Bivariate Effective Width Method to Improve the Normalization Capability for Subjective Speed-accuracy Biases in Rectangular-target Pointing

Title of the Paper

Bivariate Effective Width Method to Improve the Normalization Capability for Subjective Speed-accuracy Biases in Rectangular-target Pointing

Paper Information

• Subject Area: Human-Computer Interaction (HCI), Performance Evaluation of User Interfaces
• Keywords: Fitts' Law, Pointing Tasks, Graphical User Interface, Human Motor Performance, Crowdsourced Study

Research Background and Problem

• Identified Problem or Challenge: In current user interface performance evaluations, traditional Fitts' Law is primarily applied to one-dimensional strip targets or two-dimensional circular targets, making it unsuitable for the more common rectangular targets. Researchers have yet to explore how to integrate effective height (He) into the normalization of speed-accuracy trade-offs for rectangular targets.
• Significance: Rectangular targets are more prevalent in real-world graphical user interfaces (e.g., desktop and mobile applications). Additionally, since different input devices exhibit varying operational precision in different directions, developing effective methods to enhance external validity (i.e., simulating more realistic tasks) is of great importance.
• Research Motivation and Related Work:
  • Traditional Fitts' Law predicts movement time (MT) based on target width (W) and distance (D) and improves the normalization of speed-accuracy biases through "effective width" (We).
  • For rectangular targets, previous studies have proposed extensions but have not sufficiently validated the impact of He on speed-accuracy biases.
  • This study aims to construct a bivariate effective width method by integrating We and He to improve the normalization of speed-accuracy trade-offs for complex target shapes.

Solution

• Proposed Solution:
  • Propose the "Bivariate Effective Width Method," introducing effective height (He) into the Fitts' Law formula.
  • Use a data-driven approach to validate the normalization capability of We and He for speed-accuracy biases in rectangular targets through experiments.
• Innovations:
  • Integrate the independent applications of We and He into existing Fitts' Law models (including the weighted Euclidean model by Accot and Zhai).
  • Achieve normalization across one-dimensional and two-dimensional tasks while enhancing the model's adaptability to real GUI scenarios.
• Implementation Steps and Key Techniques:
  • Experimental Design:
    • Two experiments (remote and crowdsourced) were conducted to validate participants' adherence to speed-accuracy instructions and the model's normalization capability.
    • Independent variables included three bias conditions (accuracy-priority, neutral, speed-priority), target distance (D), target width (W), and target height (H).
  • Statistical Analysis and Methods:
    • Regression models were used to evaluate the fit (Adjusted R² and AIC) of movement time (MT) with different target dimensions.
    • Comparative analysis of the unified performance metric "Throughput" (TP) predicted by the models.
  • Experimental Tools: A custom-developed experimental system was used to record click position distributions (SDx and SDy) and movement time (MT).

Research Findings

• Specific Findings:
  • Improved Model Fit:
    • The Accot and Zhai model using effective dimensions (We and He) achieved the highest fit under mixed conditions, with adjusted R² and AIC values outperforming traditional models based on nominal dimensions (W and H).
    • The effective method significantly reduced throughput differences across the three bias conditions.
  • Experimental Comparisons:
    • The remote experiment supported the superior normalization capability of the Accot and Zhai model.
    • The crowdsourced experiment further confirmed that the introduction of effective height significantly improved the model's adaptability and validated its stability with a larger sample size.
• Advantages:
  • Provides a method to normalize speed-accuracy biases, facilitating fair comparisons across different user groups, devices, or interaction techniques.
  • The experimental design supports the application of this method to more realistic graphical user interface scenarios.
• Limitations and Future Directions:
  • Limitations:
    • Complex target scenarios (e.g., distractors, varying target angles) were not tested.
    • The diversity of devices in remote and crowdsourced experiments may introduce uncontrollable variables.
  • Future Directions:
    • Further validation of research findings in controlled laboratory environments.
    • Extension to other interaction tasks such as touch or drag-and-drop, and exploration of applications in eye-tracking or object selection tasks.
    • Investigate the impact of target approach angle (θ) on model performance and optimization.

By proposing and experimentally validating the "Bivariate Effective Width Method," this study provides a novel solution for normalizing rectangular target pointing tasks and makes a substantial contribution to research methodologies in the HCI field.