). Epistemology in the Age of AI: Rethinking Knowledge, Polymathy, and Human Cognition. White Paper on the Epistemological Boundaries of Artificial Intelligence and the Future of Human Cognitive Augmentation.
Golway, Tom. (2025). The Pantheistic Fallacy: Why Machines Cannot Become Everything. SSRN.
Golway, Tom. (2025). The cognitive boundary of AI: Why human judgment remains irreplaceable. SSRN. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6154769
Golway, Tom. (2024). Human gain-of-function: How AI expands, rather than replaces, human capability. SSRN. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5554819
Golway, Tom. (forthcoming 2026). Generative Dynamics: The New Science of How Complex Systems Transform, Create, and Transcend.
Golway, Tom. (2026). Toward a Mathematics of Living Systems.
Golway, Tom. (2026). The One Algorithm: What Tolkien’s Ring Tells Us About AI as Moral Amplifier.
Polanyi, Michael. The Tacit Dimension. Garden City, NY: Doubleday, 1966.
Whitehead, A. N. (1929). Process and reality: An essay in cosmology. Macmillan.
Searle, J.R. Minds, Brains, and Programs. Behavioral and Brain Sciences, 3(3), 1980.
Dreyfus, H.L. What Computers Cannot Do: A Critique of Artificial Reason. Harper & Row, 1972.
Dreyfus, H. L., & Dreyfus, S. E. (1986). “Mind over machine: The power of human intuition and expertise in the era of the computer.” Free Press.
Zuboff, Shoshana. The Age of Surveillance Capitalism. PublicAffairs, 2019.
Wilson, E. O. (1998). Consilience: The unity of knowledge. Knopf.
Kuhn, T.S. The Structure of Scientific Revolutions. University of Chicago Press, 1962.Clark, A., & Chalmers, D. (1998). The extended mind. Analysis, 58(1), 7–19.
Kahneman, D. (2011). Thinking, fast and slow. Farrar, Straus, and Giroux.
Polanyi, M. (1966). The tacit dimension. University of Chicago Press. Searle, J. R. (1980). Minds, brains, and programs. Behavioral and Brain Sciences, 3(3), 417–457.
