We investigated whether participants’ perception of waiting time would be affected by progress-bar thickness. Specifically, we prepared nine progress bars with different thicknesses (i.e., heights of 2, 5, 10, 25, 50, 100, 150, 200, and 400 [px]) with the length fixed at 400 [px] and the duration fixed at 5.0 [s]. As a result, at heights below 25 [px], the thinner the bar, the shorter the waiting time was perceived to be, and at heights of 50 [px] or more, there were no significant differences in the participants’ perception of waiting time.

Throughput is a widely used performance metric, combining speed and accuracy into a single measure, while reducing the effect of subjective speed–accuracy trade-offs. Despite its wide application in 2D steering tasks, its direct extension to 3D presents unique challenges since 3D trajectories exhibit higher variability, and perceptual–motor factors undermine existing formulations. Consequently, throughput has not been systematically adopted for evaluating steering in 3D virtual environments. In this paper, using a controlled virtual reality user study with a ring-and-wire task, we introduce and validate a novel throughput formulation for 3D steering based on the bivariate standard deviation of the trajectory for the effective width calculation. Our results show that this formulation provides smoother throughput values across subjective speed–accuracy differences and improves model fit compared to traditional approaches. This work advances our theoretical understanding of the Steering law in 3D contexts, provides researchers and practitioners with a robust evaluation method, and establishes a foundation for future studies of complex 3D trajectory interactions.

Typical success-rate prediction models for tapping exclude targets near screen edges; however, design constraints often force such placements. Additionally, in scrollable UIs any element can move close to an edge. In this work, we model how target–edge distance affects 1D touch pointing accuracy. We propose the Skewed Dual Normal Distribution Model, which assumes the tap coordinate distribution is skewed by a nearby edge. The results of two smartphone experiments showed that, as targets approached the edge, the distribution’s peak shifted toward the edge and its tail extended away. In contrast to prior reports, the success rate improved when the target touched the edge, suggesting a strategy of “tapping the target together with the edge.” By accounting for skew, our model predicts success rates across a wide range of conditions, including edge‑adjacent targets, thus extending coverage to the whole screen and informing UI design support tools.

In crowdsourced user experiments that collect performance data from graphical user interface (GUI) interactions, some participants ignore instructions or act carelessly, threatening the validity of performance models. We investigate a pre-task screening method that requires simple GUI operations analogous to the main task and uses the resulting error as a continuous quality signal. Our pre-task is a brief image-resizing task in which workers match an on-screen card to a physical card; workers whose resizing error exceeds a threshold are excluded from the main experiment. The main task is a standardized pointing experiment with well-established models of movement time and error rate. Across mouse- and smartphone-based crowdsourced experiments, we show that reducing the proportion of workers exhibiting unexpected behavior and tightening the pre-task threshold systematically improve the goodness of fit and predictive accuracy of GUI performance models, demonstrating that brief pre-task screening can enhance data quality.

For evaluations of 2D target selection using Fitts’ law, ISO 9241-411 recommends using the effective target width (\(W_\text{e}\)) calculated using the univariate standard deviation of selection coordinates. Related research proposed using a bivariate standard deviation; however, the proposal was only tested using a single speed-accuracy bias condition, thus the assessment was limited. We compared the univariate and bivariate techniques in a 2D Fitts’ law experiment using three speed-accuracy biases and 346 crowdworkers. Calculating \(W_\text{e}\) using the univariate standard deviation yielded higher model correlations across all bias conditions and produced more stable throughput among the biases. The findings were also consistent in cases using randomly sampled subsets of the participant data. We recommend that future research should calculate \(W_\text{e}\) using the univariate standard deviation for fair performance evaluations. Also, we found trivial effects when using nominal or effective amplitude and using different perspectives of the task axis.

Steering involves continuous movement along constrained paths, well-studied in 2D. The extensions to 3D using the Ring-and-Wire and Ball-and-Tunnel tasks were often treated as interchangeable in previous work. In this paper, we directly compare these two tasks through a within-subjects user study (n = 18) with varying 3D path orientations. The results show that Ring-and-Wire significantly outperformed Ball-and-Tunnel, with 17.17% lower task time, 21.65% higher throughput, and 21.52% faster average speed. Participants also preferred Ring-and-Wire and reported lower workload. Visual ambiguity, especially near the tunnel’s rear surface, complicated spatial perception in the Ball-and-Tunnel task. We thus recommend that future studies choose 3D steering tasks carefully for experiments, as the two tasks are not interchangeable.

