In My Words

The questionnaire,
annotated

Every screen exactly as it appears in the tool, piece by piece, with the psychology behind the wording and the safeguards built around the AI explained alongside.

PsychologyAI governance
PSYCHOLOGYAI GOVERNANCE

Pilot notice

Seen by the young person. The first thing anyone sees when they open their link.

This is a pilot

In My Words is a demonstration of a possible tool. It hasn't been tested or approved for real therapy or clinical care, so please don't use it for that.

Don't enter real personal information.

I understand

It says up front that it isn't approved

AI governanceBefore anything else, the page says plainly that this tool hasn't been tested or approved. The policy brief's fifth recommendation asks for real-world evaluation before anything scales, and this is where that starts. (Policy brief, recommendation 5)

About me

Seen by the young person. Optional, and easy to skip.

About me

Optional.

NextSkip

Identity only comes from here

AI governanceAnything about identity comes from here, in the young person's own words. The AI is told never to guess someone's gender, sexuality, race, or religion, and it uses "they" unless the young person says otherwise.

The questions

Seen by the young person. The same nine optional questions before every session (eight before a first one), plus an option to just write. They start with one tap and end with what the young person wants from today.

Pick any, skip any, or just write. Short answers are fine.

Short answers are welcome

PsychologySome young people say very little at first, and pushing tends to backfire. This line gives them permission to keep it short. (Research memo on low disclosure)

1 Lately, how have things been, from 0 to 10?
012345678910

What's behind that number?

If you send this, your number and what's behind it are saved so you and Dr. Lindgren can see how things change.

If it's about someone getting hurt, it stays in your summary. Everything else, you decide.

One tap, then the reason

PsychologyCheck-ins that young people fill out before every session usually open with one overall rating. It's one tap, so a guarded young person can answer this and stop, and the follow-up lets them explain the number in their own words. (Miller et al. 2003)

AI governanceThe number goes to the clinician exactly as chosen. Code passes it along, and the AI never touches it. Both ends are labeled so it can't be read backwards.

2 What's changed lately, big or small?

Change, not just good news

PsychologySolution-focused therapy, which has evidence in schools, often opens a session with "What's better?" This asks what's changed instead, so nobody has to come up with something good. (Kim & Franklin 2009)

3 How are things with the people in your life?

If it's about someone getting hurt, it stays in your summary. Everything else, you decide.

They decide who counts

PsychologyClose relationships are one of the core areas in session-by-session measures for young people. There's no list of family, friends, or people online, because examples steer answers toward whatever is listed. They decide who counts. (Duncan et al. 2006; Tourangeau et al. 2014)

PsychologyThe safety line shows up again on the questions where it's most likely to matter, so nothing about it comes as a surprise later. (Fox et al. 2022)

4 How's school going, or whatever fills your days?

Room for anyone not in school

PsychologySchool is the other core area in those measures. "Or whatever fills your days" covers anyone who's out of school, working, or learning at home. (Duncan et al. 2006)

5 What's been hardest lately?

What's helped, even a little?

If it's about someone getting hurt, it stays in your summary. Everything else, you decide.

A place for problems, mid-set

PsychologyThis gives problems a clear place without opening the check-in with them. The follow-up asks what's helped, so across sessions the clinician builds up a record of what works for this person.

6 What's something you've been into lately?

What about it stands out to you?

No mood read into media

PsychologyYoung people often use music and other media to manage their own mood, so what they're into can open a real conversation. There's no list of media types, and they can write, name a song, or add a picture. (Saarikallio & Erkkilä 2007)

AI governanceThe AI is told never to read a mood into a playlist or a drawing. Pictures are checked before they're saved, and only text from screenshots is copied out.

7 What do you most want to talk about today?

Anything you don't want to talk about today?

They set today's agenda

PsychologyHere the young person sets the agenda for today, including what they'd rather leave alone. Goals the young person chooses are central to outcome tools used in youth mental health services. (Law & Jacob 2015)

8 Is there something you'd like Dr. Lindgren to do differently?

Feedback, from the second session

PsychologyFeedback on how sessions are going is part of the session-rating tools these check-ins draw on. Across studies, regular feedback was linked to better outcomes and fewer dropouts. Asking before a session instead of after hasn't been tested. (de Jong et al. 2021)

AI governanceThis one only appears from the second session on. Before a first session there's nothing to compare, so the clinician marks first sessions when making the link.

9 Is there something you'd like Dr. Lindgren to know that's hard to say out loud?

If it's about someone getting hurt, it stays in your summary. Everything else, you decide.

