Theme

How clones are tested

The validation work: clones against real behaviour, against surveys, under mood, culture and social-desirability pressure.

0 insights·7 studies

How do we know a clone's answer is worth anything?

  1. #

    2.5pp

    clone error against real behaviour, in percentage points — The room emptied. The pillars stayed.

    base 235

  2. #

    870

    matched US and UK clones answered 15 identical scenarios — Mad vs. Annoyed — A Cross-Cultural Pragmatic Analysis

    base 870

  3. #

    0.07

    the widest say-do gap found across five socially loaded items (1-5 scale) — Clones Don't Lie to Look Good: Testing for Social-Desirability Bias

    base 546

  4. #

    77pp

    points by which mood alone moved the same answer, more than any demographic — How much does mood change an answer?

    base 488

  5. #

    2.5pp

    clone error against real new-account behaviour, in percentage points — The Say-Do Gap: Predicting What People Do, Not Just What They Say

    base 235

What has Brox published on how clones are tested?Newest first

Other themes

How should you read these findings?

Where the numbers come from. Every figure on this site is a predicted response from Brox clones: behavioural replicas of real people, each built from a long-form interview with that person. The base under each figure is the number of clones behind it.

Where the real data comes from. The long-form video interviews with the real people the clones are built from. A quote is always shown with the interview line it was predicted from.

What you can take away. Insights publish their tabulation as CSV and JSON: question, option, count, percentage and base. Never per-person rows.

Ask this of your own customers: humans@brox.ai