This study investigated the use of the deformation of the fingertips of a virtual hand to enhance depth perception during freehand interaction in a virtual reality (VR). Artificial fingertip deformation may generate mapping of real hand position and pseudo-haptics, improving UI usability. We conducted two experiments focusing on depth manipulation in both pointing and steering tasks. Our results revealed that changes in fingertip shape reduced operation time in pointing tasks and improved accuracy in steering tasks. Additionally, we conducted subjective evaluation surveys for both experiments, which showed improvements in pseudo-haptics, spatial perception, and user experience. Based on these results, we propose several applications and demonstrate that fingertip deformations in virtual hands can contribute to better 3D UI design.

Human-computer interaction studies evaluate the effectiveness of the user interface (UI) elements using either in-vitro as controlled situation or in-vivo experiments in realistic situation. Although both methods are used for comprehending complementary and multifaceted insights into the target in biology, most HCI studies usually corroborate in-vitro results with in-vivo ones. In this paper, we reported a concrete example that the results of in-vitro and in-vivo experiments were different and it was only achieved by conducting both experiments that these reasons of these discrepancies were captured. Specifically, we focused on a throbber as a target UI element and investigated how the different attributes of throbbers affect the users’ waiting time perception. The results of four in-vitro experiments showed that the throbber with slower rotational velocities was significantly perceived as being shorter. However, the results of three in-vivo experiments showed that the experimental sites displaying slower throbbers during the waiting time were significantly perceived as being longer. In fact, we could figure out that such discrepancy would be caused by users’ everyday interpretations like “A slower throbber means the network speed is slow” that could not be assumed in in-vitro experiments. We believe that even a seemingly negative result, such as different results obtained in-vitro and in-vivo studies, can deepen the knowledge of the target UI elements. We then strongly argue that a complementary comparison of the UI elements with the findings in in-vitro and in-vivo is very important for HCI researchers.

Two formulations of Fitts’ index of difficulty ID are empirically compared under different subjective speed-accuracy biases: the nominal form IDn and the effective form IDe using endpoint variability. The effective forms have typically been considered beneficial for capturing the actual accuracy of users’ performance, while the nominal form is better for single-biased data. In our analysis of the data from 210 crowdworkers, the best model tended to switch. At times, this switch was statistically significant, especially when limited portions of the entire workers and trials were used, such as the first eight clicks (out of 16) performed by 20 workers who were randomly sampled from a comprehensive group of 210 participants. Our findings caution against assuming a model’s capability based on only a few experiments using a limited number of participants or just a few trials. They also emphasize the need for performing replications on even well-investigated models.

Null hypothesis significance testing (NHST) is widely used in the field of human-computer interaction (HCI), and caution has been advised against deciding whether to stop or continue an experiment based on checking p-values. However, in clinical trials, NHST with interim analysis is commonly employed by adjusting the α levels to control Type I error rates. This approach, known as a group sequential design, can also be beneficial in the HCI field. We analyzed existing datasets of target-pointing tasks using group sequential designs and experimentally demonstrated that, depending on the α-level correction method, we could reduce the number of participants by 40.4–56.5%. Therefore, employing group sequential designs enables significant time and cost savings for both participants and researchers in the HCI field.

Prior work has shown that skills acquired through repeated computer-based tasks can transfer to similar tasks. While typical experiments span only up to several days, it remains unclear whether such transfer effects would be evident over longer learning periods. To address this question, we measured the target-tapping performance of 912 crowdworkers and analyzed its relationship with their usage logs of Yahoo! JAPAN services. Our analysis included data of unprecedented scale, featuring users holding Yahoo! JAPAN accounts for over 24 years, performing over 100,000 taps annually, or viewing over 180,000 pages annually (with at least ten users in each category). Results showed that a longer account age was correlated with improved tap performance, whereas a higher number of taps over the past year was negatively correlated with tap performance. More in-depth analyses also revealed that the performance trends varied depending on the specific services that were heavily used. These findings clarify the limitations of a naive assumption that long-term and heavy use of web services leads to better task performance through transfer effects.

In Fitts’ law studies to investigate pointing, throughput is used to characterize the performance of input devices and users, which is claimed to be independent of task difficulty or the user’s subjective speed-accuracy bias. While throughput has been recognized as a useful metric for target-pointing tasks, the corresponding formulation for path-steering tasks and its evaluation have not been thoroughly examined in the past. In this paper, we conducted three experiments using linear, circular, and sine-wave path shapes to propose and investigate a novel formulation for the effective parameters and the throughput of steering tasks. Our results show that the effective width substantially improves the fit to data with mixed speed-accuracy biases for all task shapes. Effective width also smoothed out the throughput across all biases, while the usefulness of the effective amplitude depended on the task shape. Our study thus advances the understanding of user performance in trajectory-based tasks.

