Affective Response Prediction Discrepancy Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for predicting affective responses in users face inaccuracies due to discrepancies between predicted and measured responses, necessitating a method to identify and address these inaccuracies to improve model accuracy.
Innovation Solution
A system comprising a feature generator, Emotional Response Predictor (ERP), discrepancy detector module, and investigation module that interacts with users to determine the cause of inaccuracies by presenting event factors and receiving user comments, allowing for the identification of discrepancies in event descriptions and model inaccuracies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a model of the user is generated using sensor measurements over long periods to predict affective response, then the detail and personalization of the model is improved, but the accuracy of predictions deteriorates due to discrepancies between predicted and measured responses
Solution Approach 1:
The system implements a feedback mechanism where sensor measurements of actual affective response are compared with model predictions. When discrepancies are detected, the system investigates the cause and uses user comments to refine the model, creating a continuous improvement loop that maintains personalization while enhancing prediction accuracy
Solution Approach 2:
The patent replaces pure automated model prediction with a hybrid approach that incorporates human feedback. The investigation module presents event factors to users and collects their comments, substituting automated decision-making with human judgment to identify and correct model inaccuracies
2Reliability
If sensor-based measurements are used as ground truth for evaluating model accuracy, then the reliability of evaluation is improved, but the complexity of the system increases due to the need for continuous measurement and comparison
Solution Approach 1:
The system extracts only the necessary sensor measurements needed for evaluation purposes rather than continuously monitoring all physiological parameters. The discrepancy detector module selectively compares specific model predictions with corresponding sensor measurements, reducing the overall system complexity while maintaining evaluation reliability
3Measurement precision
If the investigation module interacts with users to identify discrepancies in event descriptions and model inaccuracies, then the accuracy of model correction is improved, but the time required for correction increases
Solution Approach 1:
The investigation module proactively presents event factors to users before model correction is needed, allowing users to provide feedback on event descriptions in advance. This preliminary action reduces the time required for correction by having users mentally prepare and organize their thoughts about potential discrepancies
Solution Approach 2:
The investigation process is segmented into discrete event factors that are presented to users individually. This segmentation allows users to focus on specific aspects of the event description and model prediction, making the correction process more efficient and less time-consuming while maintaining high accuracy
Data Source
AI summary
Some aspects of this disclosure include systems, methods, and/or computer programs that may be used to determine a cause of an inaccuracy in predicted affective response to an event that involves a user who has an experience. Some embodiments described herein involve identifying when a difference between a measurement of affective response corresponding to an event is different from the predicted affective response corresponding to the event. When such a discrepancy is identified, a presentation is made to the user of at least one of the following: one or more factors characterizing the event, and effects of the one or more factors on the user, as determined based on a model of the user. Based on a comment made by the user in response to the presentation, at least one of the following is identified: a discrepancy in a description of the event, and an inaccuracy in the model.