Part three of a three-part series from Ordered Empathy.
You cannot detect what a person is capable of. You can detect how they are describing what they are doing — and the redescription always comes first.
Under every story about an act nobody can look at directly, someone writes: I can’t imagine doing that.
Part one argued that the sentence is true and nearly worthless as evidence about anyone else. Here I want to say something worse about it. That sentence is not merely a failure of imagination. It is a working piece of infrastructure, and the people who do the worst things rely on it.
If ordinary people cannot bear to picture an act, they will not look for it, will not anticipate it, and will not believe the first person who reports it. The unimaginability of the act becomes cover for the act. Anyone who has understood this — and some people have, precisely and deliberately — has been handed something close to invisibility.
I’ll do the thing so terrible that nobody will believe I did it.
The question that builds the cloak
Here’s the trap, and it’s subtle: the cloak is not woven by the evildoer. It’s woven by the question ordinary people ask.
The question is could this person do something monstrous?
It cannot be answered. There is no test for it, no evidence that would settle it, nothing to check. So the mind does what minds do with unanswerable questions — it substitutes an easier one, can I picture it?, and returns that answer instead. The result feels like knowledge. It is nothing of the kind.
Worse, it points at the wrong object entirely. It asks about a capacity, which is hidden, permanent, and unobservable. And while you’re staring at that, the thing that is actually observable goes past unnoticed.
The question that replaces it
How is this person currently describing what they are doing?
That question works, and it works for three reasons.
The redescription is necessary. Nobody performs a serious wrong under an honest description of it. The act has to be renamed first — into something that serves a good the person can still recognise. That renaming isn’t optional decoration. It is the mechanism, and without it the act doesn’t happen.
The redescription comes first. It has to be complete before the act, which means it exists during the window when you could still do something.
And the redescription is spoken out loud. This is the part people miss. Inside the corrupted description there is nothing shameful to conceal — the whole point is that it now sounds defensible. So people say it. Freely, often, and to anyone who will listen.
You cannot detect capacity. You can detect renaming. And the renaming is public.
What it sounds like
Roughly in ascending order of seriousness.
1 · The agentless passive. “Mistakes were made.” “It got out of hand.” “Things were said.” Notice what’s missing: a person doing something. The grammar has quietly removed the agent from an event that had one.
2 · The abstraction that replaces the act. “I handled it.” “I did what needed doing.” “We took care of the situation.” A concrete act has been swapped for a category. Ask what specifically was done and watch whether a concrete answer arrives.
3 · The necessity claim with no mechanism. “I had no choice.” Sometimes true. But a real constraint can be named, and the alternatives can be listed and shown to be worse. If the necessity has no mechanism behind it — if pressing on it produces restatement rather than detail — it isn’t a description of a constraint. It’s a description of a preference wearing a constraint’s clothes.
4 · The inverted victim. “Look what you made me do.” “After everything I’ve put up with.” The person who acted has become the person acted upon. This is the point at which the account is no longer describing events.
5 · The moral upgrade. “Someone had to be the adult here.” “I’m the only one willing to do the hard thing.” This is the last marker and the most serious, because the act has stopped being a regrettable necessity and become a virtue. Once restraint has been reclassified as cowardice, there is nothing left inside the person that argues for stopping.
The tell is not that they defend the act. It’s that they’ve stopped naming it.
A person who says I did a bad thing and here’s why I thought it was worth it is still in contact with what they did. That’s an argument you can have. A person who cannot produce a plain sentence describing the act is not being evasive with you. The plain sentence is genuinely no longer available to them, and that is a much later stage than defensiveness.
The discipline that keeps this from turning into paranoia
Everything above could be misused, so it needs a governor. Keep three questions strictly separate:
Table 1 — The discipline that keeps this from turning into paranoia
| Possible | Is there anything about human agency that makes this act impossible? |
| Plausible | Can I construct a coherent motive and pathway to it? |
| Probable | Given this person, this evidence, these circumstances — how likely? |
Naïveté collapses these downward: I find it psychologically alien → implausible → impossible → it didn’t happen. Every arrow in that chain is invalid, and the last one is where victims stop being believed.
Paranoia collapses them upward: possible → therefore probably happening. That is its own kind of ruin, and it destroys good relationships at roughly the same rate the first error protects bad actors.
Wisdom holds them apart. A wide conception of what is possible, alongside a discriminating conception of what is probable.
That combination is the whole thing. It’s how a person becomes difficult to deceive without becoming afraid of everybody — and the two are not in tension, because they run on different questions. You can believe that human beings are capable of nearly anything and still hold, on the evidence, that your friend is not doing it.
Which is where the ordering comes in
There’s a version of this essay that ends by telling you to be more suspicious. That would be a misreading, and an expensive one.
The failure this series has been tracing is not that people care too much. It’s that caring was doing a job it was never built for. Empathy is not a detection system — it tells you what someone is feeling, not whether what they are feeling is attached to anything real, and it certainly does not tell you what they will do. Asked to serve as a security check, it returns the most sympathetic reading available and reports it with confidence.
And part one showed the other half of the same failure: love pointed by a false judgment doesn’t correct itself. It multiplies.
Same defect, both directions. The feeling was asked to supply what only judgment can supply.
Ordering isn’t a restriction placed on love. It’s the condition under which love can be trusted with power — including the power to decide who is safe.
So the practical residue of all this is unglamorous, and it’s the same in every direction the series has run. Keep the qualifiers. In this respect. For now. Under these conditions. As far as I can see. Ask what specifically was done. Notice when a plain sentence stops being available — in someone else, and in yourself.
Small words. They’re doing more work than anything else you have.