When a chatbot becomes an oracle: the difference between an answer and authority

The argument published today

Anthropologist Webb Keane and AI researcher Ambuj Tewari argue in a September 29 essay that the history of oracles can illuminate people’s relationship with conversational AI. Their central concern is authority: an answer may seem to come from beyond an ordinary person’s knowledge, while the process producing it remains difficult to understand. They explicitly stop short of claiming that most people think AI is divine. Source: theconversation.com

That is an interpretive argument, not a new poll measuring worship of machines. It opens a serious question for religious readers without requiring a sensational conclusion. When does consulting a useful resource become yielding judgment to it? What makes the answer persuasive: evidence, expertise, reassurance, mystery, or simply its arrival at a vulnerable moment?

Why the historical comparison needs boundaries

The authors discuss Delphi, where a priestess was understood to convey Apollo’s responses, and other traditions of seeking guidance. These practices belonged to particular religious settings. A contemporary chat interface does not inherit those institutions simply because the word oracle offers a vivid comparison.

Diana Espírito Santo’s entry in the Open Encyclopedia of Anthropology provides a broader safeguard against oversimplification. Divination takes varied forms, can involve trained interpretation as well as inspiration, and is embedded in social relationships. Reducing every tradition to random answers would miss much of what practitioners and anthropologists describe. Source: anthroencyclopedia.com

The useful comparison is therefore limited. People may seek answers beyond their own resources in both settings, but the sources of authority, rituals and beliefs can differ profoundly. A Christian critique should be precise about the practice it is discussing. It should not use other religions as interchangeable scenery for an argument about technology.

Likewise, calling an output oracular is not evidence that a developer intended to create a religion. A naming rationale requires a company statement or another reliable record. A historical resemblance can be worth exploring while the motive behind a product remains unknown.

Three different kinds of trust

Readers can distinguish factual trust, practical reliance and spiritual authority. Factual trust concerns whether an answer is correct: for example, whether a quoted passage actually appears in the source. Practical reliance concerns whether a recommendation is suitable for a particular circumstance. Spiritual authority concerns whose account of truth and obligation a person allows to shape their life.

These can come apart. A chatbot could correctly locate a verse while giving an unhelpful interpretation of a family conflict. A confident summary could leave out the very fact that makes someone’s situation different. A person might appreciate the useful part without giving the system the right to settle the larger question.

Consider a hypothetical student asking whether they should abandon a course to pursue a calling. A chatbot could help list options or organize questions for a mentor. It cannot make the resulting decision responsible merely by producing a moving account of the student’s destiny. The appropriate role of the tool depends on the task, the evidence and the human relationships involved.

This distinction also protects ordinary learning. Consulting a digital reference is not automatically an act of religious devotion. Treating every use as idolatry would make the criticism too broad to be helpful. The more specific concern is an answer being allowed to override evidence, conscience and accountable counsel without adequate examination.

Christian perspective: examine the answer without surrendering judgment

Acts 17:11 describes the Bereans examining Scripture as they considered Paul’s message. The passage concerns a particular encounter with apostolic preaching, not a modern product evaluation. Its relevance here is the discipline of examination: an impressive speaker does not make careful attention to the source unnecessary. Acts 17:11 (NIV)

Applied to a chatbot, that discipline means opening the referenced passage, reading its context and checking whether the proposed interpretation follows. It also means noticing when a source is absent. A warm tone and detailed explanation can invite confidence, but neither supplies the missing evidence.

James 1:5 directs those lacking wisdom to God. In Christian practice, that invitation belongs with prayerful dependence rather than a promise that every convenient answer carries divine endorsement. A believer can use a tool to organize information without treating its response as God’s instruction. James 1:5 (NIV)

The pastoral response should make room for the need underneath the question. Someone seeking repeated reassurance may need patient conversation, belonging or practical help. Simply mocking their reliance on software is unlikely to supply those things. A church can explain the limits of a tool while offering a more trustworthy place to bring difficult questions.

A practical test for families and classrooms

Take one ordinary AI answer and divide its contents into facts, interpretations and recommendations. Check a factual statement against an original source. Ask what assumptions the interpretation requires. Then identify who would bear the consequences of following the recommendation. This exercise is our proposed teaching application, not an intervention tested by the scholars’ essay.

The exercise becomes especially useful when an answer sounds personally revealing. Ask which information the user supplied and whether the response could plausibly fit many people. The aim is not to deny every helpful encounter. It is to prevent emotional force from substituting for reasons.

Keane and Tewari offer a way of asking about authority, not a license to declare that a new religion has arrived. For AI Faith Monitor, the strongest follow-up is equally concrete: document how particular people use these systems, distinguish their own explanations from outside interpretations, and test claims without turning uncertainty into either ridicule or reverence.