In current graphical user interfaces, there exists a (typically unavoidable) end-to-end latency from each pointing-device movement to its corresponding cursor response on the screen, which is known to affect user performance in target selection, e.g., in terms of movement time (MT). Previous work also reported that a long latency increases MTs in path-steering tasks, but the quantitative relationship between latency and MT had not been previously investigated for path-steering. In this work, we derive models to predict MTs for path-steering and evaluate them with five tasks: goal crossing as a preliminary task for model derivation, linear-path steering, circular-path steering, narrowing-path steering, and steering with target pointing. The results show that the proposed models yielded an adjusted R2 > 0.94, with lower AICs and smaller cross-validation RMSEs than the baseline models, enabling more accurate prediction of MTs.

Existing guidelines for designing targets on smartphones often focus on single-tap operations for accurate selection. However, smartphone interfaces can support both tap and swipe actions. We explored user-performance differences between tap and swipe in two crowdsourced experiments using bar and square targets. Results indicated longer operation times, higher error rates, and significantly shifted touch points for swipe compared to tap. Our findings imply that current target-size guidelines may not apply to swipe-operated targets, and they reveal new research opportunities for swipeable-target designs.

The Steering Law is a robust model to predict the movement time (MT) for steering through a constrained path, and the most representative example in human-computer interaction (HCI) is navigating cascaded menus. In typical implementations of cascaded menus, however, users can deviate from the path for a short time; we call this error-accepting delay, or Tdelay. Yamanaka modified the Steering Law to predict MT under several Tdelay conditions, and our goal is to investigate the reproducibility of his model with more various Tdelay values. In addition, HCI researchers have recently formed a consensus that the goodness of models should be judged by the prediction accuracy for future (untested) task conditions. Thus, for the sake of completeness, we conducted two analyses: a shuffle-split cross-validation and leave-one-Tdelay-out cross-validation. The results showed that, regardless of the all-data and cross-validation analyses, Yamanaka’s modified model outperformed the baseline Steering Law, which strengthened his original experimental report.

Steering a cursor through a constrained path is required for operating graphical user interfaces, such as when navigating a cascaded menu. Recently, the accuracy with which users successfully accomplish a task is emerging as an important performance indicator. In this study, we evaluated the performance of a Servo-Gaussian model to predict the success rates in linear and circular paths with 212 and 166 crowdsourced participants, respectively. The results showed that, for linear paths, the model achieved r2=0.9676, MAE=2.036%, and RMSE=2.692%, and for circular paths, it achieved r2=0.9787, MAE=3.199%, and RMSE=3.927%. Shuffle-split cross-validation with five train-test data-size ratios also demonstrated the robust prediction accuracy of the model. These findings will provide designers with a useful tool to judge if an interface can be accurately operated without running costly user studies for new task conditions.

The Dual Gaussian Distribution Model can be utilized for predicting the success rates of tapping targets. However, previous studies have shown that the prediction error increases to as much as 10 points, where "points" represent the percentage difference between the observed and predicted values of the tap success rate, particularly for a small target width W such as 2 mm. We hypothesize that this could be due to the experimental designs with sparse W levels performed by few participants, rather than the model itself. Our experiment involving horizontal and vertical bar targets with W = 2-8 mm (step: 0.2 mm) performed by more than 180 participants showed that the maximum prediction errors were relatively small: 2.769 and 3.185 points, respectively. Furthermore, the correlation between W and the prediction error was statistically small (Pearson's |r| < 0.2), and W was not a significant contributor to changing prediction errors (p>0.05). As these results do not support the concerns that the Dual Gaussian Distribution Model has an issue when used with small targets, the development of applications and refined models is encouraged to continue.

Previous studies on the Finger-Fitts law (FFitts law) are lacking in sufficient experiments to verify its inherent potential. Since the FFitts law is originally a modified version of the effective width method to normalize speed-accuracy biases, the model fit would improve if multiple biases were mixed together and the throughputs would be more stable than using the nominal target width. In this study, we conduct an experiment in which participants tap 1D-bar and 2D-circular targets under three subjective biases: balancing the speed and accuracy, emphasizing speed, and emphasizing accuracy when they perform the tasks. The results showed that applying the effective width to Ko et al.'s refined FFitts law, which represents the touch ambiguity with a free parameter, was the most successful in normalizing biases. Reanalyzing another dataset on ray-casting pointing also led to the same conclusion. We thus recommend using Ko et al.'s model with effective width when researchers compare several experimental conditions such as devices and user groups.

