--- schema: 1 type: video title: >- Persona Engineering: A Field Guide to AI Synthetic Personas — Ishan Anand, InsightSciences.ai description: >- A team ran about a thousand people through a market research survey, then had LLM agents replay the same questions, and the agents matched the humans closely while carrying less noise than the humans did themselves. That is exactly the tell: a synthetic respondent smooths over the messiness that makes a real population real. Ishan Anand walks through where it breaks. Nudge one variable in the prompt template and purchase probability swings, because the model infers latent confounders nobody stated, a little like it is playing improv. So how you ask matters as much as which model you pick, and the persona and the study have to be specified richly, from the participant's point of view. Accuracy can hide the damage. A model that looks right on average can still distort subgroups and flatten the shape of the distribution, collapsing the variation that mattered. The fix here is a noise floor: score humans against other humans first, so you know the best agreement any method could reach, then judge synthetic against human relative to that floor rather than against a perfect match that never existed. Treat a synthetic persona as an economic actor, not ground truth, since even the human study is not ground truth, and validate it against real outcomes before you let it drive a decision. Speaker info: - https://x.com/ianand - https://www.linkedin.com/in/ishananand/ - https://ishananand.com/ Timestamps: 0:00 - Introduction: synthetic personas for market research 1:43 - Why this talk: separating signal from noise 2:46 - Forecasting people like we forecast the weather 4:04 - A thousand humans vs their agent replicas 5:22 - Purchase probability and prompt sensitivity 6:41 - Invented confounders and specifying the persona 7:45 - Question order and framing effects 8:50 - Predicting stated attitudes vs experts 10:07 - Prompting techniques and subpopulation methods 13:08 - Reconstructing and scoring the full distribution 16:09 - Calibrating personas against real forecasts 17:38 - Setting a noise floor with human vs human 18:40 - Treat personas as economic actors, and what's next --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/YnNF55QV0zs-persona-engineering-a-field-guide-to-ai-synthetic-personas-ishan-anand-insightsciences-ai Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- canonical: >- https://liquidgold.cc/channel/aidotengineer/explore/video/persona-engineering-a-field-guide-to-ai-synthetic-p-f64471adbf7b markdown: >- https://liquidgold.cc/channel/aidotengineer/video/persona-engineering-a-field-guide-to-ai-synthetic-p-f64471adbf7b.md llms: null parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-24T06:29:46.354Z' language: null children: [] videoId: YnNF55QV0zs durationSeconds: 1269 publishedAt: '2026-07-29T20:15:35.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 13 endSeconds: 24 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=13' heldReason: rights - ordinal: 1 text: null startSeconds: 24 endSeconds: 103 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=24' heldReason: rights - ordinal: 2 text: null startSeconds: 24 endSeconds: 103 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=24' heldReason: rights - ordinal: 3 text: null startSeconds: 183 endSeconds: 251 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=183' heldReason: rights - ordinal: 4 text: null startSeconds: 183 endSeconds: 251 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=183' heldReason: rights - ordinal: 5 text: null startSeconds: 261 endSeconds: 328 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=261' heldReason: rights - ordinal: 6 text: null startSeconds: 339 endSeconds: 404 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=339' heldReason: rights - ordinal: 7 text: null startSeconds: 339 endSeconds: 404 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=339' heldReason: rights - ordinal: 8 text: null startSeconds: 414 endSeconds: 469 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=414' heldReason: rights - ordinal: 9 text: null startSeconds: 414 endSeconds: 469 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=414' heldReason: rights - ordinal: 10 text: null startSeconds: 479 endSeconds: 536 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=479' heldReason: rights - ordinal: 11 text: null startSeconds: 479 endSeconds: 536 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=479' heldReason: rights - ordinal: 12 text: null startSeconds: 479 endSeconds: 536 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=479' heldReason: rights - ordinal: 13 text: null startSeconds: 547 endSeconds: 612 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=547' heldReason: rights - ordinal: 14 text: null startSeconds: 612 endSeconds: 623 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=612' heldReason: