Syncretic ThinkAloud
Think-aloud sessions that keep going when nobody is in the room.
ThinkAloud plays the part of the moderator in an unmoderated study. It briefs the participant, has them practise thinking aloud, notices when they go quiet, and nudges them back — while capturing screen and microphone to storage that survives a crash.
Think aloud, in three lines
- Say what you are doing while you do it. Not afterwards, not a summary — a running commentary of what you are looking at and what you expect to happen.
- Describe, don't explain. "I'm looking for the price" is data. "I clicked that because I assumed…" is a story told after the fact, and it changes the thinking being measured.
- Silence is the failure mode. People stop talking within a minute of starting a task. A moderator says "keep talking." Everything distinctive here exists to replace that one sentence.
What the participant sees
One decision per screen
Consent, device check, practice, tasks, post-task questions, done. Large type, a single forward action, and a visible way out on every screen.
Step 3 of 6
A meter, not a nag
A live speech meter runs for the whole session. It is the main intervention: unbroken feedback that they are being heard, which prevents most silences before a reminder is ever needed — and it doubles as a mic-failure indicator.
Listening — keep talking when you're ready
Nudges that cannot lead
When silence and input-idleness coincide, ThinkAloud shows one of four fixed, content-free reminders. Twice per silence episode, then it stops. The strings are product constants — a study author cannot put a leading question into the nudge path.
Never lose a session
Media is written to durable storage before it is uploaded, in chunks, from the first second. A refresh, a dead network or a closed tab is recoverable — not a lost participant.
What it does not do
ThinkAloud is a capture instrument.
It does not transcribe, code, score or summarise anything. What it leaves behind is a
predictable object layout, a versioned manifest.json and an event log with millisecond
offsets, so the tools that do analysis can align transcript, task and nudge without reverse-engineering
us.
No webcam recording. Speech detection runs on the participant's own device — no audio is sent anywhere for the purpose of nudging.