In typical experiments to evaluate novel pointing-facilitation techniques, participants are asked to perform a task as rapidly and accurately as possible. However, the balance can differ among participants, and the techniques’ effectiveness would change if the majority of participants give weight to either speed or accuracy. We investigated the effects of three subjective biases (emphasizing speed, neutral, and emphasizing accuracy) on the evaluation results of pointing-facilitation techniques, namely Bubble Cursor and Bayesian Touch Criterion (BTC). The results indicate that Bubble Cursor outperformed the baseline in terms of movement time and error rate under all bias conditions, while BTC underperformed a simpler target-prediction technique, which was an inconsistent outcome to the original study. Examining multiple biases enables researchers to discuss the (dis)advantages of novel or existing techniques more precisely, which can be beneficial to reach a more reliable conclusion.

Error rates (ERs) in target-pointing tasks are typically modelled in two steps: predicting the click-point variability (σ) based on target sizes and then computing the probability that a click falls outside a target. This is an indirect approach if the researcher’s purpose is to achieve the accurate prediction of ERs because the model coefficients are optimized to predict σ accurately in the first step. We compared the prediction accuracies of this method with a more direct technique in which the coefficients used for σ are determined in such a way as to optimize the closeness between observed and predicted ERs. Our re-analysis of eight datasets from mouse- and touch-based pointing studies showed that the latter approach consistently outperforms the conventional one if the starting values for the parameter search are appropriate (which can be achieved by hyperparameter optimization), thus enabling the interface configuration on the basis of accurately predicted ERs.

Studies on gaze-based interactions have utilized natural eye-related information to detect user intent. Most use a machine learning-based approach to minimize the cost of choosing appropriate eye-related information. While those studies demonstrated the effectiveness of an intent detection system, understanding which eye-related information is useful for interactions is important. In this paper, we reanalyze how eye-related information affected the detection performance of a previous study to develop better intent detection systems in the future. Specifically, we analyzed two aspects of dimensionality reduction and adaptation to different tasks. The results showed that saccade and fixation are not always useful, and the direction of gaze movement could potentially cause overfitting.

Touch surfaces are widely utilized for smartphones, tablet PCs, and laptops (touchpad), and single and double taps are the most basic and common operations on them. The detection of single or double taps causes the single-tap latency problem, which creates a bottleneck in terms of the sensitivity of touch inputs. To reduce the single-tap latency, we propose a novel machine-learning-based tap prediction method called PredicTaps. Our method predicts whether a detected tap is a single tap or the first contact of a double tap without having to wait for the hundreds of milliseconds conventionally required. We present three evaluations and one user evaluation that demonstrate its broad applicability and usability for various tap situations on two form factors (touchpad and smartphone). The results showed PredicTaps reduces the single-tap latency from 150--500 ms to 12 ms on laptops and to 17.6 ms on smartphones without reducing usability.

A prior work has recommended adding a 4-mm gap between a target and the edge of a screen, as tapping a target located at the screen edge takes longer than tapping non-edge targets. However, it is possible that this recommendation was created based on statistical errors, and unexplored situations existed in the prior work. In this study, we re-examine the recommendation by utilizing crowdsourced experiments to resolve the issues. If we observe the same results as the prior work through experiments including diversities, we can verify that the recommendation is suitable. We found that increasing the gap between the target and the screen edge decreased the movement time, which was consistent with the prior work. In addition, we newly found that increasing the gap decreased the error rate as well. On the basis of these results, we discuss how the gap and the target should be designed.

Evaluating the validity of an existing user performance model in a variety of tasks is important for enhancing its applicability. The model studied in this work is the steering law for predicting the speed and time needed to perform tasks in which a cursor or a car passes through a constrained path. Previous HCI studies have refined this model to take additional path factors into account, but its applicability has only been evaluated in GUI-based environments such as those using mice or pen tablets. Accordingly, we conducted a user experiment with a driving simulator to measure the speed and time on curved roads and thus facilitate evaluation of models for pen-based path-steering tasks. The results showed that the best-fit models for speed and time had adjusted r^2 values of 0.9342 and 0.9723, respectively, for three road widths and eight curvature radii. While the models required some adjustments, the overall components of the tested models were consistent with those in previous pen-based experimental results. Our results demonstrated that user experiments to validate potential models based on pen-based tasks are effective as a pilot approach for driving tasks with more complex road conditions.

