Machine Learning Emulation of Virtual Respondents for Survey Data

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing data collection methods face challenges in obtaining reliable data due to issues like unreliable responses from autonomous programs, distracted or inattentive respondents, and malicious actors, as well as difficulties in collecting data from a significant number of individuals across various populations.

Innovation Solution

A machine learning-based system and method for emulating virtual respondents by generating simulated responses to surveys and questionnaires. This involves collecting reference survey artifacts, obtaining demographic data, constructing demographic vectors, and using nominal or ordinal response generation models to generate responses based on iterative contextual parameters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional survey methods are used to collect data from individual members of a population, then data can be obtained through surveys and questionnaires, but the data may be unreliable due to autonomous programs posing as respondents, distracted respondents, or malicious actors

Engineering Contradiction:

Improvedata reliabilityVSAvoidunreliable responses from autonomous programs, distracted respondents, and malicious actors

Core Design Contradiction:

ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent creates virtual respondents that are copies or simulations of real respondents, using machine learning models to generate responses that mimic human behavior patterns. These virtual respondents serve as reliable data sources without being susceptible to the harmful factors affecting real human respondents, such as distraction or malicious intent.

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent replaces the mechanical system of human respondents physically completing surveys with an automated computational system. Machine learning models generate responses algorithmically, substituting human cognitive processes with computational processes that are not subject to distraction, fatigue, or malicious behavior.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Quantity of substance

If traditional survey methods are used to collect data, then information can be extracted from individuals, but it is challenging to collect data from a significant or required number of individuals of various populations

Engineering Contradiction:

Improvenumber of respondentsVSAvoiddata collection efficiency

Core Design Contradiction:

Quantity of substanceVSProductivity

Solution Approach 1:

The patent generates large quantities of virtual respondents through computational processes, creating synthetic data from numerous simulated individuals. This approach bypasses the need to recruit and coordinate large numbers of real human participants, enabling rapid generation of extensive datasets with diverse population characteristics.

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent varies parameters such as demographic characteristics, response patterns, and behavioral attributes across virtual respondents to simulate diversity across different population groups. By adjusting these parameters computationally, the system can efficiently generate data representing significant numbers of individuals from various populations without the logistical challenges of real-world recruitment.

Inventive Principle:

Principle #35Parameter changes

3Productivity

If machine learning models are used to generate responses for virtual respondents, then simulated responses can be created efficiently, but the system complexity increases with multiple models for nominal and ordinal questions

Engineering Contradiction:

Improveresponse generation efficiencyVSAvoidsystem complexity with multiple response generation models

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent divides the response generation task into separate specialized models: one for nominal questions and another for ordinal questions. Each model is optimized for its specific question type, improving generation efficiency and accuracy while allowing the overall system to handle diverse questionnaire formats through modular architecture.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent creates a unified virtual respondent system that can handle multiple question types (nominal and ordinal) through its ensemble of specialized models. The system universally processes different survey formats by routing questions to appropriate models, maintaining efficiency across varied questionnaire structures while managing complexity through organized model specialization.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS20250037157A1Systems and methods for machine learning-based emulation of virtual respondents

Publication Date: 2025.01.30 SIMSURVEYS LLC

AI summary

A system and method for automated generation of simulated responses from one or more virtual emulated respondents includes collecting a reference survey artifact comprising unanswered questions, obtaining demographic data for the one or more virtual emulated respondents, constructing a demographic vector for each virtual emulated respondent representing the characteristics of the associated virtual emulated respondent, generating responses for each target virtual emulated respondent to each unanswered question by transmitting inputs including the associated demographic vector, each unanswered question, and an iterative contextual parameter to a response generation model, and updating a simulated response dataset with the generated responses.