<?xml version="1.0" encoding="UTF-8"?>
<feed xmlns="http://www.w3.org/2005/Atom" xml:lang="en">
<title>Brox Insights — Research methods</title>
<subtitle>How Brox clones are tested: against real behaviour, against surveys, against social-desirability bias, mood and culture. The evidence behind every other page here.</subtitle>
<id>https://reports.brox.ai/industries/research-methods/atom.xml</id>
<link rel="self" type="application/atom+xml" href="https://reports.brox.ai/industries/research-methods/atom.xml"/>
<link rel="alternate" type="text/html" href="https://reports.brox.ai/industries/research-methods"/>
<updated>2026-08-04T00:00:00.000Z</updated>
<generator uri="https://astro.build">Brox Insights (Astro)</generator>
<author><name>Brox Insights</name><uri>https://reports.brox.ai</uri></author>
<entry>
<title>The room emptied. The pillars stayed.</title>
<id>https://reports.brox.ai/studies/the-room-emptied-the-pillars-stayed</id>
<link rel="alternate" type="text/html" href="https://reports.brox.ai/studies/the-room-emptied-the-pillars-stayed"/>
<published>2026-08-04T00:00:00.000Z</published>
<updated>2026-08-04T00:00:00.000Z</updated>
<author><name>Brox.AI</name></author>
<category term="Study"/>
<category term="research-methods"/>
<category term="How clones are tested"/>
<category term="The say-do gap"/>
<summary type="text">Research ethics was written for a bounded encounter between two humans. Synthetic research keeps none of those conditions. Consent, privacy, honesty and harm, pillar by pillar — with Brox.AI evidence: stated intent missed reality by 38.8 points, the clone by 2.5. Headline finding: 2.5pp — clone error against real behaviour, in percentage points (base 235). Full report: https://reports.brox.ai/en/case-studies/the-room-emptied-the-pillars-stayed</summary>
</entry>
<entry>
<title>Mad vs. Annoyed — A Cross-Cultural Pragmatic Analysis</title>
<id>https://reports.brox.ai/studies/mad-vs-annoyed-a-cross-cultural-pragmatic-analysis</id>
<link rel="alternate" type="text/html" href="https://reports.brox.ai/studies/mad-vs-annoyed-a-cross-cultural-pragmatic-analysis"/>
<published>2026-06-25T00:00:00.000Z</published>
<updated>2026-06-25T00:00:00.000Z</updated>
<author><name>Brox.AI</name></author>
<category term="Study"/>
<category term="research-methods"/>
<category term="How clones are tested"/>
<summary type="text">A cross-cultural pragmatic analysis of anger talk in 870 matched US and UK clones across 15 scenarios (26,098 first-person responses, ~754,000 words), and what it reveals about the fidelity of population-scale simulation. Headline finding: 870 — matched US and UK clones answered 15 identical scenarios (base 870). Full report: https://reports.brox.ai/en/case-studies/mad-vs-annoyed-a-cross-cultural-pragmatic-analysis</summary>
</entry>
<entry>
<title>Clones Don&apos;t Lie to Look Good: Testing for Social-Desirability Bias</title>
<id>https://reports.brox.ai/studies/clones-dont-lie-to-look-good</id>
<link rel="alternate" type="text/html" href="https://reports.brox.ai/studies/clones-dont-lie-to-look-good"/>
<published>2026-06-22T00:00:00.000Z</published>
<updated>2026-06-22T00:00:00.000Z</updated>
<author><name>Brox.AI</name></author>
<category term="Study"/>
<category term="research-methods"/>
<category term="How clones are tested"/>
<category term="The say-do gap"/>
<summary type="text">We stress-tested clones for social-desirability bias: 546 clones answered five socially-loaded questions as a survey answer and as their real intent. The say-do gap stayed within ±0.07 — the rosy-survey-answer bias simply didn&apos;t appear. Headline finding: 0.07 — the widest say-do gap found across five socially loaded items (1-5 scale) (base 546). Full report: https://reports.brox.ai/en/case-studies/clones-dont-lie-to-look-good</summary>
</entry>
<entry>
<title>How much does mood change an answer?</title>
<id>https://reports.brox.ai/studies/how-much-does-mood-change-an-answer</id>
<link rel="alternate" type="text/html" href="https://reports.brox.ai/studies/how-much-does-mood-change-an-answer"/>
<published>2026-06-16T00:00:00.000Z</published>
<updated>2026-06-16T00:00:00.000Z</updated>
<author><name>Brox.AI</name></author>
<category term="Study"/>
<category term="research-methods"/>
<category term="How clones are tested"/>
<category term="The say-do gap"/>
<summary type="text">We put 488 synthetic personas through ten different moods and asked them the same questions. Mood alone moved the answer by up to 77 points Headline finding: 77pp — points by which mood alone moved the same answer, more than any demographic (base 488). Full report: https://reports.brox.ai/en/case-studies/how-much-does-mood-change-an-answer</summary>
</entry>
<entry>
<title>The Say-Do Gap: Predicting What People Do, Not Just What They Say</title>
<id>https://reports.brox.ai/studies/the-say-do-gap</id>
<link rel="alternate" type="text/html" href="https://reports.brox.ai/studies/the-say-do-gap"/>
<published>2026-06-09T00:00:00.000Z</published>
<updated>2026-06-09T00:00:00.000Z</updated>
<author><name>Brox.AI</name></author>
<category term="Study"/>
<category term="research-methods"/>
<category term="The say-do gap"/>
<category term="How clones are tested"/>
<summary type="text">We tested four ways to predict consumer behaviour against what actually happened. One matched reality on its home turf — and the honest version, including where it ties or loses, is more useful than… Headline finding: 2.5pp — clone error against real new-account behaviour, in percentage points (base 235). Full report: https://reports.brox.ai/en/case-studies/the-say-do-gap</summary>
</entry>
</feed>