rights - ordinal: 15 text: null startSeconds: 623 endSeconds: 791 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=623' heldReason: rights - ordinal: 16 text: null startSeconds: 623 endSeconds: 791 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=623' heldReason: rights - ordinal: 17 text: null startSeconds: 623 endSeconds: 791 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=623' heldReason: rights - ordinal: 18 text: null startSeconds: 802 endSeconds: 976 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=802' heldReason: rights - ordinal: 19 text: null startSeconds: 802 endSeconds: 976 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=802' heldReason: rights - ordinal: 20 text: null startSeconds: 802 endSeconds: 976 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=802' heldReason: rights - ordinal: 21 text: null startSeconds: 989 endSeconds: 1064 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=989' heldReason: rights - ordinal: 22 text: null startSeconds: 989 endSeconds: 1064 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=989' heldReason: rights - ordinal: 23 text: null startSeconds: 989 endSeconds: 1064 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=989' heldReason: rights - ordinal: 24 text: null startSeconds: 989 endSeconds: 1064 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=989' heldReason: rights - ordinal: 25 text: null startSeconds: 1064 endSeconds: 1074 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=1064' heldReason: rights - ordinal: 26 text: null startSeconds: 1064 endSeconds: 1074 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=1064' heldReason: rights - ordinal: 27 text: null startSeconds: 1074 endSeconds: 1129 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=1074' heldReason: rights - ordinal: 28 text: null startSeconds: 1140 endSeconds: 1269 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=1140' heldReason: rights - ordinal: 29 text: null startSeconds: 1140 endSeconds: 1269 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=1140' heldReason: rights - ordinal: 30 text: null startSeconds: 1140 endSeconds: 1269 watch: 'https://liquidgold.cc/watch/YnNF55QV0zs?t=1140' heldReason: rights --- # Persona Engineering: A Field Guide to AI Synthetic Personas — Ishan Anand, InsightSciences.ai A team ran about a thousand people through a market research survey, then had LLM agents replay the same questions, and the agents matched the humans closely while carrying less noise than the humans did themselves. That is exactly the tell: a synthetic respondent smooths over the messiness that makes a real population real. Ishan Anand walks through where it breaks. Nudge one variable in the prompt template and purchase probability swings, because the model infers latent confounders nobody stated, a little like it is playing improv. So how you ask matters as much as which model you pick, and the persona and the study have to be specified richly, from the participant's point of view. Accuracy can hide the damage. A model that looks right on average can still distort subgroups and flatten the shape of the distribution, collapsing the variation that mattered. The fix here is a noise floor: score humans against other humans first, so you know the best agreement any method could reach, then judge synthetic against human relative to that floor rather than against a perfect match that never existed. Treat a synthetic persona as an economic actor, not ground truth, since even the human study is not ground truth, and validate it against real outcomes before you let it drive a decision. Speaker info: - https://x.com/ianand - https://www.linkedin.com/in/ishananand/ - https://ishananand.com/ Timestamps: 0:00 - Introduction: synthetic personas for market research 1:43 - Why this talk: separating signal from noise 2:46 - Forecasting people like we forecast the weather 4:04 - A thousand humans vs their agent replicas 5:22 - Purchase probability and prompt sensitivity 6:41 - Invented confounders and specifying the persona 7:45 - Question order and framing effects 8:50 - Predicting stated attitudes vs experts 10:07 - Prompting techniques and subpopulation methods 13:08 - Reconstructing and scoring the full distribution 16:09 - Calibrating personas against real forecasts 17:38 - Setting a noise floor with human vs human 18:40 - Treat personas as economic actors, and what's next --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/YnNF55QV0zs-persona-engineering-a-field-guide-to-ai-synthetic-personas-ishan-anand-insightsciences-ai Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- Canonical: https://liquidgold.cc/channel/aidotengineer/explore/video/persona-engineering-a-field-guide-to-ai-synthetic-p-f64471adbf7b · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md) > Search chunk text is held until public-source provenance is verified. ## Document chunks > Missing: no stored document chunks are available.