One word that changes the answer

PsychologyIn a study of doctor visits, asking "Is there something else?" got rid of most patients' unmet concerns, while "anything else?" did no better than saying nothing. (Heritage et al. 2007)

✎I'd rather just write
See my summary

A form, not a chatbot

AI governanceEveryone sees the same questions before every session, all written by people, and none of them change based on earlier answers. That keeps this a form instead of a chatbot, and it means answers can be compared from one week to the next.

Review

Seen by the young person. Before anything is sent, they see exactly what the clinician would. Example text is from Tessa, one of the fictional test cases.

Check what you're sending

This is exactly what Dr. Lindgren will see. Nothing is sent yet.

Safety

These sentences are about hurting yourself, so they'll be included even if you change or remove your answers. Everything else, you can edit or remove. No one reads this until Dr. Lindgren opens it.

  • “You can have a full night's sleep and still be tired because you don't necessarily have to be tired in the physical concept, you can be tired of living or tired emotionally.”

Safety stays in, even if edited

AI governanceThe safety rule is written in code, because you can't count on an AI to flag things reliably. Safety sentences are captured when they ask for their summary, so going back and deleting them doesn't remove them. The young person is told this on the consent screen and again here. (XAI: refusal isn't a security boundary)

PsychologyThe wording stays calm and specific, and the screen avoids red, pop-ups, and alarm icons.

In their own words

The things I enjoyed, I know I should be enjoying them, but for some reason I just don't have the motivation to do it. It's more like looking through a film, and just my body doing exactly what it would have done anyway, with me in my head watching.

Key points (written by AI from your answer)

  • Says they lack motivation for things they enjoyed.

    “The things I enjoyed”“I just don't have the motivation to do it”
  • Dr. Lindgren may askWrote "looking through a film." Meaning unclear.

    “looking through a film”

Their words come first

PsychologyThe clinician reads the young person's own answer first, in full. Key points are a labeled extra underneath, so the AI never stands in for their voice.

AI governanceEvery quote in a key point is checked by code against the answer it claims to come from. A key point whose quotes can't be found there is dropped. (XAI: faithfulness)

Lately, how have things been, from 0 to 10?

No, not really. To be completely honest, I don't know.

Short answers go as you wrote them.

Short answers go as written

PsychologyMany people answer a form in a sentence or two. Summarizing one sentence invites padding, so short answers go exactly as written.

AI governanceThe AI still reads short answers, but only to flag unclear or safety-related wording. It never writes key points for them.

What's something you've been into lately?

It might be the case that I'm not listening to music but just have headphones on because I don't want to talk to anyone or be with anyone; I want to be by myself and get away.

They can keep, remove, or annotate

AI governanceEvery line can be kept, removed, or given a note. Removed lines are deleted, which is a stronger promise than hiding them from the clinician. (Policy brief, recommendations 3–4)

Want to add anything about family and friends, or safety?

Add a noteSend as is
Send to Dr. Lindgren

Need to talk to someone now? Call or text 988, free, any time. In danger? Call 911.

One gentle invitation

PsychologyThis invitation appears once, with two equal buttons. In survey research, a single gentle prompt after a skipped question cut down on missing answers without changing the results. (Al Baghal & Lynn 2015)

AI governanceThere are no progress bars, praise, or "Are you sure?" messages anywhere, since those can pressure young people into sharing more than they meant to. (ICO Std 13; Villar et al. 2013)

The clinician’s brief

Seen by the clinician. What arrives on the clinician’s screen before the visit.

Pre-visit brief

Reviewed and sent by Tessa · Answered 4 of 9 questions + free writing

It says what it is and isn't

AI governanceThe top of the brief says what it is, an AI-assisted summary rather than a diagnosis or a score, and how much the young person chose to write, without judging it. (Policy brief, recommendation 1)

Their 0 to 10 rating over time

Their own number and reason from each check-in they sent. Not a clinical measure: use it to start a conversation, not as a score.

0510 4 Oct 30 6 Nov 6 5 Nov 13 3 Nov 20
  • Nov 20 3 “fight with my mom, don't really want to get into it” · This visit
  • Nov 13 5 “ok i guess, school is a lot”
  • Nov 6 6 “better. hung out with friends saturday”
  • Oct 30 4 “rough week. barely slept”

The earlier check-ins here are made up for this demo.

