L2.03.5placeholder leftoverdesignresearch

Placeholder content in templates is kept verbatim; empty slots must be unskippable

Aliases: dummy text kept · leftover example values · skippable empty slot

What it is

The template says “object: e.g. Acme Inc.” and “deadline: e.g. this Friday.” Someone in a hurry changes the one item they notice and hits send. The prompt the model receives still contains Acme and this Friday — the designer’s fake data, not the user’s. Placeholder leftover is sample text submitted as a true value. The matching design demand is that empty slots cannot be walked through in silence.

This is not expression range being shaved, and not the frame being uneditable. It is pollution at the value layer: fake content lives on into generation.

Why it happens

Placeholder copy has two faces. Grey “example” text is a hint when unfocused; once it sits inside editable text, it is an ordinary string in the data. Scanning existing words, people mark slots that “look filled” as done. The more the sample looks real (a real company, a real date), the stronger the completion illusion. Under time pressure, unfocused slots never enter working memory.

Generative models have no channel for “this is fake.” Acme and this Friday constrain the output like any other proper name. The artifact then contains a company the user does not know and a wrong deadline, while the user thinks they only filled the object. Leftover turns the designer’s example into the user’s factual error.

Studying it

Build a template with high-fidelity placeholders (names, dates, numbers in real formats). The task requires every slot to be replaced with the user’s own values. Record which slots leave verbatim, where sample values appear in the artifact, whether the user later recognizes those as not their fills. Independent variables: placeholder as grey hint versus already written into the text, whether empty slots block send, whether sample values are obviously fake (“some company” versus “Acme”). Dependent variables: leftover rate, artifact contamination, moment leftover is noticed (before send / after seeing the result / never).

Use gaze or click traces to confirm: did a leftover slot ever receive focus? Never focused and still sent is skippable, not kept on purpose after looking.

Where it stops holding

If placeholders never enter the submitted string — grey text under the field, stripped on send — the leftover channel closes. In a personal template the user built, a “placeholder” may be last week’s true value; keeping it is reuse, not pollution. Voice templates have no visible grey text; leftover appears as “the system spoke a sample sentence for you,” harder to catch. Experts select-all and delete before filling; leftover mainly hits moderately skilled people in a hurry.

Applying it

  • Do not write sample values into the string that will be submitted. Use hints outside the slot, or true placeholders that vanish on focus and stay empty on blur.
  • Block send on empty slots, and name which ones are empty. Do not use a single red “please complete all fields.”
  • If an example must live inside the field, mark it so it cannot be taken as real ([company name], fullwidth brackets), and run a “still a sample value?” check before send.
  • Check: a high-fidelity placeholder template, people in a hurry, only the one item they must change called out verbally. Inspect the submitted payload for leftover proper names. If present, check whether that name landed in text the user might send onward. Either one, and empty slots are still skippable.

Related

  • Same group: L2.03.1 Templates lower the cost of getting started · L2.03.2 Templates constrain the range of what users will express · L2.03.3 Templates must be editable, not only fillable · L2.03.4 Templates turn open input back into structured blanks, at the cost of openness · L2.03.6 Templates encode the designer’s task assumptions and actively mislead when those do not match real work · L2.03.7 A template that can only be applied whole, with its structure invisible, teaches nothing
  • Nearby: L3.03 Hallucination and the Fact-Checking Burden · L2.04 Complementarity of Parameter Controls and Natural Language · L1.06 Graceful Degradation of AI Failure
  • Search terms: placeholder leftover · dummy text · unskippable empty slot

Cards in the same group

Quick Actions

Share

Share this page

ios_share

https://hci.top/en/handbook/L2.03.5