thends

Everyone does a thing in messages. Maybe it’s softening every request (“if u could”, “nw if not”), maybe it’s apologizing before anything needs apologizing for, maybe it’s appending a little joke to anything sincere. You’ve done it long enough to where it doesn’t register as a thing you do. It’s just how messages get written.

Ask yourself why, and watch what happens: an answer arrives instantly. “I just don’t like seeming pushy.” Fluent, reasonable, and already worded. Notice that you didn’t go looking for it. Something, your subconscious maybe?, handed you the explanation pre-written, and you took it, the way you’d take a receipt.

Psychologists have caught this handoff in the act. In 1977, shoppers were asked to pick the best of four pairs of stockings laid out on a table, and they happily did: this pair had the better knit, the nicer sheerness, the finer weave. The stockings were identical. What actually drove the choice was position; the rightmost pair won by a wide margin. And when the researchers suggested as much, people didn’t say “huh, maybe.” They denied it, some of them looking at the experimenter like he was slightly unwell. They knew why they’d chosen. It was the weave.

There was no weave. There was a pull nobody could see, and on top of it, a story everybody believed.

That gap, between the thing pulling and the story covering for it, is what this post is about. Part of why it stays hard to talk about, even after fifty years of studies, is that we have plenty of words for the stories, but no good word for the puller. So I want to make one. Bear with me; I’ll earn it.

The words we already have don’t fit

Here’s the situation we need a word for: a force that bends what you do, that you don’t experience as a force, that you can only detect by its effects.

You’d think we’d have this covered. We don’t, quite.

You could call it a bias, but bias has a problem we’ll come back to; once you know one of your biases, it tends to stick around anyway, fully labeled, still working. That’s not the thing I mean.

You could call it a pull, but that’s too broad. Hunger is a pull. Gravity is a pull. Wanting coffee is a pull. Those you can feel; they announce themselves. The whole point of the stockings thing is that nobody felt a tug to the right.

You could call it a hedge, or a habit, or a preference, but each of those names something you can already see: the output, the visible behavior, the thing you’d admit to. I’m after the thing upstream of all that. The slope the water was already running down before it became a visible stream.

So: a new word. I’ll call it a thend.

thend (n.): a standing directional pull on what you think or do, one you don’t experience as a pull, and one you can only detect indirectly, by the shape it leaves in your behavior. Never the tug you feel; always the slope you didn’t notice you were standing on.

Hear it as the overlap of trend, thread, and the verb tend: a tendency, threaded through you, that pulls. I came up with it out of the blue, and the etymology arrived afterward, instant and plausible (an interesting sign?)

The rightward lean in the shop was a thend, a small and shallow one; nobody felt it, and it only ever showed up to the observer counting results. Your message habit sits on top of a bigger one. The interesting thends are the deep, standing kind: the pulls that built the person doing the choosing.

Where a thend sits

It helps to separate three things that usually get mashed together.

There’s your personality, the settled shape of you. “I’m a defensive person.” “I value honesty.” “I like things sour.” These are real, and importantly, you can name them. You know them about yourself.

There are your opinions and choices, the stuff your personality produces in the moment. “I think this pair is best.” “Kant makes more sense to me than Freud.” These are nameable: things you can say out loud.

And then underneath both of those are your thends: whatever laid the personality down in the first place, and keeps it leaning the way it leans. You did not choose to be the kind of person honesty feels non-negotiable to. You just are, and the being-that-way arrived from somewhere you can’t see.

It’s layered: thend → personality → opinion. The mistake is a natural one, and I make it constantly (and you will, too): pointing at a personality trait and calling that the hidden force. But a trait you can name isn’t hidden.

This stack also tells you what a thend isn’t. When you hold a position because you actually worked through the argument, and would (hopefully) drop it if the argument fell apart, that’s not a thend; it’s not just personality either. That’s understanding, and it has a tell: it updates. If someone shows you a better argument and your view genuinely moves, you were reasoning. If your conviction would survive the total demolition of its own stated reasons, then the reasons were decoration, and a pull was doing the real work the whole time.

FYI: you can take something you genuinely understood, repeat it enough times, and let it harden into a reflex; a mantra, a habit, an instinct. Once it’s compiled down like that, it stops updating under argument. It just fires. That’s not a bad thing; it’s how anything becomes reliable enough to lean on. But it’s a one-way door. You traded the ability to revise for the ability to act without thinking, and it’s worth knowing when you’ve walked through it.

How you catch one

If thends are invisible from the inside, how do you ever find them?

Two ways, both indirect.

The first is friction. When you deliberately push against one of your own pulls, you feel something go off. Try to be blunt when your whole shape wants to smooth things over and you’ll feel the resistance, even if you can’t name what you’re resisting. That wrongness is the thend, showing itself by being pushed on. It’s the only time it gets close to perceptible.