Target-selection tasks have been frequently conducted to evaluate novel input devices and pointing-facilitation techniques. Recently, Sharif et al. showed that a unified metric to evaluate the pointing performance, called throughput TP, was not stable across two sessions performed by the same participant group, which indicates poor test-retest reliability. Because there are cases in which using TP is inappropriate depending on the research topic, we extend their finding to two other metrics: movement time MT and error rate ER. We demonstrated that, even for the participants who kept their TPs across two sessions stable, they would exhibit unstable MTs and ERs. Thus, if time allows, researchers should design their experiments to run multiple sessions for obtaining the central tendency of user performance, which increases the validity of their user studies.

The effective width method of Fitts’ law can normalize speed-accuracy biases in 1D target pointing tasks. However, in graphical user interfaces, more meaningful target shapes are rectangular. To empirically determine the best way to normalize the subjective biases, we ran remote and crowdsourced user experiments with three speed-accuracy instructions. We propose to normalize the speed-accuracy biases by applying the effective sizes to existing Fitts’ law formulations including width W and height H. We call this target-size adjustment the bivariate effective width method. We found that, overall, Accot and Zhai’s weighted Euclidean model using the effective width and height independently showed the best fit to the data in which the three instruction conditions were mixed (i.e., the time data measured in all instructions were analyzed with a single regression expression). Our approach enables researchers to fairly compare two or more conditions (e.g., devices, input techniques, user groups) with the normalized throughputs.

In this paper, we first position the current dwell selection among gaze-based interactions and its advantages against head-gaze selection, which is the mainstream interface for HMDs. Next, we show how dwell selection and head-gaze selection are used in an actual interaction situation. By comparing these two selection methods, we describe the potential of dwell selection as an essential AR/VR interaction.

We propose a method that predicts the success rate in pointing to 2D rectangular targets by using 1D vertical-bar and horizontal-bar task results. The method can predict the success rates for more practical situations under fewer experimental conditions. This shortens the duration of experiments, thus saving costs for researchers and practitioners. We verified the method through two experiments: laboratory-based and crowdsourced ones. In the laboratory-based experiment, we found that using 1D task results to predict the success rate for 2D targets slightly decreases the prediction accuracy. In the crowdsourced experiment, this method scored better than using 2D task results. Thus, we recommend that researchers use the method properly depending on the situation.

Models of lassoing time to select multiple square icons exist, but realistic lasso tasks also typically involve encircling non-rectangular objects. Thus, it is unclear if we can apply existing models to such conditions where, e.g., the width of the path that users want to steer through changes dynamically or step-wise. In this work, we conducted two experiments where the objects were non-rectangular, with path widths that narrowed or widened, smoothly or step-wise. The results showed that the baseline models for pen-steering movements (the steering and crossing law models) fitted the timing data well, but also that segmenting width-changing areas led to significant improvements. Our work enables the modeling of novel UIs requiring continuous strokes, e.g., for grouping icons.

We investigated the relationship between dwell-time and the model human processor (MHP). First, we devised an equation that can represent the time taken for recognizing an image based on MHP. Then, we evaluated whether the equation can represent the time and wheter the time estimated by the equation matches the user’s preferred dwell-time. The experiment consisted of two tasks: image selection with a button (button-task) and image selection with a dwell (dwell-task). From the results of the button-task, we found that the equation derived by MHP can estimate the time; the time taken for button selection was 662 ms on average, and the time estimated by the equation was 660 ms on average. Also, we showed that the estimated time represented the user’s preferred dwell-time; all participants in the experiment answered that 500 ms and 600 ms were their preferred dwell-times.

The usage of crowdsourcing to recruit numerous participants has been recognized as beneficial in the human-computer interaction (HCI) field, such as for designing user interfaces and validating user performance models. In this work, we investigate its effectiveness for evaluating an error-rate prediction model in target pointing tasks. In contrast to models for operational times, a clicking error (i.e., missing a target) occurs by chance at a certain probability, e.g., 5%. Therefore, in traditional laboratory-based experiments, a lot of repetitions are needed to measure the central tendency of error rates. We hypothesize that recruiting many workers would enable us to keep the number of repetitions per worker much smaller. We collected data from 384 workers and found that existing models on operational time and error rate showed good fits (both R^2 > 0.95). A simulation where we changed the number of participants N_P and the number of repetitions N_repeat showed that the time prediction model was robust against small N_P and N_repeat, although the error-rate model fitness was considerably degraded. These findings empirically demonstrate a new utility of crowdsourced user experiments for collecting numerous participants, which should be of great use to HCI researchers for their evaluation studies.

