Natural Language Dispute Interface for Incremental Transaction Details
Solution Overview
Problem
Existing user interfaces overwhelm non-technical users with complex information, requiring them to navigate through multiple layers and decipher obscure terminology, leading to frustration and inefficiency, especially when disputing transactions.
Innovation Solution
Implementing a generative AI interface that converses with users in natural language, providing transaction details incrementally and automatically flags users exceeding a dispute threshold, conserving computational resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If detailed transaction information is presented to users, then decision-making quality improves, but user interface complexity and user frustration increase
Solution Approach 1:
The patent segments transaction information into multiple hierarchical levels: summary level (high-level overview), detailed level (specific transaction data), and evidentiary level (supporting documents). Users can progressively drill down from summary to detailed information as needed, rather than being presented with all information at once. This resolves the contradiction by providing complete information when needed while maintaining interface simplicity during normal operation.
Solution Approach 2:
The interface dynamically adapts its information presentation based on user context, device type, and interaction history. The system adjusts the amount and type of information displayed in real-time, transforming from a static information hierarchy to a dynamic response system. This allows the interface to provide sufficient detail for decision-making while simplifying presentation for usability in different contexts.
2Loss of information
If comprehensive transaction details are displayed, then user understanding improves, but information overload and user frustration increase
Solution Approach 1:
Transaction information is divided into contextual segments (merchant information, timing, location, amount, category) that can be independently accessed. Users receive only the relevant contextual segments needed for their specific dispute situation, rather than being overwhelmed by all possible transaction details. This segmentation maintains ease of operation while ensuring complete contextual understanding.
Solution Approach 2:
Different portions of the interface provide different levels of information density appropriate to their function. Summary sections provide high-level context, while detailed sections provide specific data points. This local differentiation of information quality allows users to understand transaction context without information overload, as each area presents appropriate detail for its purpose.
3Reliability
If surveillance measures are applied to all disputed transactions, then fraud detection reliability improves, but computational resource consumption increases
Solution Approach 1:
The system changes the parameter of surveillance intensity based on risk assessment. Low-risk disputes receive minimal surveillance (standard processing), while high-risk disputes trigger enhanced surveillance measures (detailed analysis, additional verification). This parameter adjustment maintains fraud detection reliability for suspicious cases while conserving computational resources for the majority of low-risk transactions.
Solution Approach 2:
The system performs self-assessment of dispute risk using automated analysis of transaction patterns, user history, and dispute characteristics. This self-service risk classification enables the system to automatically determine which cases require intensive surveillance, reducing reliance on uniform high-cost processing while maintaining detection reliability for problematic cases.
Data Source
AI summary
Presenting a natural language response to a user query for a disputed transaction involves gathering transaction details of the user, sending the transaction details and the query to a generative model, receiving a natural language response from the generative model, and presenting the natural language response to the user in a user interface. The transaction details comprise the transaction and a transaction history of the user. The generative model is trained to present a natural language response to the user about the transaction. The natural language response comprises a datum of the transaction present. Further user interaction involves receives an additional query about the transaction; and presenting an additional natural language response to the user that contain an additional datum of the transaction. Flagging a user for a disputed transaction involves determining is based on a user's natural language answer, transaction history, and a disputed transaction threshold.



