Can AI Read Between the Lines?
Can it uncover connections in texts that humans might miss?
Imagine that a language model could process an extensive corpus of religious texts, together with their translations, narratives, and centuries of accumulated debate.
It could compare passages that no individual scholar would have the time to examine all at once, identifying relationships that are difficult to discern from any single perspective. It could trace the recurrence of a symbol across different historical periods, compare how a metaphor changes from one translation to another, or identify a tension between passages that have traditionally been studied separately.
These connections would not necessarily constitute new discoveries, but they could offer valuable starting points for further investigation.
The question I would like to explore in this article is not whether God exists, nor whether artificial intelligence believes, but what it can understand and interpret about sacred texts.
We tend to assume that interpreting them requires something more than knowing their contents. A religious authority belongs to a tradition, participates in its practices, and inherits its history; their relationship with the text is never exclusively intellectual.
Religion makes this asymmetry particularly complex because, for many people, faith is an integral part of the text’s very meaning. For a believer, a sacred text is not merely a document to be studied from a distance, but a living, everyday relationship with the divine.
This relationship cannot be reduced to the information contained in the text alone. A single passage can be read as a historical document, a moral teaching, a liturgical expression, or a revelation. Its actual meaning depends on the interpretive framework through which it is approached.
Lived faith vs. computational analysis
Understanding seems to require lived experience, whereas interpretation can exist without it. A historian does not need to have lived in ancient Rome to interpret its texts, just as a translator does not need to share a religious faith to translate a sacred text.
One form of knowledge arises from within, from inhabiting a shared tradition; another emerges from the outside, from the connections between texts, translations, and the history accumulated around them. These are two distinct and non-interchangeable paths, and the difference between them becomes crucial when we reconsider the very concept of interpretation.
If interpretation takes place entirely on this second level — that of historical and textual relationships — AI presents itself as a capable computational exegete. Its ability to process symbols and reconstruct semantic relationships, however, does not resolve the question of understanding.
A system can analyze what a symbol means within a tradition without this demonstrating that it experiences that meaning. The distance between analyzed meaning and lived meaning thus remains the point at which the question of intelligence meets that of consciousness.
It is precisely from this apparent contradiction that a third path of interpretation emerges: what if lived experience were not the only way to access meaning?
Get Stefania Marin’s stories in your inbox
Join Medium for free to get updates from this writer.
By mapping vast amounts of data and analyzing how human beings have described, translated, and debated the sacred over the centuries, AI does not participate in faith as a lived experience, but can reconstruct aspects of how human beings have understood and expressed it.
Rather than replacing the knowledge that arises from a lived tradition, computational analysis maps its invisible architecture. What emerges is a new representation of the patterns through which human beings have organized, transmitted, and interpreted their relationship with the sacred.
This perspective confronts us with an unavoidable provocation: if, one day, an LLM were to uncover a previously unrecognized connection between sacred texts, a connection that could shed new light on a long-standing theological contradiction, how would we react?
Would we dismiss the discovery simply because it had emerged from an algorithm without a soul or faith?
Such a rejection would, on closer examination, be entirely plausible. In human beings, after all, experience, faith, interpretation, and authority are often inseparable. This possibility, however, is not without its limitations.
The bias of the algorithm
The fact that AI can compare different traditions without adhering to any of them does not make it impartial. The neutrality of an algorithm cannot be taken for granted: the distribution of sources, the available translations, and the criteria used to select them can privilege some perspectives over others. Moreover, identifying a connection does not automatically establish its relevance. Two passages may share a word, an image, or a narrative structure without expressing the same idea.
The point, then, is not to replace the scholar or the believer, but to use AI to do something that has until now been extremely difficult to achieve on a large scale: map the differences.
By bringing together textual relationships that are difficult to perceive at a glance, an algorithm does not resolve theological conflicts, but can expose the assumptions from which they arise. It is up to us to examine and verify them in context, within the intricacies of history and tradition.
Here, however, we encounter a fundamental distinction. An interpretation can be linguistically flawless and historically plausible without being spiritually authoritative in any way. These are different domains, and confusing them means attributing a legitimacy to computation that it simply does not possess.
The real challenge is not to decide whether an algorithm truly “understands” sacred texts, but to understand what we ourselves mean by understanding. If we reduce it to a series of digital patterns, we risk losing the meaning a text acquires within a tradition; if we tie it exclusively to personal experience, we give up an extraordinarily powerful analytical lens.
Where does the value of this knowledge lie? It all depends on what we are looking for when we open a sacred text.
AI Disclosure
AI was used solely to support conceptual brainstorming, translation, and structural formatting within a strictly author-directed process.
This is Part 3 of a five-part series exploring the boundaries of mind and organization. If you missed the previous chapters, you can read:
