1. A 2017 "cafeteria lunch = she's into me" bug report is still unbeatable nine years later
In September 2017, a post called "A Confession Guideline by a Science Grad Student" appeared on Hatena Anonymous Diary, a Japanese site where people post anonymous essays. [1] The writer tried to treat things a woman in his lab did, like eating a meal together, walking home together, and asking him for advice, as green lights for confessing his feelings.
The problem is that what he actually observed and the meaning his head tacked on afterward got mixed together.
- Observation: We went to the campus cafeteria together.
- Interpretation 1: She doesn't mind being alone with me.
- Interpretation 2: I'm treated as someone special.
- Interpretation 3: She's romantically interested in me.
- Conclusion: It's fine to confess.
Only the first line is real data. The rest was generated inside his head.
On top of that, this was the same lab. With an ordinary friend, you can keep your distance after a failed confession. In a lab, you see each other the next day, and the seminars, experiments, conferences, advisor and shared work all keep going. A study of 942 graduate students in STEM fields found that the person expressing interest tends to underestimate how uncomfortable it is for the other person to turn down an unwanted advance, and how much it can affect their work afterward. [3]
So the stake here isn't just "the odds that the confession works." The existing relationship in the lab is on the table as collateral.
And the biggest reason the post is still fun to read is not the main text but the updates. Every time he got a reaction, he rewrote his own model a little.
First report: Maybe she was giving me signals. Under investigation: Well, a cafeteria lunch and walking home are pretty weak. Root cause analysis: Come to think of it, guys eat together in pairs all the time even when they're not close. Final report: Most likely I got the distance wrong. I'll leave this up as an example of failure.
It's less a romance flame war and more a public incident-response log.
2. Your feelings for someone are not a sensor. They're the gas pedal
The core of this post can be explained by a fairly ordinary cognitive bias.
In two studies of opposite-sex friendships, people with stronger romantic or sexual desire for the other person tended to overestimate how much the other person wanted them back. [2] The studies cover a limited range of people and eras, so this can't be generalized as is to every romance and every gender. Still, the way your own desire leaks into your estimate of someone else's feelings fits this poster's behavior very well.
So the practical rule is simple.
Your own attraction is a good reason to try getting closer. It is not evidence that the other person feels the same.
To put it more crudely, your libido is a gas pedal, not a sensor.
"She laughs a lot," "she asks me for advice," "we go to meals together." Romance can explain all of that. But so can friendship, politeness, being coworkers, and ordinary lab life. This isn't data where only the romance hypothesis wins.
If you then start building a "more accurate list of signs she likes you," you end up right back at the 2017 guideline. What you need is not better mind-reading. It's to guess less.
3. If "I like you! Go out with me!" is your first move, romance becomes an all-in leveraged bet
The old-school one-shot confession crams too many decisions into a single sentence.
"Are you romantically interested in me?" "Do you want to go on a date, just the two of us?" "Do you want to get closer?" "Do we put the label 'couple' on this?"
Skip all of that and go straight to:
I like you! Go out with me! ↓ Um, I never thought of it that way...
and the fall is steep when it fails. The person who turned you down may also start wondering, "Was he reading all his kindness as romance this whole time?" or "If I talk to him like before, will I be giving him hope again?" In other words, a confession can create problems that didn't exist before it. [3]
The safer approach isn't to change the relationship in one go. It's to check only the next small step.
- "Want to go out just the two of us, as a date this time?"
- Each of you shows through actions whether you want to see the other again.
- When you bring up romantic topics, see whether the other person joins in on that.
- For physical closeness, check with the other person at each stage.
- Make the final "Shall we go out?" less like a big announcement and more like putting a name on a relationship that's already moving.
This is not a playbook that guarantees success if you follow the order. It's a design that makes it harder for a failure to blow up the existing relationship in one hit.
In investing terms: don't put your friendship, your existing asset, up as collateral and bet everything on one reading of "she likes me."
4. What worked on the poster: not "you're wrong," but running a unit test on his own model
The interesting thing in the updates is that the first direct rebuttals didn't fully change his mind. Even when told "a cafeteria lunch or walking home is not a sign she likes you," he kept defending himself for a while, saying "reason and emotion are different" and "so I'd like her to avoid eating alone with me." [1]
Then a question came in that hit hard.
"If you went to the cafeteria with a guy, would that make him your best friend?"
After that, the poster started answering by himself, "No, guys do that all the time." [1]
This isn't a lecture. It's a controlled experiment using a worldview he had already adopted.
Old model: Eating together alone is special, so it's a romantic sign. Control condition: Eating together alone with a male friend. Observation: Not special at all. Result: Old model FAIL.
