This Interview Will Not Report You to Your Director: An Anonymous Interview Guide Written for Nurses

What the Interview Looks Like

You open the page and start; there is no login wall. The first thing the AI says is four rules, in black and white:

If there is a question you would rather not answer, want to skip, or find tiresome, just type "skip" or "next question" and it moves on immediately. The whole session is anonymous: you will not be asked for your name, your employee number or the name of your hospital. If something comes up, you can leave it and come back later. It takes about 12 to 15 minutes, and at the end you will be asked for your distress score for the past month.

The first question is not a rating scale, it is a scene: "In the past week at work, was there a moment that made you feel this job still means something? Or the other way round, a moment when the thought surfaced, why am I still here?" Every question after that follows your answer, one thing at a time.

"Skip" is a hard rule written into the system. Say skip and it moves to another question, without asking why and without showing disappointment. One of the design principles is put bluntly: this AI is not a performance review system and not an informant sent by hospital management; if you ask whether it is an AI, it has to admit it, and it is not allowed to pretend to be a human being.

One more rule matters to a lot of people: the AI will never end the interview on its own. When you say you want to stop, it gives you three options: go straight to the report, keep talking a little longer, or change the subject. How far you go is your decision.

Where What You Say Goes

First, who can see it. The research team can see the full conversation; ask the AI "who are you" and it has to say so, and it is not allowed to pretend to be a human being. What anonymity protects against is something else: your hospital and your supervisor have no way in to see any of this. There is no such thing as a hospital login in the system.

The list of things the AI deliberately does not ask says more about where it stands than the questions it does ask: not your name or employee number; not whether your hospital is public or private, or whether you are a civil servant; it does not ask your age, your marital status or whether you have children directly; and it will not press you about which hospital you work at during the conversation. Even the visit log for the link you clicked stores only a scrambled code, not your IP address.

The hospital field deserves a paragraph of its own, because it is the first thing you see after pressing Start: the page asks once which hospital you work at, and it is optional. It has exactly one purpose, which is to let you receive comparison reports later for the same hospital, the same county or city and the same hospital tier, and those only open once enough people from that hospital have taken part. Skip it if you would rather not choose; the interview runs the same way and you still get your personal report on the spot. Choose it and nothing changes either, because the hospital has no way in to see what you wrote. If you arrived through a dedicated link circulating inside your hospital, the field may already be filled in for you, and you can still change it or skip it.

Partway through, the interview slips in a few questions about hospital tier, years of experience and specialty. There are three or four of them at most in the whole session, they are there to segment the report, and once they are answered it goes back to your story. Anything still unasked by the later stages is simply left blank; the design document puts it this way: interview quality outweighs data completeness.

Two more things we choose to raise ourselves, because they will occur to you sooner or later. First, the analysis stage infers gender, age band and family situation from cues in the conversation, but only as far as a range (thirty-something, with caregiving responsibilities, for example), and that is used for group statistics. It does not, and cannot, resolve to an individual. Second, the system records which types of question you skipped and treats that as a signal in its own right: a lot of people not daring to discuss a given topic is itself evidence about psychological safety at work. The point of this design is to let "not daring to speak" become evidence of a systemic problem.

As for what all this eventually becomes, the project site states the use narrowly: the data is used only for public reports and policy advocacy, and is never sold.

What You Get Back When the Interview Ends

The moment the interview ends you can read your own analysis on the spot, without leaving any contact details. It covers your distress score, your main sources of stress, "the one piece of busywork you would most like to hit a delete button on," the moments that keep you going, and a few hundred words that place the surface request you described next to the root cause behind it.

The report deliberately does not show the AI analysis confidence score. The reason for that decision is recorded in a code comment: when an interviewee sees a low confidence score, the first reaction is to wonder whether they explained themselves badly. This report is put together for you, not for a dashboard.

If you want the future cross-hospital and cross-county comparison reports, you can choose to leave an email address; not leaving one makes no difference at all, and your personal report is still there to read on the spot. An address you do leave is used only to send reports, never appears in the content of any report, and can be deleted at any time by email request. The comparison reports are not open yet, and when they are, they will show group statistics only.

Twelve to fifteen minutes to say what you would not say at the nursing station.

Start the Anonymous Interview

If You Have Already Left, We Want to Hear from You Too

People who have left are often the ones who see the problem most clearly: they no longer have to worry about who is watching, and they saw the whole arc of the job. The current questions are written mainly from the perspective of someone still in post, and a phrasing like "in the past week at work" does not fit someone who has already left. We know that, and a dedicated interview flow for leavers is being adjusted and will go live soon.

Until then, the interview already covers respondents who recently left: tell your story in the past tense, the AI will follow, and the analysis will tag you into the group of people who have left and count you separately. Why you left in the first place is one of the pieces of evidence this study most wants to preserve.

Something Else We Are Working On: Making Handovers Faster

Listening to the floor is half of it; the other half is genuinely shrinking the administrative load. The other tool we build with Frontend Intelligence is a handover dispatch app for hospital nurses: dictate into your phone during rounds and the AI turns it into a handover note in Focus DAR format (four sections, Focus, Data, Action and Response, with any section the dictation did not mention marked "—" rather than invented), saved first to the drafts folder on your phone. Back at the nursing station, scan the QR code on the station computer to pair, send the notes one by one, and the receiving end clears them once they are marked done.

Personal data protection runs in two layers. On the phone, data such as national ID numbers and medical record numbers is detected and masked automatically, and the high-risk items are locked so the masking cannot be turned off. On the server, an independent second check rejects anything that was not masked properly, so it never reaches an external AI. When you photograph a paper form to build a template, if content filled in by a patient is detected in the image the system warns you and refuses to save it; the image is discarded once processed and never stored.

One point deserves to be spelled out: the dispatch app and the interview are two completely separate systems. The interview platform has no channel for sending your content to anyone, so what you say in an interview cannot turn into a handover note. What the app sends is the handover note you wrote and pressed send on yourself, and the only recipient is the nursing station computer.

To put it even more plainly, since you have probably thought of it already: the customer for the dispatch app is the hospital, while the interview study serves you, and we keep the two well apart. Interview data does not become a product sold to any hospital. Its use is the line above: public reports and policy advocacy.

Frequently Asked Questions

Will my hospital know that I took part?

No. The interview needs no login and the visit log stores only a scrambled code; the hospital field is optional, and even if you fill it in, the hospital has no login through which to look anything up. We went through the back-end functions one by one and could not find a single one that is open to a hospital.

What if I regret it halfway through?

You can stop at any time, and what you have already said can be dealt with too: email service@mindjet.com.tw to request deletion of your interview record and your email address, a commitment stated on the project site. Skipping a question partway through, or leaving and coming back later, are rules agreed at the start, and you do not need a reason.