Understanding User Behavior in Window Selection using Dragging for Multiple Targets

Research Background and Problem Statement

• Challenges and Issues:
  Although window selection is a fundamental multi-target interaction method in graphical user interfaces (GUIs), existing research primarily focuses on modeling single-target dragging behavior, with limited studies on drag-based multi-target selection behavior. Window selection involves multiple parameters, such as target size, target spacing, target quantity, and target layout, whose effects on user interaction behavior have not been clearly quantified.

• Significance:
  In practical desktop environments, applications such as file managers, spreadsheets, or illustration tools frequently involve window selection. Since it entails multi-target selection, optimizing this interaction can significantly enhance user efficiency and experience.

• Research Motivation and Related Work:
  This study is based on human motor models (e.g., Fitts' Law and the Steering Law) to investigate whether these models can effectively describe drag-based multi-target window selection behavior. Moreover, considering that dragging is a common fundamental operation in GUIs, studying the impact of parameters related to window selection in actual GUIs on user behavior holds significant value.

Solution

• Methods and Innovations:
  The authors propose a new model that integrates existing models (such as Fitts' Law and the Steering Law) to study user behavior during dragging. By introducing task parameters such as target width (W), target spacing (I), target quantity (N), and layout (L), an in-depth experimental analysis of user behavior was conducted.

• Innovations:

  • Extended the parameter settings of traditional Fitts' Law and the Steering Law, adjusting the modeling approach to suit dragging behavior in window selection.
  • Proposed new behavioral metrics (e.g., Falling Rate, FR) to analyze the distribution of release points and explore shifts in user behavior.
  • Developed a new predictive model capable of accurately forecasting performance time and user behavior based on key parameters.

• Implementation Steps:

  1. Experimental Design:
     • Conducted multiple experiments to study the effects of target spacing, quantity, layout, and dragging characteristics on user performance.
  2. Behavioral Modeling:
     • Developed and compared various models, including extended versions of traditional Fitts' Law, the Steering Law, and the Targeted Steering Model.
     • Proposed a refined model to describe time performance in window selection, incorporating target parameters and constraints on the dragging path.
  3. Model Validation:
     • Validated the robustness of the model in partially constrained real-world GUI scenarios, including tasks with obstacles.

Research Findings

• Experimental Results:

  • Target spacing (I) and target quantity (N) significantly affect completion time (Movement Time, MT, or Dragging Time, DT), while target width (W) has no significant impact on performance.
  • For larger spacing (I), user behavior resembles pointing or crossing actions; for smaller spacing, it manifests as steep path control (Steering operations).
  • Model validation demonstrated that the comprehensive Targeted Steering Model is optimal for multi-target selection dragging operations.

• Model Performance:

  • The proposed refined model performed exceptionally well in fitting user behavior, achieving an adjusted coefficient of determination (R²_adjusted) of 0.82 (DT) and 0.95 (MT).
  • The model maintained excellent predictive capability even in partially constrained scenarios, indicating its adaptability to complex real-world work environments.

• Advantages:

  • Compared to traditional models, the new model exhibits superior predictive capability in complex user behavior scenarios involving multi-target selection.
  • Provides practical guidance for GUI designers to optimize icon layouts and interaction methods.

• Limitations and Future Directions:

  • The study's task scenarios are limited to desktop environments and have not been extended to mobile devices (e.g., touch or stylus input).
  • The selection criterion was fixed as "partial overlap of the target point is sufficient for selection"; future studies could explore other selection methods (e.g., full coverage).
  • The range of some experimental parameters, such as target width/spacing values, remains relatively narrow, potentially limiting the generalizability of the findings.
  • The experimental design did not consider dynamic targets or complex visual encoding environments, such as graphically dense gaming scenarios.

Conclusion

This study makes significant contributions to understanding drag-based window selection behavior by analyzing and validating a comprehensive performance prediction model. These findings not only enrich research on human motor models in HCI but also provide practical guidance for optimizing GUI design and improving efficiency. Future work could extend the research to more operational devices and complex scenarios to verify the broad applicability of the model.