In touch interfaces, a target, such as an icon, has two widths: the visual width and the touchable width. The visual width is the target's appearance, and the touchable width is the area in which users can touch a target and execute an action. In this study, we conduct two experiments to investigate the effects of the visual and touchable widths on touch pointing performance (movement time and success rate). Based on the results, we build candidate models for predicting the movement time and compare them by the values of adjusted R^2 and AIC. In addition, we build a success rate model and test it through cross-validation. Existing models can be applied only to situations where the visual and touchable widths are equal, and we show that our refined model achieves better model fitness, even when such widths are different. We also discuss the design implications of the touch interfaces based on our models.

We demonstrate a gaze-based command activation technique that is robust against unintentional command activations using a series of dwelling on a target and performing a specific gesture (dwell-thengesture manipulation). The gesture adopted is a simple two-level stroke, which consists of a sequence of two orthogonal strokes. To achieve robustness against unintentional command activations, we designed and fine-tuned a gesture detection system based on how users move their gaze, as revealed through three experiments. Although our technique seems to simply combine well-known dwelland gesture-based manipulations, implying a low rate of success, our technique is actually the first technique that consists of a short time dwelling for target selection and a simple gesture for command activation. In addition, our technique will be the first technique adopting a marking menu, which is a traditional menu for command activation used in mouseor pen-based interactions to gaze-based interactions.

For hover-based UIs (e.g., pop-up windows) and scrollable UIs, we investigated mouse-pointing performance for users trying to avoid overshooting a target while aiming for it. Three experiments were conducted with a 1D pointing task in which overshooting was accepted (a) within a temporal delay, (b) via a spatial gap between the target and an unintended item, and (c) with both a delay and a gap. We found that, in general, movement times tended to increase with a shorter delay and a smaller gap if these parameters were independently tested. Therefore, Fitts' law cannot accurately predict the movement times when various values of delay and/or gap are used. We found that 800 ms is required to remove negative effects of distractor for densely arranged targets, but we found no optimal gap.

Personal identification numbers (PINs) and grid patterns have been used for user authentication, such as for unlocking smartphones. However, they carry the risk that attackers will learn the PINs and patterns by shoulder surfing. We propose a secure authentication method called SheetKey that requires complicated and quick touch inputs that can only be accomplished with a sheet that has a pattern printed with conductive ink. Using SheetKey, users can input a complicated combination of touch events within 0.3 s by just swiping the pad of their finger on the sheet. We investigated the requirements for producing SheetKeys, e.g., the optimal disc diameter for generating touch events. In a user study, 13 participants passed through authentication by using SheetKeys at success rates of 78-87%, while attackers using manual inputs had success rates of 0-27%. We also discuss the degree of complexity based on entropy and further improvements, e.g., entering passwords on alphabetical keyboards.

The steering law is a model for predicting the time and speed for passing through a constrained path. When people can view only a limited range of the path forward, they limit their speed in preparation of possibly needing to turn at a corner. However, few studies have focused on how limited views affect steering performance, and no quantitative models have been established. The results of a mouse steering study showed that speed was linearly limited by the path width and was limited by the square root of the viewable forward distance. While a baseline model showed an adjusted R2 = 0.144 for predicting the speed, our best-fit model showed an adjusted R2 = 0.975 with only one additional coefficient, demonstrating a comparatively high prediction accuracy for given viewable forward distances.

We propose a Servo-Gaussian model to predict success rates in continuous manual tracking tasks. Two tasks were conducted to validate this model: path steering and pursuit of a 1D moving target. We hypothesized that (1) hand movements follow the servo-mechanism model, (2) submovement endpoints form a bivariate Gaussian distribution, thus enabling us to predict the success rate at which a submovement endpoint falls inside the tolerance, and (3) the success rate for a whole trial can be predicted if the number of submovements is known. The cross-validation showed R^2>0.92 and MAE<4.9% for steering and R^2>0.95 and MAE<6.5% for pursuit tasks. These results demonstrate that our proposed model delivers high prediction accuracy even for unknown datasets.

