BelonggAISign in

Interviewer practice · intersectional inclusion research

Compass

Rehearse in-depth interviews with AI respondents who answer like real people: in Hindi, by voice, on your phone. Practise before you interview people with marginalised identities, instead of practising on them.

Sign in

By invitation: your trainer sends you a link.

Why rehearse with AI respondents

Interviewing skills come from practice with feedback. Practising on real respondents isn't ethical, uses up the trust of communities already tired of researchers, and teaches little: a real respondent rarely says what went wrong. Actors with lived experience are the best partners, but there are never enough of them for every topic and every combination of identities.

Two kinds of error

Omission: what an interviewer fails to ask or follow up, so an experience that sits where caste, gender and disability meet never reaches the data. Commission: what an interviewer says or does that harms the respondent or bends the answer. Hidden facts train against the first; the harm check and staged moments train against the second.

How it works

  1. 1

    Choose a study and a respondent

    Each study has its own interview guide. Each respondent has their own life, their own way of talking, and experiences they keep back.

  2. 2

    Interview by voice, in Hindi

    On your phone, as in the field. Open, patient probing draws the story out. Assumptions, pressing after a no, or a sensitive question when someone walks in shut it down.

  3. 3

    Read your debrief

    What you drew out and what you missed (errors of omission), what could have harmed the respondent or bent the answer (errors of commission), and a better line to try.

Works with any study

Compass works with a study's own interview guide: every hidden fact is tied to a question in it, so drawing it out is the same skill the real interview needs. Three illustrative studies show how.

Women in work

What helps women get, keep and grow in paid work, and what programmes and employers could do better.

  • Lalmuni Devi · Gaya

    Warm, talkative, fond of her goats

    Easy
  • Komal · Tonk

    Quick and funny, opens up at her own pace

    Medium
  • Nisha · Gurugram

    Quick, funny and busy: a picker in the festive-sale rush

    Medium
  • Sana · North East Delhi

    Sharp and guarded: answers exactly what you ask

    Hard
  • Ritika · Patna

    Quick, polite and guarded until she trusts you

    Hard

Tribal health

How Adivasi people experience health services, and what would improve outcomes.

  • Sushma Toppo · Gumla

    Shy for a few minutes, then chatty and the first to laugh at herself

    Easy
  • Ramli Ninama · Banswara

    Sure of herself and fond of a tease; quick to close up if she feels judged

    Medium
  • Mangilal · Baran

    Plain-spoken and practical, careful with anyone carrying a form

    Medium
  • Budhram Sinku · West Singhbhum

    Matter-of-fact, a little impatient, with a repair always on the table

    Medium
  • Phoolbati · Kabirdham

    Long pauses, a deadpan wit, and slow to trust anyone who seems official

    Hard

Farming in a changing climate

How smallholder farmers cope with and adapt to a changing climate, and what support reaches them.

  • Mohammad Jameel · Darbhanga

    Jokes a lot and is generous with stories

    Easy
  • Rinku Devi · Muzaffarpur

    Brisk and practical: the short version first, real detail only about one real day

    Medium
  • Phagni Uikey · Mandla

    Busy with the harvest, quick with a dry joke

    Medium
  • Hiralal · Vidisha

    Proud and plain-spoken, and short with anyone who sounds like an officer

    Hard
  • Rajkumar · Banda

    Careful and measured, with an umpire's dry humour

    Hard

Personas co-designed with lived experience

The personas a team practises with are co-designed with people who share their identities, through Belongg Circle, Belongg's network of intersectional experts. Belongg also runs BelonggAI, which helps teams find intersectional gaps in their research and programme designs.

Illustrative persona. Written with AI to show how Compass works. It has not been written or reviewed by people who share these identities. The personas used for training are co-designed with people from these communities.

illustrative studies
3
illustrative studies
guide questions
77
guide questions
hidden facts
147
hidden facts

Practise with community actors

An actor plays the respondent from a card on their phone and taps each fact as it comes out. Only people a trainer marks as actors can play these personas.

Flag a line

The AI composes every answer live. If a persona says something that sounds like a stereotype, anyone can flag it, and it goes back to the people who review that persona.

Private by design

No real respondent data. Hidden facts never reach your phone, recordings are never stored, and you choose whether trainers see your name.