Electronic Device Recommendation Explanation Module
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recommendation systems fail to provide users with an explanation for why specific items are recommended, leading to a lack of transparency and user dissatisfaction, as they only output a list of recommended items along with their scores without explaining the reasoning behind the recommendations.
Innovation Solution
An electronic device and method that receive a list of items and their corresponding scores from an external device, identify relevancies between item features and user preferences, and output these relevancies along with the recommended items, allowing users to understand why certain items are recommended.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a recommendation system outputs only a list of recommended items and scores, then the system complexity is low and processing is fast, but the user does not understand why items are recommended leading to user dissatisfaction
Solution Approach 1:
The patent introduces an explanation generation module as an intermediary component that takes the recommendation results and generates human-understandable explanations. This module acts as a mediator between the complex recommendation algorithm and the user, translating technical scoring mechanisms into natural language explanations without requiring changes to the core recommendation system architecture.
Solution Approach 2:
The system segments the recommendation process into distinct components: the recommendation engine that generates scores, and the explanation generation module that creates human-readable justifications. This segmentation allows the complex scoring process to remain opaque while the explanation layer provides transparency, resolving the contradiction between maintaining simple output and providing understanding.
2Measurement precision
If complex calculations are performed to determine recommended items, then the accuracy of recommendations improves, but no explanation is provided as to why items are output
Solution Approach 1:
The system implements feedback by analyzing the complex calculation results and generating explanatory information that feeds back to the user. The explanation generation module processes the scoring data and returns human-understandable reasons, ensuring that the information loss from complex calculations is compensated through structured feedback in the form of explanations.
Solution Approach 2:
The patent transforms the parameters from raw numerical scores into qualitative explanations. By changing the parameter representation from quantitative scoring metrics to natural language descriptions, the system preserves the accuracy benefits of complex calculations while eliminating the information loss about reasoning processes.
3Reliability
If recommendation scores are provided with item lists, then some transparency is achieved, but users still do not know why calculated scores are higher for specific items
Solution Approach 1:
The patent replaces the mechanical scoring system with a linguistic explanation system. Instead of presenting users with numerical scores and implicit calculation mechanisms, the system substitutes these with natural language explanations that are inherently more interpretable. This substitution maintains reliability by preserving the original scoring logic while making the reasoning accessible through language rather than mathematics.
Data Source
AI summary
An electronic device is provided. The device includes a processor configured to receive, from a first external device, a list of a first number of items generated based on a request of a user, and a first score of each of the first number of items, identify a first relevancy between each element of a first feature set and each of the first number of items and a second relevancy between each element of the set and the user, identify a first list of a second number of items, identify, for at least one item of the first list, a third relevancy between a first feature of the set which satisfies a condition and the at least one item and a fourth relevancy between the first feature and the user, based on the first relevancy and the second relevancy, and output the third relevancy and the fourth relevancy with the first list.