In research on updating beliefs, refutation-style texts work better than simply correcting a misconception. They name the common mistake, refute it, and replace it with the right explanation. A meta-analysis pooling 44 comparisons and 3,869 people reported a moderate positive effect. [4]
Also, people sometimes change their attitudes by coming up with the reasons for being persuaded themselves. [6] In a 2024 experiment, people who were prompted to think counterfactually ("what if the conditions were different?") were more likely to update their beliefs when they met contradicting evidence. [7]
What matters, though, is not stopping after you've poked a hole with a counterexample. Reviews of misinformation correction point out how important it is to offer an alternative explanation to take the old one's place. [5]
For the poster, it would be:
Old: Eating together alone, so romance. Demolition: Even guys eat together alone as a normal thing. New: Eating together alone may just be the result of closeness or circumstances. Whether it's romance, check separately, in a small way.
Once he gets that far, the model swap is complete.
This doesn't mean that piling on makes people smarter. The original page is also full of personal attacks and abuse. [1] Insults are not a method of changing minds. What probably worked were concrete counterexamples he could recalculate with his own assumptions, and the alternative explanation he was able to build from them.
5. Six steps for "unit testing" when you want to change how someone thinks
A unit test is a small automated check that confirms one piece of logic gives the result you expect. For real-life use, this order is quite manageable.
- Separate observation from interpretation. "We ate together" is an observation. "She likes me" is an interpretation.
- Ask for their model. "Why do you think that's evidence she likes you?"
- Bring in a comparison case they also accept. "What if a same-sex friend did the same thing?"
- Let them make the prediction. Don't instantly announce the right answer.
- Build an alternative model. "Maybe it could be explained by closeness, politeness, or being coworkers, not romance."
- Make the next action reversible. Instead of a one-shot confession, get the next small agreement.
Research on motivational interviewing, a counseling approach that supports behavior change, also shows that open questions and reflective listening are more strongly linked to the person talking about change than confrontation, warnings or one-sided advice. But having people rehearse their reasons for not wanting to change for a long time can push things the other way, so it's no all-purpose conversation technique. [8]
In short, persuasion is not a contest to see who can read out the right answer the loudest. It's the work of looking together at where the person's own model misses its predictions.
6. So could this method have stopped Suguru Geto's rampage?
From here on, there are major spoilers for Jujutsu Kaisen.
Suguru Geto is not simply someone who "was taught a bad ideology and fell into darkness." The official synopsis of anime episode 29 states that after Satoru Gojo became "the strongest" and began taking missions alone, Geto also had more time on his own, and that during a summer with frequent cursed spirits he did nothing but exorcise and absorb them. [11]
On top of that, things pile up: the death of Riko Amanai, the people who rejoiced at her death, the split between him and Gojo in strength and role, the pain of absorbing cursed spirits, his conversation with Yuki Tsukumo, the death of Yu Haibara, and the incident at the village where girls with sorcerer talent were being abused.
Director Shota Goshozono also explained in an official interview that he read Geto's fall not as "Geto's problem alone" but as a collapse that happened when the stress built up in the jujutsu world piled onto Geto's own. [12] Yuichi Nakamura, who voices Gojo, said that Geto took Amanai's death harder than Gojo did, and that Gojo failed to notice it. [13]
So if what Geto needed was just
"Killing every non-sorcerer is logically wrong."
that would be far too weak.
At this stage his ideology is no longer a mere hypothesis. It has started to fuse into:
- a story that explains his own pain
- an explanation for why his comrades die
- a sense of what he lives for
- anger at the jujutsu world
- a new place to belong for someone who is isolated
The "needs, narratives, networks" three-part model from radicalization research doesn't diagnose Geto, but it's an interesting lens for reading fiction. [9] When a story that explains the world, and a group that backs it up, connect to someone who has lost their sense of meaning and significance, the ideology becomes stronger than a mere opinion.
Research on moral injury is also a useful metaphor for Geto's environment. It describes how repeatedly living through morally unbearable situations, and feeling betrayed by your organization, can bring anger, disgust, helplessness, cynicism and isolation. [10] This isn't a diagnosis either. The point is that fixing only the person's thinking while leaving the workplace as is won't work.
7. To stop Geto, skip the "debate squad" and split the roles among Gojo, Yaga and Tsukumo
The easiest time to stop him is not the moment he carries out the mass killing at the village. It's before that: after Amanai's death, between his isolation, the talk with Tsukumo, and Haibara's death.
Gojo: don't outsource the moral call to Geto, ask about Geto's own state
Gojo became overwhelmingly strong, but his Limitless technique doesn't extend to reading his best friend's inner world. [13]
What was needed was not outsourcing a good-or-evil ruling with "Should we kill these guys?" but
"Here's what I'm feeling right now. What are you feeling, Suguru?"
and asking about Geto's own state.
