You asked the AI for the risks. It listed the safe ones.

You typed "tell me the risks of co-signing this lease." The chatbot gave you a clean list. Market rent, early-termination fees, credit impact, the cost of one person leaving a joint-and-several agreement.

Every item was correct. Not one of them was the thing keeping you up.

You nodded at the list. You still didn't sign. Because the real risk never made it into the prompt.

The dilemma

Here's a composite of the kind we see constantly in this category — stitched together from many readings, no real person.

Maya has been with Dev eight months. The lease on his flat is up; her roommate is moving out; the math says sharing one place is cheaper than two. He floated it lightly over dinner — "we basically live there anyway" — and she heard herself say "makes sense."

Then she couldn't sleep. So she did the modern thing: she asked a chatbot to list the downsides of a joint lease. It produced a competent memo. Twelve months of shared liability. Her credit tied to his late-paying habit. The fee if either of them breaks early. Furniture nobody wants to split. She read it twice and felt no clearer, because the memo answered a question she hadn't actually asked.

What she couldn't type was this: things with Dev have been flat for two months. The easy texting stopped. The silences at dinner aren't comfortable, they're careful. And somewhere under the spreadsheet, she suspects she wants the lease precisely because it would be a reason to stay — a twelve-month wall against a conversation she's dreading. The lease isn't the risk. The lease is the hiding place.

She asked a tool to inventory the logistics. The logistics were never the problem. The problem was that signing would let her postpone finding out whether the relationship is ending, by making leaving expensive.

The read

A checklist can't find a risk you didn't hand it. A reading works the gap between what you typed and what you steered around. Four lenses bite here.

Motive: the real engine is the lease as an excuse

The motive lens asks what's actually driving the move, under the stated reason. The stated reason is "it's cheaper." The engine is "a signed lease means I don't have to decide about Dev this year." Those are different animals. One is a budgeting decision. The other is using a legal contract to outsource a choice she's afraid to make. The lens pulls hard toward naming it: Maya isn't weighing a lease, she's buying twelve months of not-knowing and calling it rent.

Honesty: the question she asked was built to be safe

The honesty lens checks whether the framing is telling the truth. "List the risks" sounds like courage. It's often the opposite — a way to feel thorough while keeping the one dangerous item out of the room. You can ask for the risks all night and never type the sentence that would make a stranger worried for you. The lens pulls toward the omission: the most honest input isn't the twelve logistics, it's "I think I want this so I won't have to leave."

Non-harm: a contract is a bad way to hold a person

The non-harm lens asks who gets hurt, and how. Binding Dev to twelve months as a hedge against her own doubt isn't neutral. If the relationship is fading, the lease doesn't save it — it traps two people in a slow fade with a cancellation fee attached. The harm lands on both of them, later, with interest. The lens pulls toward not using a legal instrument as emotional insurance.

Discernment: separate the money question from the person question

Here's where two lenses openly disagree. Discernment — the capacity to tell what's really being decided — notices that there are two decisions wearing one coat. There's a clean logistics question (is a joint lease smart for two committed people?) and there's a buried relationship question (are we committed?). The checklist answered the first. Motive insists the second is the only one that matters. Discernment doesn't pick a side so much as refuse to let the money question keep standing in for the person question. You cannot answer "should we share a lease" until you've answered "are we staying together" — and that's the one she's steering around.

The verdict: don't sign the lease to answer a question about Dev. Have the Dev conversation first; let the lease be a consequence of the answer, not a substitute for it.

A general AI chatbot is built to be helpful about the question you hand it — so it diligently inventories the risks you were brave enough to name and quietly skips the one you weren't. That's agreeableness, a documented behavior in these systems, not a glitch. KarmaLens runs the other direction: it asks for the facts that would actually change the call, then commits to one reading of what's really being decided and won't talk that call back down because you flinched. It sets the options beside the verdict with the cost each one carries — sign now, wait and talk, split the difference — and quotes scripture only where a verse lands, verbatim, never improvised to flatter you. The point isn't a longer list. The point is that the call gets made once, after the thing under the list gets named.

The takeaway you can run tonight

Free, no account. Call it the embarrassment test.

1. Write the risks you'd happily read aloud to a friend. (For Maya: fees, credit, furniture.)
2. Now finish this sentence, privately, in plain words: "The risk I'd be embarrassed to type is ___."
3. Read what you wrote. Notice that it's a sentence about a person or a fear, almost never a line item.
4. Ask: does the decision I'm leaning toward make that embarrassing risk easier to face or easier to avoid?

If the choice makes the real risk easier to avoid, it isn't solving your problem — it's hiding it, the way a smooth status page hides the outage underneath. The embarrassing sentence is the one worth a reading. The googleable list is just the part you were already willing to look at.

A chatbot can only inventory what you typed. The move that changes your decision is typing the thing you were steering around.

Maya's twelve-item list wasn't deliberation. It was avoidance wearing thinking as a costume. If you've got a list that keeps growing while the one real question stays off the page, write the dilemma to KarmaLens — and put the embarrassing sentence first.

You can always see how a committed reading differs from an agreeable one first. What's the risk you've been careful not to type?