Sequential Decision Drift Detection With Gaze-Based Regularization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing decision-making models ignore drift, assuming fixed environments and constant utility differences between choice options, leading to biased decisions due to fluctuations in evidence accumulation over time.
Innovation Solution
A system and method that utilize sensors to detect user gaze and touch inputs, learn drift using machine learning, and automatically regularize choice options by predicting drift direction and classifying risk-averse or risk-seeking behaviors to output a desirable behavior model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing decision-making models assume fixed environments and constant utility differences, then model simplicity is maintained, but decision accuracy deteriorates due to ignoring drift in evidence accumulation
Solution Approach 1:
The patent transforms static decision-making models into dynamic systems that continuously adapt to environmental changes. Drift detection mechanisms monitor evidence accumulation patterns over time, and the system dynamically adjusts utility estimates and choice recommendations based on detected drift states, allowing the model to respond to changing conditions while maintaining operational simplicity through automated adjustments
Solution Approach 2:
The system implements feedback loops where drift detection results feed back into the decision-making process. By continuously monitoring choice patterns and evidence accumulation, the system detects drift and feeds this information back to adjust utility estimates and regularization parameters, creating a closed-loop system that improves decision accuracy through continuous learning and adaptation
2Reliability
If drift detection and regularization mechanisms are implemented, then decision accuracy is improved, but system complexity increases
Solution Approach 1:
The system applies partial regularization selectively based on detected drift states rather than uniformly across all decisions. By identifying specific drift conditions and applying regularization only when necessary, the system achieves improved reliability without the full computational overhead of continuous regularization, implementing just enough correction to maintain decision quality
Solution Approach 2:
The patent dynamically changes model parameters such as utility estimates and regularization strengths based on detected drift patterns. By adjusting these parameters in response to drift detection, the system maintains high reliability without requiring a complete redesign of the decision-making architecture, achieving improved performance through parameter adaptation rather than structural complexity
3Measurement precision
If choice options are automatically regularized when drift is detected, then bias in decisions is reduced, but processing time increases
Solution Approach 1:
The system performs preliminary drift detection and assessment before final decision presentation. By detecting drift patterns early in the decision process and pre-computing regularization adjustments, the system reduces bias without adding significant processing time to the final decision moment, as much of the computational work is done in advance during the drift monitoring phase
Data Source
AI summary
Embodiments of the present disclosure are directed to systems and methods to display a plurality of items in a spatial layout, monitor an eye gaze of the user while making sequential decisions about the plurality of items, determine that a drift has occurred, automatically regularize a plurality of alternative choice options, learn and identify an upper boundary and a lower boundary based on whether the user chooses or rejects at least one alternative choice, model the drift using machine learning based at least in part on a history of user decision responses, predict a drift direction of the drift and a relevance impact of different attributes into a drift state, classify the drift states into a risk-averse classification and a risk-seeking classification by clustering drift parameters over different behavioral measures to learn an attentive drift of the user, and output a desirable behavior model for machine learning.