Software Usability Testing via Predictive Model Integration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for evaluating software usability do not effectively integrate predictive models with test results, and entering information defining input sequences into these models is complex, limiting the ability to provide accurate predictions of task completion times and event counts.
Innovation Solution
A computer program product that generates electronic records of input events during test sessions and applies them to predictive models, such as the GOMS model, to produce predictions of task completion times and event counts, while analyzing coding to determine relevant characteristics and deducting irrelevant time, thereby improving the integration of predictive outputs with actual test results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If predictive models (GOMS) are used to calculate theoretical time for user inputs, then prediction accuracy for task completion time is improved, but the complexity of entering information defining input sequences into the model increases
Solution Approach 1:
The patent uses screen capture images as visual copies of the software interface to train the neural network, eliminating the need to manually encode complex input sequences into predictive models. The neural network learns directly from visual representations, copying the interface structure and user interaction patterns from images rather than requiring detailed textual descriptions of each input event.
Solution Approach 2:
The patent replaces traditional mechanical GOMS modeling (which requires manual decomposition of tasks into discrete operations with fixed time values) with a neural network system that automatically learns interaction patterns from visual data and test results, substituting manual model construction with automated machine learning.
2Ease of manufacture
If traditional test sessions are monitored and evaluated manually or with basic recording equipment, then implementation simplicity is maintained, but the ability to integrate predictive models with test results is reduced
Solution Approach 1:
The patent creates a multi-functional system where the neural network serves multiple purposes: it predicts task completion times, identifies relevant characteristics from screen captures, processes various input devices uniformly, and integrates both predictive modeling and actual test result analysis within a single framework, replacing multiple separate tools.
Solution Approach 2:
The system automatically processes test sessions by capturing screen images, extracting relevant characteristics, comparing predicted versus actual performance, and generating usability metrics without requiring manual monitoring or complex post-processing, making the system self-sufficient while maintaining simplicity.
3Reliability
If total time in test sessions is measured including all activities, then complete monitoring is achieved, but irrelevant time (talking with instructor, receiving instructions) skews the test results
Solution Approach 1:
The patent extracts only the relevant portions of test time by analyzing screen capture sequences to identify when actual software interactions occur versus when users are engaged in irrelevant activities like listening to instructions. The system separates useful interaction time from irrelevant time based on visual evidence from screen captures rather than relying on total elapsed time.
Solution Approach 2:
The system uses feedback from comparing predicted interaction patterns with actual screen capture sequences to identify and exclude irrelevant time periods. By continuously monitoring whether user actions match expected task sequences, the system can detect and filter out time spent on non-task activities, improving measurement accuracy.
Data Source
AI summary
Usability of a software program can be tested by generating a first electronic record of input events that a user executes with at least one input device during a test session. The operations comprise applying the generated first electronic record to a predictive model to generate a prediction for performing the predefined task. Usability of a software program can be tested by registering input events that a user executes with at least one input device during a test session in which the user performs a predefined task of a software program. The operations comprise generating a first electronic record associated with the test session. The first electronic record includes at least one characteristic of the registered input events and a prediction for performing the predefined task. The prediction is generated by applying the registered input events to a predictive model.