Yaga: change the working conditions before the ideological coaching
Cut down solo missions. Pair him up. Let him rest. Review the decisions behind Haibara's mission. Stop endlessly pushing "exorcise it and swallow it" onto a student whose technique, Cursed Spirit Manipulation, is such a heavy burden.
Fix the thinking, leave the work schedule alone. That's the worst kind of improvement program.
Tsukumo: before answering as a researcher, ask "Is this a thought experiment, or are you asking how to actually do it?"
Tsukumo makes sense as a researcher. Her framing of the problem is to look for a root-cause cure, where cursed spirits never get born, instead of treating symptoms by exorcising them forever.
So when Geto brought up an extreme solution, continuing the conversation as a theoretical possibility is understandable. But the person across from her was an exhausted, isolated Special Grade sorcerer, the top rank, who also had the power to carry it out.
What was needed here was a safety check:
"Are you asking this as a research thought experiment? Or are you hurting badly enough that you actually want to do it?"
And then separating the goal from the method:
"Is it really non-sorcerers you want to get rid of? Or is it this system in which sorcerers keep dying?"
Takahiro Sakurai, who voices Geto, interprets in an official interview that, after taking in several things including Tsukumo's words, Geto ended up deciding not merely on "resolve" but on his "way of living." [14]
That's why making Tsukumo the "culprit who pushed him into darkness" is sloppy too. It's closer to an accident in which someone nearly broken was handed a structural analysis of the world and picked the worst possible way to implement it.
As a researcher, top marks. As a crisis responder: "Don't leave it there."
8. What makes Modulo interesting is that it genuinely reconsiders "change the system" instead of "get rid of people"
Break down Geto's problem and at its core is anger: "Why do only sorcerers keep dying to clean up the curses born from non-sorcerers?"
An alternative can be built from here.
Old plan: The cause of cursed spirits is non-sorcerers, so get rid of non-sorcerers. Alternative: The problem is how leaked cursed energy turns into cursed spirits, so change the jujutsu system, not the people.
Volume 3 of Jujutsu Kaisen Modulo (sequel series to Jujutsu Kaisen), which collects chapters 22 to 23, came out in May 2026. [15] As far as secondary sources can confirm, around this point the story seriously considers a plan to remove cursed energy from Japanese people and others to stop cursed spirits from appearing. It also deals with the fact that removing cursed energy alone doesn't make the cursed spirit problem disappear entirely, and with the ethical weight of changing other people's bodies and possibilities without their consent. [16]
Here it's important to double-check the premise.
"Remove cursed energy and everything is happy" is not the answer.
Even so, compared with Geto's mass killing plan, the subject of debate shifts from "deleting humans" to "how do we change the way the world works." That change of direction is a big deal in itself.
If you were to offer a different path to the Geto of 2007, this would be a strong one:
"You don't have to throw out your problem statement. Just change the solution. Study root-cause treatment with Tsukumo."
You can keep all of that energy: wanting to protect your comrades, wanting to end the endless marathon, wanting to go after the cause instead of treating symptoms. Only the massacre gets cut away.
Gojo stops the accidents in the field. Yaga fixes the organization. Tsukumo and Geto study the jujutsu system itself.
Not "delete all the users," but "debug the cursed energy layer."
9. Conclusion: in love or in jujutsu, don't run DROP TABLE on the live system
When you leap from "we went to the cafeteria together" to "she likes me,"
and when you leap from "curses are born from non-sorcerers" to "so just get rid of non-sorcerers," the scale is wildly different, but the bug is the same.
You mistake an explanation you built in your head for a spec sheet delivered by the world.
(DROP TABLE is the command that wipes out a whole table of data in a database. Running it on a live system is the classic way to destroy everything at once.)
The countermeasures look like this.
- Separate observation from interpretation.
- Don't count your own desire as evidence.
- Test the model with a counterexample the person already accepts.
- Let the person find the contradiction themselves.
- When you break the old model, put an alternative in its place.
- Keep the next action as small and reversible as you can.
- When pain or isolation is severe, change the environment and ensure safety before trying to win the argument.
In romance, instead of creating a relationship through a confession, build the relationship bit by bit and put a name on it at the end.
In changing minds, instead of "you're wrong," run the unit test: "Would your theory predict the same thing under different conditions?"
And if someone is already as close to breaking as Geto was, let them rest before you start the study group.
The distance from a cafeteria lunch to wiping out humanity is enormous. But the last lesson is the same.
Don't go all in. Check in small steps. When the model breaks, update it. When a person is about to break, protect the person before the model.
