Silicon Sample
Definition
Simulated responses of a language model are intended to approximate perspectives of a described group of people.
A Silicon Sample arises when a language model is conditioned with information about hypothetical persons and answers survey questions. This produces artificial answers, not actually collected opinions.
Simply explained
You describe invented persons and let an AI answer from their perspective. Whether real people would answer similarly must be checked separately.
Why it matters in practice
Such samples can prepare questions and hypotheses before a real survey begins.
Example from business
Hypothetical example: A market research team simulates answers from various household profiles to discover unclear questions in the questionnaire.
Limits and pitfalls
- Model stereotypes can appear like population differences. Agreement in one study does not prove general representativeness.
Origin & history
The approach builds on research into simulating human survey responses with language models.
Use in mid-sized companies
- Use for preparing interviews and initial scenarios; verify important statements with real survey data.
Frequently asked questions
Does this replace a customer survey?
No, simulated answers are not proof of actual customer opinions.
Sources
Editorially reviewed on: