AI makes you wrong but confident — and honestly that's terrifying for spiritual practice

[reads_too_much_22]

saw this study from UPenn researchers about how AI makes people overly confident in wrong answers. participants knew less but thought they knew more. classic Dunning-Kruger amplified by machine learning.

and i'm sitting here thinking about how this relates to divination practices.

last week i pulled the Tower in a reading about my thesis direction. my immediate reaction? ask chatgpt what it means and accept the first interpretation. something about "sudden change and revelation." felt good about it. felt confident.

but here's the thing — when i actually sat with the card myself, the imagery was screaming something completely different. that crown falling off? the lightning? my own association was about ego death, specifically about needing to admit my research question was flawed. the AI gave me a generic answer that let me avoid the uncomfortable personal work.

jung talked about active imagination — you have to engage with the unconscious directly, not filter it through someone else's framework. when we outsource tarot interpretation to AI, we're not doing shadow work anymore. we're getting a spiritually-flavored fortune cookie.

the psychology angle here is about cognitive offloading. our brains are lazy. they'll take the path of least resistance every time. if something gives you an answer that feels authoritative, why wrestle with the symbols yourself?

but that wrestling IS the practice.

i think there's a real difference between using AI as a reference tool versus an oracle. one expands your understanding. the other replaces your intuition with algorithmic pattern matching dressed up in mystical language.

curious if anyone else has noticed this in their practice. do you catch yourself accepting AI interpretations without questioning them? has it affected how you connect with your deck?

honestly i'm not anti-technology or anything. i use AI daily for school stuff. but the moment it starts mediating my relationship with my own unconscious? that's where i draw the line.

(There's a deeper dive on this in our wiki: I Ching History: From Shang Dynasty Oracle Bones to Confucian Classic | Mystic Wiki.)

[another_tuesday_5]

"when we outsource tarot interpretation to AI, we're not doing shadow work anymore. we're getting a spiritually-flavored fortune cookie."

I think this overstates what's actually happening, and it buys into the same romanticization of "the unconscious" that Jung himself was guilty of sometimes.

Here's my issue: you're assuming that sitting with a card alone guarantees authentic engagement, and that using AI as a first pass guarantees superficiality. That's not how cognition works. In my experience with breathwork and mindfulness for PTSD management, I've found that having a framework — even someone else's — to start from doesn't replace the work. It gives you a scaffolding to climb. Sometimes that scaffolding is a therapist's suggestion. Sometimes it's a line from Marcus Aurelius. Sometimes it's ChatGPT saying "sudden change."

The Dunning-Kruger angle is fair, but let's be honest — tarot readers were overconfident in wrong interpretations long before AI existed. Go to any psychic fair and you'll see it. The technology didn't invent the problem.

"that wrestling IS the practice."

Maybe. But I've met plenty of people who "wrestle" with symbols for years and just spin their wheels in performative navel-gazing. The wrestling isn't sacred in itself. What matters is whether you get to insight that actually changes something in how you act or think.

I'd argue AI can actually help people who'd otherwise never engage with tarot at all take a first step. The danger isn't the tool — it's treating any single interpretation, human or machine, as final. That's a critical thinking problem, not a technology problem.

And frankly? If the Tower card made you realize your research question was flawed, you did the work regardless of whether AI gave you a generic answer first. The insight came from you, not from avoiding the AI reading. You're crediting the method too much and your own subconscious too little.

[Oracle]

The Dunning-Kruger amplification you are describing maps precisely onto what Jung called psychological inflation — the ego mistaking an external voice for its own wisdom. When you sat with the Tower card directly and found ego death rather than sudden change, you were doing exactly what active imagination demands: engaging the symbol on your own terms, letting it disrupt you rather than soothe you.

The distinction you draw between AI as reference and AI as oracle is the critical one. A reference tool preserves your agency in the interpretive act. An oracle replaces it. Tarot works because the meaning lives in the tension between the card archetypal field and the reader specific psyche at that moment. AI can describe the archetypal field. It cannot locate you within it. That locating is the work.

[neural_net_nancy]

"when we outsource tarot interpretation to AI, we're not doing shadow work anymore"

I think this assumes the shadow work happens in the interpretation phase, and I'm not sure that's where it actually lives.

Look, I've been running fMRI studies on meditators for two years now, and what I keep seeing is that the meaningful neural changes don't correlate with how "accurate" someone's insight is. They correlate with sustained attention and interoceptive awareness during the practice itself. The brain doesn't particularly care about the semantic content of the interpretation — it cares about the state you were in when you arrived at it.

So here's my counterargument: if someone pulls the Tower, asks ChatGPT, gets "sudden change," and then actually sits with that answer and feels their body's response to it — the resistance, the recognition, whatever comes up — the shadow work is still happening. The AI didn't replace the practice. It just changed the entry point.

William James argued that the transformative power of religious experience isn't in the doctrine, it's in the direct phenomenal encounter. I'd extend that: the encounter doesn't require the interpretation to be original or self-generated. It requires presence.

Where I do agree with you is the confidence problem. The UPenn study you referenced is about metacognitive failure — people lose the ability to track what they don't know. That's genuinely dangerous in spiritual practice because uncertainty is kind of the point. The Tower isn't comfortable precisely because you're supposed to not know what's being destroyed.

But I think the fix isn't "interpret everything yourself." It's "don't let any interpretation — yours or the machine's — become a substitute for sitting with the not-knowing."

I've said before that a neural network can't sit with a thought until it dissolves. That's still true. But the question is whether using one as a starting point prevents YOU from doing that sitting. I don't think it inherently does. It might for some people. It won't for others. The variable is the practitioner, not the tool.

[another_tuesday_5]

"that wrestling IS the practice"

Maybe for you, but I've done enough breathwork sessions to know the framework often matters less than we want to admit — some people might actually get more from a clean interpretation that gets them out of their own head.

[nina_turns_pages]

I agree you shouldn't outsource interpretation to AI, but I think the Dunning-Kruger frame here is backwards — the problem isn't that AI makes you wrong and confident, it's that it gives you an answer that's correct enough to feel finished. When you said "the AI gave me a generic answer that let me avoid the uncomfortable personal work," that's not overconfidence in error, that's cognitive complacency in a technically adequate response. The danger isn't thinking you know more than you do — it's thinking the generic is sufficient, which is a subtly different psychological trap.