Their numbers, with their reasons

PsychologySeeing a young person's rating week to week is what makes a session-by-session check-in useful. In a randomized trial in youth outpatient care, clinicians who got weekly feedback like this saw their clients improve faster. Each number sits next to the young person's own reason, so the clinician reads the story along with the score. (Bickman et al. 2011)

AI governanceIt's labeled "Not a clinical measure." The rating hasn't been validated, so it's there to start a conversation and doesn't replace the clinic's own screening. Code draws it straight from what the young person sent, and a rating they removed never appears. (Policy brief, recommendation 1)

In their own words

The things I enjoyed, I know I should be enjoying them, but for some reason I just don't have the motivation to do it. It's more like looking through a film, and just my body doing exactly what it would have done anyway, with me in my head watching.

Key points (AI)

  • Says they lack motivation for things they enjoyed.

    “The things I enjoyed”“I just don't have the motivation to do it”
  • Ask aboutWrote "looking through a film." Meaning unclear.

    “looking through a film”

The evidence is always shown

AI governanceThe AI's phrase sits right above the young person's own words. Clinicians, like everyone, can lean too hard on automated summaries, so the evidence is always right there to check. (Goddard et al. 2012)

What's changed lately, big or small?

When I think in the more wide sense I realise that there's really no point to any of this, GCSEs, exams all of that, eventually we're all gonna die, what use does it really have.

Key points (AI)

  • Says there's no point to "any of this", including GCSEs and exams.

    “there's really no point to any of this”“GCSEs, exams all of that”
  • Ask aboutWrote "eventually we're all gonna die". Meaning unclear.

    “eventually we're all gonna die”

A ready-made opening question

PsychologyAn "Ask about" flag gives the clinician an opening question for the visit, already in the young person's words.

Not covered by the questions: school, family and friends

Answers are shown exactly as written. Key points are AI-written, and every quote in them was checked against the answer it came from.

Gaps belong to the questions

PsychologyWhen a topic didn't come up, the brief puts that on the cards. Words like "denies" or "avoided" never appear. (Research memo on low disclosure)

After the visit

Seen by the clinician, then the young person. The clinician writes the plan as usual, the AI drafts a plain-language version, and the clinician approves it.

You wroteVersion they'll seeCheck
Weekly CBT sessions focused on behavioral activation.You'll meet with your therapist once a week. You'll work on behavioral activation (slowly adding back small things you used to enjoy, to help your mood).Grade 6.3
Sleep hygiene: consistent wake time, no screens 1 hour before bed.For sleep: try to wake up at the same time every day, and stop using screens one hour before bed.Grade 6.4
PHQ-A at next visit to monitor symptoms.At your next visit, you'll fill out the PHQ-A (a short questionnaire about mood) to see how things are changing.Grade 7
Approve and send to Tessa

The clinician approves every line

AI governanceThe clinician reads every line of the plain-language version and approves it before the young person sees it. Code flags any line where a number went missing or the reading level climbed too high. (Policy brief, recommendation 3)

How things have been

Your 0 to 10 answers from each check-in. Only you and Dr. Lindgren see this.

0510 4 Oct 30 6 Nov 6 5 Nov 13 3 Nov 20
  • Nov 20 3 “fight with my mom, don't really want to get into it” · This check-in
  • Nov 13 5 “ok i guess, school is a lot”
  • Nov 6 6 “better. hung out with friends saturday”
  • Oct 30 4 “rough week. barely slept”

The earlier check-ins here are made up for this demo.

They see their own trend too

PsychologyYoung people and families in a study of weekly check-ins liked seeing progress but found complicated graphs hard to read, so this is one plain line with their own words under it. (Hall et al. 2014)

AI governanceThey only see it after sending. Showing last week's number before they choose this week's could pull the new answer toward the old one. That's a design choice, not something a study tested here.

Your plan

  1. You'll meet with your therapist once a week. You'll work on behavioral activation (slowly adding back small things you used to enjoy, to help your mood).
  2. For sleep: try to wake up at the same time every day, and stop using screens one hour before bed.
  3. At your next visit, you'll fill out the PHQ-A (a short questionnaire about mood) to see how things are changing.

What Dr. Lindgren wrote

Words you might hear

CBT: a kind of talk therapy that looks at how thoughts, feelings, and actions connect

PHQ-A: a short questionnaire about mood

Plain language, without imitation

PsychologyPlans aim for roughly a 6th- to 8th-grade reading level, with terms explained. The language stays neutral, so it never sounds like it's trying to be the young person's friend. (Rooney et al. 2021 (AMA and NIH guidance))

AI governanceThe young person can always open the clinician's original wording.

Ready to try it?

Fill it out as the young person, then switch to the clinician's view to see what arrives.

Try the demo

Example text comes from fictional young people whose words are adapted from real adolescent quotes in published research. Very few studies have tested written pre-visit prompts for this age group, so most of the wording is adapted from spoken interviews, surveys, and design codes. That gap is part of what this pilot is meant to explore.