[Paper Review] Calibration Methods of Touch-Point Ambiguity for Finger-Fitts Law.

Shota Yamanaka, Hiroki Usuba|arXiv (Cornell University)|Jan 13, 2021

Tactile and Sensory Interactions1 citations

TL;DR

This study evaluates calibration methods for the finger tremor factor (σₐ) in Finger-Fitts Law, analyzing 1D and 2D touch-pointing experiments and reanalyzing prior data. It finds no optimal method for measuring σₐ that universally maximizes prediction accuracy, challenging the consensus on best practices for parameter calibration in touch-based pointing tasks.

ABSTRACT

Finger-Fitts law (FFitts law) is a model to predict touch-pointing times that was modified from Fitts' law. It considers the absolute touch-point precision, or a finger tremor factor sigma_a, to decrease the admissible target area and thus increase the task difficulty. Among choices such as running an independent task or performing parameter optimization, there is no consensus on the best methodology to measure sigma_a. By integrating the results of our 1D and 2D touch-pointing experiments and reanalyses of previous studies' data, we examined the advantages and disadvantages of each approach to compute sigma_a, and we found that there is no optimal choice to maximize the prediction accuracy of FFitts law.

Motivation & Objective

• To investigate the impact of different calibration methods for the finger tremor factor σₐ on the prediction accuracy of Finger-Fitts Law.
• To resolve the lack of consensus on whether to use independent tasks or parameter optimization for measuring σₐ.
• To assess the reliability and validity of existing approaches for estimating σₐ using empirical touch-pointing data.
• To determine whether any single method consistently outperforms others in predicting touch-pointing times.

Proposed method

• Conducted controlled 1D and 2D touch-pointing experiments to collect empirical data on finger tremor and pointing performance.
• Reanalyzed data from previous studies to compare consistency and accuracy across different calibration techniques.
• Evaluated multiple approaches to estimate σₐ, including independent measurement tasks and parameter optimization within the FFitts model.
• Used statistical modeling to compare prediction accuracy of FFitts Law across different σₐ estimation methods.
• Applied model fitting techniques to assess how well each method predicts observed pointing times.
• Assessed the trade-offs between measurement complexity, data requirements, and predictive performance of each method.

Experimental results

Research questions

• RQ1 Which calibration method for σₐ yields the highest prediction accuracy in Finger-Fitts Law across different touch-pointing tasks?
• RQ2 How do independent measurement tasks compare to parameter optimization in estimating the finger tremor factor?
• RQ3 Is there a universally optimal method for measuring σₐ that outperforms others across diverse experimental conditions?
• RQ4 To what extent does the choice of σₐ calibration method affect the reliability of FFitts Law predictions?

Key findings

• No single calibration method for the finger tremor factor σₐ consistently maximized prediction accuracy across all experimental conditions.
• Both independent tasks and parameter optimization methods showed limitations in reliably estimating σₐ with high predictive fidelity.
• The reanalysis of prior datasets revealed significant variability in σₐ estimates depending on the chosen calibration approach.
• The lack of a universally optimal method suggests that current practices for σₐ calibration may need reevaluation in applied human-computer interaction research.
• The study highlights the inherent ambiguity in defining and measuring touch-point precision, which undermines the consistency of FFitts Law predictions.
• The results indicate that the choice of calibration method significantly influences model performance, but no method emerged as clearly superior.