Prioritized AI Risk Checklist Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AI system development and risk assessment methods require significant man-hours to materialize AI systems and identify ethical risks, as checklists do not specify which components to focus on, and risk chain models need manual organization of risk elements for AI system components and stakeholders.
Innovation Solution
A risk analysis program that acquires relational information about AI system configurations, determines priorities based on object person attributes, and outputs a prioritized checklist for AI system developers and providers to address ethical risks, automating the generation of checklists based on graph structures representing AI system interactions with stakeholders.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If checklists from AI ethics principles and guidelines are applied to AI systems, then ethical risk recognition is improved, but the materialization work becomes highly difficult and requires a large amount of man-hours
Solution Approach 1:
The patent segments the comprehensive AI ethics checklist into multiple attribute categories (object person attributes, processing attributes, data attributes). By dividing the monolithic checklist into structured segments with hierarchical relationships, the system enables selective application of check items based on specific AI system configurations, thereby reducing the overall man-hours required while maintaining thorough ethical risk coverage.
Solution Approach 2:
The patent introduces priority levels as a new parameter to classify check items based on their importance and relevance to different AI system configurations. By changing the parameter structure from a flat checklist to a prioritized, multi-attribute hierarchy, the system allows developers to focus on high-priority items first, significantly reducing the time investment required for ethical risk assessment.
2Reliability
If risk chain models are used to organize risk constituent elements, then risk control examination is improved, but manual organization of risk elements for AI system components and stakeholders is required
Solution Approach 1:
The patent creates a universal attribute hierarchy structure that can be applied across different AI system types and configurations. The standardized attributes (object person, processing, data) serve multiple functions: they organize risk elements systematically, enable automated priority determination, and work across various AI application domains. This universality eliminates the need for manual customization of risk element organization for each new AI system.
Solution Approach 2:
The system enables automated priority determination of check items based on the AI system's configuration and the attribute hierarchy. Instead of requiring manual organization and prioritization of risk elements, the system automatically assigns priorities based on predefined rules and attribute relationships, making the risk element organization process self-service and significantly reducing manual effort.
3Reliability
If comprehensive checklists are applied to all AI system components and stakeholders, then ethical risk coverage is improved, but the complexity of implementation increases
Solution Approach 1:
The patent segments the comprehensive checklist into three main attribute categories (object person, processing, data) with multiple hierarchical levels. This segmentation allows the system to maintain comprehensive ethical risk coverage while reducing implementation complexity by enabling selective application of check items based on the specific AI system configuration and relevant stakeholders.
Solution Approach 2:
The patent implements partial action by allowing AI system developers to apply only the necessary subset of check items based on their system's specific attributes and configurations. Instead of requiring implementation of all checklist items universally, the system enables selective application of relevant check items, thereby maintaining comprehensive coverage where needed while reducing complexity where not required.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A risk analysis program for an artificial intelligence (AI) system, the risk analysis program being a program including instructions which, when executed by a computer, cause the computer to execute processing, the processing including: acquiring a plurality of pieces of relational information that include at least two attributes among an attribute of a type of an object person, an attribute of a type of processing, and an attribute of a type of data, wherein the relational information is determined on a basis of a configuration of the AI system; determining a priority of the plurality of pieces of relational information on a basis of the attribute of the type of the object person; and outputting one or a plurality of check items selected on a basis of the determined priority from among a plurality of check items associated with each attribute as a checklist for the AI system.