Sushasan Labs Policy · Innovation · Impact
Interview protocol

The questions asked in every interview, and the terms on which people take part.

Interviews are held against this published protocol, with fixed questions, a fixed format and stated attribution options, so that they can be read together.

Version 1.0, . Changes are versioned, and each published interview will state the version used. हिंदी में पढ़ें

Why this is published

The protocol is published so that a reader can see what was asked and what was not, another researcher can repeat it, and a participant knows what they are agreeing to before the recorder starts.

The nine standing questions

Each interview asks these questions, in this order, about one programme. Follow-up questions are added as needed.

  1. 1 What reaches you from above: in what form, from whom, and on what timetable?
  2. 2 What do you have to do with it that the instruction does not describe?
  3. 3 What are you measured on, and who sees that number?
  4. 4 What do you record that nobody asks for?
  5. 5 Where does the timetable break, and what happens when it does?
  6. 6 What can you not do, that people assume you can?
  7. 7 What would you change first, if the change cost nothing?
  8. 8 Who else in this system would see it differently, and why?
  9. 9 What should I have asked that I have not?

Who is approached

  • Approached: former officials, retired frontline staff, serving staff in states where Sushasan Labs has no professional engagement, and practitioners reached through professional networks.
  • Not approached: anyone in a current working or reporting relationship with Sushasan Labs or its people, anyone met through a live professional engagement, and anyone working on a programme on which Sushasan Labs or its people are engaged.

Before you agree to take part

  • Taking part is voluntary. Interviews can be held in Hindi or English.
  • You are told who is interviewing you before the interview.
  • Recording is optional.
  • If you are in government service, you may wish to check your service rules before agreeing.

“I am speaking to you for Sushasan Labs. Before we start, four things.

One. I am writing about how this programme works in practice. What is published will be a written, edited interview on a public website, free to read.

Two. You can be named, or described by role and state only, or not identified at all. Tell me which. You can change your mind until publication.

Three. If I want to quote you directly, I will send you the exact quotation and the description of your role before publication. You can correct it, withdraw it, or ask for a different attribution. I will not send you the whole text.

Four. Anything you say is not for publication stays out entirely, including as paraphrase or background. You can say so at any point, including afterwards.

May I record this, so that I quote you accurately? Recording is optional. The recording is not published. It is transcribed on a device controlled by Sushasan Labs, not uploaded to any outside service, and deleted once the interview is published.”

Why the script says where the recording is processed

Where a person’s voice recording is sent is a fact about that person, and it cannot be undone afterwards. Participants are told for that reason.

How an interview is published

Length
1,500–2,500 words, edited for length and clarity.
Attribution
Named, by role and state, or unattributed, as the participant chooses. Stated at the head of the interview.
Annotation
Substantive claims are footnoted against the document that says what should happen.
Quotation approval
Direct quotations and the description of the participant’s role are sent to the participant before publication. The full text is not.
Numbering
Interviews are numbered in one series, so that they can be cited together.
Withdrawal
Up to publication, without giving a reason.

What an interview is evidence of

Interviews are good evidence of mechanism: how something works, where it breaks, what is done with an instruction. They are weak evidence of magnitude and are not used to support numerical claims. The number of interviews is stated, and findings from them are not generalised to a state.