Implicit User Interaction Scoring via Gaze and Device Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated systems struggle to accurately estimate a person's attention level during a presentation, often leading to false negatives or false positives due to assumptions about gaze direction, which can result in inappropriate content delivery and user frustration.
Innovation Solution
Incorporating client device interaction levels into a model to dynamically score implicit user interactions, allowing for a more accurate estimation of attention by analyzing gaze direction in conjunction with device interactions such as social media engagement and search queries related to the presented content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If automated systems use imaging systems to capture images and estimate attention level based on gaze direction, then the system can automatically measure attention without explicit user interaction, but the system produces false negatives and false positives when users look away to engage with client devices related to the presentation content
Solution Approach 1:
The patent combines multiple data sources including imaging system gaze detection, client device interaction data, and presentation content analysis to create a comprehensive attention estimation model. This merging of previously separate measurement approaches allows the system to distinguish between genuine disengagement and productive engagement with presentation-related content.
Solution Approach 2:
The patent introduces an intermediary analysis layer that processes client device interactions and presentation content to mediate between raw gaze data and final attention determination. This intermediary layer contextualizes gaze-away events by analyzing whether device interactions are presentation-related, preventing false negative classifications.
2Device complexity
If the system assumes that looking at the presentation indicates high attention and looking away indicates low attention, then the model is simple to implement, but the model fails to capture true attention level when users engage with client devices
Solution Approach 1:
The patent transforms the static gaze-based attention model into a dynamic multi-factor model that continuously integrates new data streams. The system adapts its interpretation of gaze behavior based on real-time analysis of client device interactions and presentation content, allowing the model to evolve from simple geometric assumptions to contextual understanding.
Solution Approach 2:
The patent changes the parameters used for attention estimation from purely spatial (gaze direction coordinates) to include behavioral and contextual parameters (device interaction type, content relevance, timing patterns). This parameter expansion allows the system to distinguish between different types of attention states without excessive complexity.
3Productivity
If the system incorrectly identifies disinterested users based on gaze direction alone, then the system can make quick decisions about content delivery, but the system wastes bandwidth and frustrates users by providing inappropriate content
Solution Approach 1:
The patent performs preliminary analysis of client device interactions and presentation content before making final attention determination and content delivery decisions. By pre-processing and contextualizing device interaction data, the system avoids hasty incorrect decisions about content delivery, reducing wasted bandwidth on disinterested users.
Data Source
AI summary
A method of dynamically scoring implicit interactions can include receiving, by an interaction analysis server from an imaging system, a plurality of images of an environment captured in a period of time corresponding to display of a presentation, retrieving, by the interaction analysis server, content information corresponding to content of the presentation, and identifying, by a presence detector of the interaction analysis server, that a face appears in at least one image of the plurality of images. The method can further include matching, by a facial recognition system of the interaction analysis server, the face with a user identifier, retrieving, by a client device information retriever, client device information associated with the user identifier and corresponding to the period of time, and calculating, by a client device interaction score calculator, a client device interaction score based on one or more correspondences between the client device information and the content information.