Loading…
How it works
Behind the source evidence sits a shared map of meaning. Interview statements and survey answers are read together, so what different people meant in different words can be compared on the same ground.
We call this shared map the Latent Cognitive Model.
The Latent Cognitive Model (LCM) is an intermediate, structured model that reconstructs an individual's underlying system of thought from the collected evidence, treated as observable but incomplete proofs. Rather than describing only what a person believes, the LCM infers how they weigh evidence, organise their values and handle uncertainty, making it possible to generate virtual agents that answer stably and coherently even to entirely novel questions.
Method questions
Short answers to the questions visitors ask most about the method. Each answer states how HDTwin works and where its limits apply.
Synthetic research uses a virtual panel of Human Digital Twins to explore how a target audience may respond to questions, messages, and decisions. It extends real-world research by making existing evidence reusable for follow-up exploration, and it never replaces fieldwork with real people.
A synthetic panel is the group of Human Digital Twins built for one study: each participant carries characteristics, attitudes, and information drawn from the target description and any research evidence supplied. The panel stays available and can be questioned again as new hypotheses emerge.
ChatGPT produces a single plausible answer from one model. HDTwin runs a defined panel of Human Digital Twins through iterative surveys and simulated group discussion, so positions, arguments, opinion groups, disagreement, and consensus stay inspectable instead of collapsing into one response.
A panel starts from a written description of the target audience. Where interviews and surveys exist, their evidence calibrates the Human Digital Twins; the resulting virtual panel stays available for further questions without new recruitment.
Every language model brings its own tendencies. HDTwin adapts models to the character of each participant through Digital Imprinting, which reduces unwanted model influence without removing it. Remaining tendencies are reported as limitations.
A synthetic persona is one Human Digital Twin inside the panel: a configured participant representing relevant characteristics, attitudes, and information of a member of the target population. Personas deliberate together, so the outcome reflects a population rather than a single voice.
A personalized demo walks through how your target description becomes an explorable panel, using a question from your own work.
