Behavior Change Advice Using Risk Causal Effect Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies fail to accurately derive the influence of risk factors on user risks, making it difficult to encourage effective behavioral changes despite presenting risk and risk factor information to users.
Innovation Solution
A behavioral change promotion device that builds a first learning model to estimate user risks and a second learning model to estimate risk causal effects, generating advice information that includes both risk and risk causal effects to enhance user perception and encourage effective behavioral changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If advice information including only risk and risk factors is presented to users, then the information presentation is simple, but users cannot accurately perceive the influence of risk factors on their risks
Solution Approach 1:
The patent segments the risk assessment into two distinct learning models: a first learning model that estimates overall risk, and a second learning model that estimates risk causal effects (the influence of risk factors). This segmentation allows the system to provide detailed causal information to improve user perception accuracy while keeping each individual model relatively simple and manageable.
Solution Approach 2:
The patent introduces a second learning model as an intermediary component that specifically processes the relationship between risk factors and risk outcomes. This intermediary model acts as a mediator that quantifies and communicates the causal influence of risk factors to users, enabling them to accurately perceive how specific behaviors or conditions affect their risk levels without overwhelming them with raw data.
2Loss of information
If a single learning model is used to estimate risk, then the device complexity is low, but it cannot determine the influences of risk factors on the risk
Solution Approach 1:
The patent divides the risk assessment function into two specialized learning models: the first learning model focuses on estimating overall risk levels, while the second learning model specifically estimates risk causal effects. This segmentation ensures that no information is lost - each model handles a specific aspect of risk analysis, and together they provide complete information about both the risk level and the influence of risk factors.
Solution Approach 2:
The patent creates a multi-functional risk assessment system where two learning models work together to provide comprehensive risk information. The first model provides overall risk estimation, while the second model provides causal effect analysis. This universal approach allows the system to handle multiple types of risk information simultaneously, ensuring that neither risk levels nor risk factor influences are lost.
3Loss of information
If risk information is presented without risk causal effects, then the information processing is simple, but users cannot understand the degree of increase of risk according to risk factors
Solution Approach 1:
The patent implements feedback by having the second learning model estimate and return risk causal effects to the advice generation unit. This feedback loop provides users with specific information about how much their risk increases due to particular risk factors, enabling them to understand the quantitative impact of their behaviors and make more effective behavioral changes.
Solution Approach 2:
The second learning model serves as an intermediary that processes the relationship between risk factors and risk increases, translating complex causal relationships into understandable information about risk increase degrees. This intermediary component bridges the gap between raw risk data and actionable behavioral insights, enabling users to comprehend and respond to risk information effectively.
Data Source
AI summary
A behavioral change promotion device includes a learning unit configured to build a prediction model used for estimating a risk by performing learning with learning user information and risk information being associated with each other and build a causal model used for estimating a risk causal effect by performing learning with information relating to a risk factor and the risk information being associated with each other; an estimation unit configured to estimate a risk of the user by inputting estimation user information to a prediction model and estimate a risk causal effect by inputting information relating to a risk factor to a causal model; an advice generating unit configured to generate advice information including at least the risk and the risk causal effect; and an output unit configured to output the advice information.