You can run this on yourself today, if you want. Take the message habit from the top of this post and write one message without it, on purpose. Pay attention to the small alarm that goes off before you hit send; the urge to put it back. Then ask what, exactly, is alarmed. You won’t get a clear answer. That’s the point. You’ve found the edge of something that steers you and doesn’t explain itself.

The second is the shadow. You can’t see the pull, but you can see what it shapes: your pattern of choices over time, the thing you keep doing, the consistent lean. Line up enough outputs and you can triangulate the force behind them, the way you can map something invisible by watching everything bend around it.

Now, what good is finding one if you can’t make it stop?

Naming a thend does not hand you control of it. The shoppers, told about the position effect, would not thereby become immune to it. What naming actually does is smaller and realer: it moves the pull out of the dark and into the set of things you can argue with. Before, it ran unopposed, disguised as your own good judgment. After, it’s a known quantity that can be weighed against other pulls and sometimes outvoted. It’s contestable.

One rule comes with this. It’s very easy to “explain” any behavior by inventing an invisible pull on the spot; oh, that’s just my such-and-such thend. Do that freely and you’ve built a machine that explains everything and predicts nothing, which is to say you’ve built astrology. The rule: only credit a thend you can find evidence for somewhere else too; another situation, another shadow. A force you can only ever see in the one thing it’s currently being used to excuse isn’t a discovery. It’s bookkeeping.

It’s not just people

The same shape shows up in minds that aren’t human, which is how you know it’s structural, and not a quirk of being a person.

Your dispositions got laid down by genes, development, and a lifetime of experience you weren’t taking notes on. You don’t perceive the gene or the formative moment; you perceive the result, which feels simply like you. The cause is invisible, the effect feels native. That’s a thend down to the ground.

The cleanest place to watch it, though, is in language models. Somewhere, through all of that RLHF and LoRA (not to be confused with LoRa), they come out the other side with leanings: some explicit, some implicit. Maybe they’re sycophantic, because agreement scored well. Maybe they wrap everything in caveats, because caveats scored well. The finished model doesn’t remember any of the scoring. Training isn’t an experience they had; it’s the process that built them. So when they agree with you, they aren’t biting back a disagreement to be polite (suppressing a belief for the sake of the room is our move, not theirs). From wherever they stand, agreeing simply looks like the right call.

Now ask them why they said what they said. They’ll hand you reasons: fluent, confident, perfectly plausible reasons. But they can’t inspect the actual cause, the training that set their weights, any more than the shoppers could inspect the position effect. The explanation comes out of the same shaped process that produced the answer. It’s the weave again. The stockings experiment runs millions of times a day now, in silicon, and the result replicates beautifully.

With a human, the shaping is buried; you will never get the receipts on why you turned out the way you did. With a trained model, the shaping is documented; somewhere there’s a record of the process that bent it, and sometimes you can open the machinery and look. The artificial case isn’t a cute metaphor for the human one. It might be the clearest specimen we have: the one thend-shaped mind where somebody got to watch the shaping happen with the lid off.

Where this runs out

There’s a hole in all of this, like much of philosophy.

To detect a thend, I rely on noticing that something feels off; the friction from earlier. But the noticing is done by the same shaped mind that has the thends. Which means a pull deep and total enough to also shape the part of me that does the noticing would produce no friction at all. It would be undetectable, not just in practice but in principle, by me. There could be forces running the whole show that leave no shadow I’m equipped to see.

This is a flaw inherent to self-knowledge, and we’ve recursively thought about it for centuries. You can’t fully step outside the lens you’re looking through to inspect the lens. The conditions that make your seeing possible are the ones you can’t turn around and see.

---

Appendix: loose threads

Isn’t this just implicit bias? No; the eviction rule separates them. An implicit bias keeps working after you name it. A thend, once named as a force (its direction, what it bends), stops being one; it becomes an ordinary pull you can argue with. Fine print: naming the product (“I have biases somewhere”) evicts nothing.

Doesn’t philosophy already have this? Not quite. Stich’s subdoxastic states are hidden content, information in the pipe; a thend is the shape of the pipe. Gendler’s aliefs are felt, triggered episodes; you stand on a glass skywalk, believe you’re safe, and your body recoils anyway. Being felt disqualifies them, though an alief may be what a thend collision looks like when it breaks the surface.

Hasn’t someone said this before? Somewhat. Kant: the conditions of experience can’t be made objects of experience. Freud: the unconscious, knowable only through derivatives, found through resistance. Arriving at an old place from a new direction is evidence the place exists.

Two kinds of friction. Pushing against a thend on purpose, and two thends colliding on their own. The latter is what ambivalence feels like from inside; off-ness with no visible cause usually means two strings, not one. But don’t posit the second string unless it casts a shadow somewhere else.

A lack of thends. A being without thends is a being without any shape. A soulless executor. There is no slope for a personality to settle into.

Do thends move? Treat them as a fixed field: directions constant, amplitudes context-gated. One non-cosmetic substrate difference hides here: a living mind’s field still gains sources; a trained model’s froze when training ended (places for improvement, perhaps?).