Three Pigeons in a Row
When I can’t make a decision, sometimes I look for a sign. If I see three pigeons in a row on the next block, fine, that’s a yes. Other times I ask AI. Not to decide for me, but to consult, or just to listen.
I’m a second-year MFA Interaction Design student at SVA, and my thesis looks at why people ask AI to tell their fortune. Underneath it is the question I started with in my first post: how does a person take an output from a system they don’t understand, and turn it into a decision they’re willing to act on?
A question like that can’t stay in my head. So I’ve started saying it out loud, to classmates and to someone who actually uses AI this way, to hear how they read it and where they push back. These are the first three conversations. More are coming, and I’ll add them here.
1. Kathy Choi, a classmate
September 15, in our thesis class. We paired up and explained our theses to each other. I went first.
I told her that when we face uncertainty, we ask everyone: experts, friends, parents. Sometimes there's no right or wrong, just a decision, like why I picked SVA. When I still can't decide, I look for a sign. The pigeons at the top of this post come from that conversation.
I also told her about my younger sister. She buys lottery tickets, and she asks GPT for the numbers. Sometimes she adds: I was born in the year of the dragon, give me seven numbers. Nobody knows the mechanism behind that. But my sister believes it.
Filling in a lottery ticket. Photo by Santeri Viinamäki, CC BY-SA 4.0, via Wikimedia Commons.
Kathy saw a tension in my framing:
"You use the name fortune teller, which sounds very mystical. But it's not relying on supernatural powers. You're still relying on your own intellect, through discussion with the other person, to build that confidence."
Then she asked something I hadn't asked myself:
"How much is too much AI, where it becomes more harmful than good, even if it is just for small decisions? Can that snowball into something they rely on too much?"
About her own use:
"Sometimes I'll be like, am I overthinking this? And usually if I have to type that out, it means yes. We just need a quote-unquote second opinion."
That was one of my hypotheses: people already have the answer in their head, but not enough confidence, so they need a second opinion. And when they hear an answer that goes against it, they ask again. Kathy proved it to me. It isn't only my idea. Maybe it's a collective one.
I'd held that part back because it felt too narrow. She said: "This is exactly where everyone is, so keep going with it."
2. A friend who uses AI for fortune-telling
September 22, in the evening, one on one. Not a formal interview yet. I can't do real research with people until the school's research ethics board (the IRB) approves my study.
I explained the idea to her. We go to AI to help us decide, without knowing how it arrives at its answer. That is a lot like fortune-telling. The I Ching, Zi Wei Dou Shu, tarot: there seems to be a mechanism, but it isn't one plus one equals two. AI has one too, just too complex for us to understand. Both are black boxes. So why do we let something we don't understand help us decide? And often we don't really want them to decide for us. We want them to confirm something we already feel inside.
The 64 hexagrams of the I Ching in the Fuxi sequence, from James Legge's 1882 translation. Public domain, via Wikimedia Commons.
She said:
"I've always felt Claude is more like a humanities student, and ChatGPT thinks like a science student."
Which is true. It's why I'm trying to reach philosophers at Anthropic, to ask how they design the system prompt.
She'll be away from October 4 to early December, and offered to screenshot her AI fortune-telling sessions in the meantime, as long as nothing is too private. When we meet properly, I want to watch how she actually talks to Claude and ChatGPT, typing or by voice.
3. Yuxuan Chen, a classmate
September 24, in Thesis Lab, a class on telling stories with data. We traded data sketches. Mine used survey data on how often Americans consult astrology, tarot, and fortune-tellers.
Yuxuan studies impulse buying on social media. Her sketch showed me how many calculations sit behind a single purchase decision.
She asked me directly: "Have you ever believed in it?"
I said I believe it when I'm especially lost. I've paid for tarot, a natal chart, a Bazi reading, maybe once every three to five years. I don't remember whether any of it was accurate. But every time I went, I was in some huge fog.
Horoscope of Prince Iskandar, grandson of Timur, cast for the moment of his birth. Wellcome Collection, CC BY 4.0, via Wikimedia Commons.
Then there was my own data. Reading all the categories together, I got an answer to "who is most likely to go to a fortune-teller," and it matched my own stereotypes almost exactly. I didn't dare make that chart. It felt discriminatory, so I compared one trait at a time instead. Yuxuan said: "That's your design choice."
Visitors filling out a questionnaire at the YOU youth fair, Berlin, 2008. Photo by Raimond Spekking, CC BY-SA 4.0, via Wikimedia Commons.
What I took from it: you have to make choices, or you get no conclusion. The table does hint at where to find interviewees, since people who use AI for fortune-telling mostly like fortune-telling already. But when I tried to finish that sentence, I stopped. Too stereotyped.
Patterns across the conversations
The same thing came up three times, from three directions. Kathy said we usually know the answer already and just want a second opinion. I told my friend that we often want AI, or a fortune-teller, to confirm something we already feel. And with Yuxuan, I admitted that the few times I paid for a reading, I was lost, and I don't even remember whether it was accurate. The answer itself seems to matter less than having something outside us say it.
The conversation that changed me most was with Yuxuan. My own data handed me a stereotype about who goes to fortune-tellers, and I had to decide not to use it. I'll face the same choice again when I pick who to interview.
My central question hasn't moved, but it's getting narrower. Less about why people believe, more about what these systems say to them, and who wrote the instructions behind it: the system prompt.
Still to come
Next: someone who builds AI fortune-telling products, and someone working on AI coaching.