The dual Gaussian distribution hypothesis has been used to predict the success rate of target pointing on touchscreens. Bi and Zhai evaluated their success-rate prediction model in off-screen-start pointing tasks. However, we found that their prediction model could also be used for on-screen-start pointing tasks. We discuss the reasons why and empirically validate our hypothesis in a series of four experiments with various target sizes and distances. The prediction accuracy of Bi and Zhai's model was high in all of the experiments, with a 10-point absolute (or 14.9% relative) prediction error at worst. Also, we show that there is no clear benefit to integrating the target distance when predicting the endpoint variability and success rate.

The “crossing time” to pass between objects in lassoing tasks is predicted by Fitts’ law. When an unwanted object, or obstacle , intrudes into the user’s path, users curve the stroke to avoid hitting that obstacle. We empirically show that, in the presence of an obstacle, modified Fitts models for pointing with obstacle avoidance can significantly improve the prediction accuracy of movement time compared with standard Fitts’ law. Yet, we also found that when an object is (only) close to the crossing path, i.e., a distractor , users still curve their stroke, even though the object does not intrude. We tested the effects of distractor proximity and length. While the crossing motion is modified by a nearby distractor, our results also identify that overall its effect on crossing times was small, and thus Fitts’ law can still be applied safely with distractors.

In steering law tasks, deviating from the path is immediately considered an error operation. However, in navigating a hierarchical menu item, which is a representative application of the law, a deviation within a short duration is sometimes permitted. We tested the validity of the steering law model with various durations of such error-accepting delays and found that it showed high fits for each delay condition (R2 > 0.96) but poor fits if the delay values were not separated (R2 = 0.58). Because the average movement speed linearly increased as the delay increased, we refined the model by taking the delay into account, and the fitness was significantly improved (R2 = 0.97). Our model will help GUI designers estimate the average operational time on the basis of the menu item length, width, and error-accepting delay.

Lassoing objects is a basic function in illustration software and presentation tools. Yet, for many common object arrangements lassoing is sometimes time-consuming to perform and requires precise pen operation. In this work, we studied lassoing movements in a grid of objects similar to icons. We propose a quantitative model to predict the time to lasso such objects depending on the margins between icons, their sizes, and layout, which all affect the number of stopping and crossing movements. Results of two experiments showed that our models predict fully and partially constrained movements with high accuracy. We also analyzed the speed profiles and pen stroke trajectories and identified deeper insights into user behaviors, such as that an unconstrained area can induce higher movement speeds even in preceding path segments.

When users operate smartphones and desktop interfaces with their fingers, there are differences between the motor and visual widths. For example, when a user selects an item from a vertical menu, the area that is physically touched by the user is often larger than the visual width (e.g., of the label for the item selected). Therefore, the user aims for the label assuming that the label width (the visual width) means the motor width. Consequently, the user performs operations more carefully than necessary. We conducted an experiment to investigate the effect of the motor and visual widths on finger pointing. After asking participants to explore the motor width, they performed an experimental task. Our experiment shows that the users' movement time depends on the motor width and can be predicted. We also analyze existing interfaces and discuss the implications.

In graphical user interfaces, users can select multiple objects simultaneously via lasso selection. This can be implemented, for example, by having users select objects by looping around their centers or entire areas. Based on differences in lassoing criteria, we presume that the performance of the criteria also differs. In this study, we compare three lassoing criteria and model lassoing motions while considering these criteria. We conducted two experiments; participants steered through straight-line and one-loop paths by using three criteria. The participants handled the lassoing criteria correctly and performed lassoing at appropriate speeds for each path shape. Although the drawn trajectories varied depending on the lassoing criteria, the criteria in the performance and subjective evaluations did not differ significantly. Additionally, from our results, we build a baseline model to predict the movement time by considering the lassoing criteria. We also discuss further experiments to predict movement time under more complex conditions.

For pen-steering tasks, the steering law can predict the movement time for both straight and circular paths. It has been shown, however, that those path shapes require different regression expression coefficients. Because it has also been shown that the path curvature (i.e., the inverse of the curvature radius) linearly decreases the movement speed, we refined the steering law to predict the movement time under various curvature radii; our pen-steering experiment included linear to circular shapes. The result showed that the proposed model had good fit for the empirical data: adjusted r2 > 0.95, with the smallest Akaike information criterion (AIC) value among various candidate models. We also discuss how this refined model can contribute to other fields such as predicting driving difficulties.

