Loading…
Evidence and trust
HDTwin turns a target description and research evidence into a virtual panel that can be questioned again and again. This page shows what supports that work, what its outputs mean, and where its limits apply.
Target fidelity
Panels start from a description of the target audience and can be calibrated with observed research evidence. Calibration can be measured and refined; it does not promise a perfect replica, but experiments with our clients have shown a level of alignment of over 98% between the answers given by the synthetic panel and those from the real panel.
Deliberation and consensus
Human Digital Twins discuss the research topic, exchange reasons, and form opinion groups. Agreement and disagreement stay traceable to individual answers, so consensus can be checked instead of assumed.
Model control
Several AI models take part under controlled settings, and known model tendencies are checked against observed evidence. Remaining tendencies are reported as limitations, never as removed by assertion.
Operational evidence
HDTwin panels support market, communication, and public research work, from early exploration to stakeholder review.
Information levels
Every output is one of three things. The distinction stays visible throughout the product.
Collected from real people through fieldwork or published studies.
Derived inside the model from observed material, shown with its assumptions.
Generated by the panel during deliberation. Useful for exploration, not a record of real behaviour.
How opinions form, where disagreement sits, and which formulations gather broader agreement.
How real people will behave. Synthetic outputs support decisions; they do not replace real-world research.
Governance
The documents that govern responsible use live in one place.
Guides and references for building and running panels.
The public experiment behind the deliberation method.
Panels built on your inputs are considered your exclusive property, are delivered directly to you and are not stored on our infrastructure, except for the strictly necessary time for processing. No marketing cookies. No behavioural tracking. No use for training our systems.
Demo requests are counted on the booking page only. This site sets no advertising cookies, runs no behavioural profiling, and uses no invasive visitor analytics.
Trust questions
How HDTwin results are checked, where they stop, how customer material is handled, and how regulation is met.
A virtual panel is checked by comparing its behaviour with evidence collected from real populations, including a human control group answering the same questions. Agreement is measured where reference material exists; outside that perimeter answers are read as exploration.
Synthetic outputs are decision-support evidence, not observations collected from real people. Confidence falls as questions move away from the source material, and panels cannot guarantee how real people will behave.
Panels build on the target description and the research material a customer shares for calibration, or on generic panels provided by HDTwin. Panels built on your inputs are considered your exclusive property, are delivered directly to you and are not stored on our infrastructure, except for the strictly necessary time for processing.
HDTwin was natively designed to fully comply with the European privacy regulation (GDPR) and the EU AI Act.
A personal demo walks through your target audience with a question from your own work.
