Use cases
Researchers: fieldwork, reading, and the ethics question
Academic writing is mostly the writing around the writing: reading notes, analytic memos, field notes. Dictation suits all three, and on-device recognition avoids a problem with your ethics approval.
Last updated
The ethics point, first
Research involving human participants runs under an approved protocol that specifies how data is handled and who processes it. If you dictate a field note about a participant into a cloud transcription service, that service is processing participant data.
Almost no protocol names a dictation vendor, because almost no researcher thought of it as data processing. It is, and an ethics committee would say so. On-device recognition means the audio is not transmitted, so there is no additional processor to declare — which is a great deal easier than an amendment.
Where dictation fits academic work
| Task | When | Why speaking wins |
|---|---|---|
| Field notes | In the moment, or immediately after | Detail decays. This is the classic case |
| Reading notes | At the end of a section | Stopping to type breaks the reading |
| Analytic memos | Whenever the thought arrives | Memos are thinking out loud in written form. Literally the same act |
| Coding notes in qualitative analysis | During coding | The reasoning behind a code, which is what makes it defensible |
| First drafts of sections | Once the argument is clear | Fast, loose, restructured after |
| Peer review comments | On request | The explanatory part, which is what makes a review useful |
| Grant narrative sections | Under deadline | Long prose, always late |
| Exact quotations and citations | Never | Type these. A misheard quotation is a correction |
The analytic memo is the strongest case
Memo-writing is a formal method in qualitative research, and it works precisely because it is thinking made explicit while the thinking is happening. The obstacle is that opening a document and typing four hundred words interrupts the thought that prompted it.
Ninety seconds of speaking does not. What comes out is messy and that is fine — a memo is not a publication. The value is that it exists and is dated, which is also what makes an audit trail of analytic decisions possible.
Fieldwork and connectivity
Fieldwork happens where there is no signal, frequently by design. Cloud dictation is unusable there. On-device recognition needs only a one-off model download before you travel — see setting up offline dictation.
The same applies to archives, which are often basements with no reception and a prohibition on photography but not on a laptop.
Setting it up
- Teach it your field's vocabulary Theoretical terms, method names, the authors you cite constantly, institutions, place names. Academic vocabulary is precisely what general models have not seen.
- Add participant pseudonyms, not real names If your protocol uses pseudonyms, teach it those. It also keeps real names out of the vocabulary file, which is itself a small piece of data hygiene.
- Make a memo snippet Date, what prompted this, what I think, what would change my mind. Then dictate into each.
- Set retention short and turn on disk encryption The local history contains participant data. Treat it as such: FileVault on a Mac, BitLocker on Windows.
- Never dictate a quotation Type or paste every direct quotation and every citation. A plausible misrecognition inside quotation marks is a correction notice.
Limits worth knowing
- Dictation does not transcribe interviews. It takes live microphone input only. VV Pro can transcribe a recorded interview from a watched folder or Vaitly Meetings (beta), on your computer — and interview transcription carries separate consent obligations whichever tool does it.
- Quotations and citations must be typed. Non-negotiable.
- Shared offices. Dictating a field note about a participant in a shared postgraduate office is a confidentiality problem regardless of the software.
- It will not fix a thin literature note. Speaking makes it faster to write down what you thought; it does not supply the thought.
Questions
Does a dictation app need to be in my ethics approval?
If it transmits audio to a vendor, it is a processor and arguably does. If recognition runs on your own machine and nothing is transmitted, there is no additional processor. We are not your ethics committee, but that is the distinction they will draw.
Can it transcribe my interview recordings?
Dictation does not, but VV Pro can: drop a recording into a watched folder or import it into Vaitly Meetings (beta) and it is transcribed on your computer, with the audio not uploaded. Asking for a summary sends the transcript's text to our server. Either way, check your protocol and consent forms cover it, because that is processing participant data.
Will it understand my field's terminology?
Not at first. Specialist vocabulary is rare in general English. Teaching it twenty terms takes fifteen minutes and covers most of a project.
Does it work in Zotero, Obsidian, NVivo and Word?
It types into any text field, so free-text notes work everywhere. Structured fields and coding interfaces are usually better handled with the keyboard.