To achieve better touch GUI designs, researchers have studied optimal target margins, but state-of-the-art results on this topic have been observed for only middle-size touchscreens. In this study, to better determine more practical target arrangements for smartphone GUIs, we conducted four experiments to test the effects of gaps among targets on touch pointing performance. Two lab-based and two crowd-based experiments showed that wider gaps tend to help users reduce the task completion time and the rate of accidental taps on unintended items. For 1D and 2D tasks, the results of the lab-based study with a one-handed thumb operation style showed that 4-mm gaps were the smallest sizes to remove negative effects of surrounding items. The results of the crowd-based study, however, indicated no specific gaps to completely remove the negative effects.

We investigate stroking motions through successive objects with styli. There are several promising models for stroking motions, such as crossing tasks, which require endpoint accuracy of a stroke, or steering tasks, which require continuous accuracy throughout the trajectory. However, a task requiring users to repeatedly steer through constrained path segments has never been studied, although such operations are needed in GUIs, e.g., for selecting icons or objects on illustration software through lassoing. We empirically confirmed that the interval, trajectory width, and obstacle size significantly affect the movement speed. Existing models can not accurately predict user performance in such tasks. We found several unexpected results such as that steering through denser objects sometimes required less times than expected. Speed profile analysis showed the reasons behind such behaviors, such as participants' anticipation strategies. We also discuss the applicability of exiting performance models and revisions.

Targets on touchscreens should be large enough so that they can be tapped by fingers. In addition to the size of a target, properties of unintended targets around the intended target (e.g., margins) could affect user performance. In this study, we investigate the negative effects of such unintended targets (or distractors), which impose a penalty time when tapped for which users have to wait. Our participants sometimes purposely tapped an empty space on the opposite side of the distractor to avoid tapping it, and such behavior was affected by (1) the size of the intended target, (2) gap between the intended target and distractors, and (3) dimensionality of pointing tasks (1D or 2D). We also found that we could not estimate user performance by using Fitts' and FFitts' laws, probably because tap positions tended to shift away from distractors.

Optimal target size has been studied for touch-GUI design. In addition, because the degree of risk for tapping unintended targets significantly affects users' strategy, some researchers have investigated the effects of margins (or gaps) between GUI items and the risk level (here, a penalty time) on user performance. From our touch-pointing tasks in grid-arranged icons, we found that a small gap and a long penalty time did not significantly change the task completion time, but they did negatively affect the error rates. As a design implication, we recommend using 1-mm gaps to balance the space occupation and user performance. We also found that we could not estimate user performance by using Fitts' and FFitts' laws, probably because participants had to focus their attention on avoiding distractors while aiming for the target.

The steering law models human motor performance and has been verified to hold for a single linear and/or circular path. Some extensions investigated steering around corners. Yet, little is known about human performance in navigating joined linear paths, i.e., successions of path segments with different widths. Such operations appear in graphical user interface tasks, including lasso operations in illustration software. In this work, we conducted several experiments involving joined paths. The results show that users significantly changed their behavior, and that this strategy change can be predicted beforehand. A simple model summing the two indexes of difficulty (IDs) for each path predicts movement time well, but more sophisticated models were also evaluated. The best model in terms of both of R2 and AIC values includes the ID of the crossing operation to enter the second path.

The performance of trajectory-based tasks is modeled by the steering law, which predicts the required time from the index of difficulty (ID). This paper focuses on the fact that the time required to pass through a straight path with linearly-varying width alters depending on the direction of the movement. In this study, an expression for the relationship between the ID of narrowing and widening paths has been developed. This expression can be used to predict the movement time needed to pass through in the opposite direction from only a few data points, after measuring the time needed in the other direction. In the experiment, the times for five IDs were predicted with high precision from the measured time for one ID, thereby illustrating the effectiveness of the proposed method.

We propose Vibkinesis, a smartphone that can control its angle and directions of movement and rotation. By separately controlling the vibration motors attached to it, the smartphone can move on a table in the direction it chooses. Vibkinesis can inform a user of a message received when the user is away from the smartphone by changing its orientation, e.g., the smartphone has rotated 90° to the left before the user returns to the smartphone. With this capability, Vibkinesis can notify the user of a message even if the battery is discharged. We also extend the sensing area of Vibkinesis by using an omni-directional lens so that the smartphone tracks the surrounding objects. This allows Vibkinesis to tap the user's hand. These novel interactions expand the mobile device's movement area, notification channels, and notification time span.