Keywords

1 Introduction

Many studies on flexible displays have been conducted as the consumer device market demands greater portability and durability. Flexible displays are thin, pliable, and yielding; they do not break easily [4]. Such displays are highly portable and durable. Flexible displays can be divided into various hardware types by the development flow shown in Fig. 1. Currently, second-generation bendable displays, such as curved displays, are being released into the device market. When displays have more freedom of flexibility, they can achieve a disposable property, such as paper-like thin displays.

Fig. 1.
figure 1

Flexible display roadmap [9] (adapted from Displaybank: flexible display technology and market (2007–2017) report).

Full size image

The paper-like thin displays are more portable. However, the displays do have certain disadvantages. For example, display with higher flexibility can be bent unintentionally, because it is relatively difficult to prop up a paper-thin display. Thus, the user might not be able to accurately touch an appropriate target on the display, because TI in paper-like thin display can induce an unintentional bending.

Consequently, the use of bending interactions (BIs) on flexible displays is being developed to allow users to interact with applications using methods other than touch. However, BIs cannot completely replace touch because there are strengths and weaknesses for each interaction type [5]. Thus, BIs and TIs have to be used for complementary use between those two interactions.

The majority of previous BI studies analyzed the performance [6] or user preference [2, 3] and [7] for BIs by comparing various flexible shapes. However, it is difficult to be certain that the extracted bending interaction can be applied to flexible displays with touch interface, because they did not compare the preference of BIs with existing touch-based interactions. Thus, results can change when the defined interactions applied to flexible displays are compared with familiar TIs currently used on many different devices.

There was study that BI was more effective in specific command by comparing TI [1, 10]. However, it is hard to know that the effective BIs are intuitively chosen in flexible display, because the study did not observe an intuitive choice between BI and TI. Moreover, in the TT Ahmaniemi et al. study, a mobile-size display was used and the display was not thin like a paper. The properties are different from properties of paper-like display which have high flexibility. Therefore, in order to suggest utilization directions of BI, it is necessary to find out why users intuitively choice TI or BI in each command when they use the paper-like tablet prototype.

We have questions regarding effective use of the BI on flexible displays along with commonly used TIs in current displays: in flexible display, can BI be used as an intuitive interaction for all commands? Is the preference for BI higher than for TI? What are the implications of a preference for familiar TI being higher than for unfamiliar BI?

2 Methods

2.1 Focus Group Interview and Pilot Test

From focus group interview (FGI), we extracted appropriate applications for comparison of preference between BI and TI. And we provided interaction options that contains BIs and TIs based on previous studies and pilot test.

Applications.

To understand the preference for TI or BI, we considered BI unfamiliarity because there is a familiarity difference between TI and BI. Thus, we extracted applications where we could find a potential for BI use. Through FGI, we considered the posture of the “Lap hand” [11] among various use postures for tablet devices. The “Lap hand” is the way in which users hold a tablet in their lap supported by their hand. We decided that holding a display with both hands (display-holding posture) offers a higher degree of flexibility than placing the display on a desk. Thus, we selected applications that could be used in the display-holding posture for most of the applications’ commands. Based on our consideration factors, we chose applications for watching contents by the FGI. Moreover, in order to understand the similarities and differences in the interaction patterns according to the application, we divided the applications into a dynamic application of video contents and a static application of eBook content.

Interaction Options.

In our experiment, we provided 18 possible interactions for executing commands, as shown in Fig. 2. To extract the interactions, we considered a common feature of the input method between BI and TI, and selected interactions that required only two hands without an additional tool, such as a touch pen. Thus, we selected “icon touch” and “touch-gesture” as the TIs. The icon TI was divided into three interactions, numbered 1 to 3 in Fig. 2, by considering the display-holding posture. Based on our FGI, the five most common touch gestures, numbered 4 to 8 in Fig. 2, were extracted in reference to the touch-gesture reference guide [8]. Further, eight BIs, numbered 9 to 16 in Fig. 2, were selected based on the experiment results of Lee et al. [7] and Lahey et al. [2]. We extracted BIs with a high preference rating in these studies, and then conducted a pilot test to analyze these BIs. We added additional BIs, numbered 17 and 18 in Fig. 2, based on our participants’ opinions from the pilot test.

Fig. 2.
figure 2

Interaction options: eight for TI and ten for BI

Full size image

2.2 Experiment

Experiment Design.

The experiment had a within-participants design factor of TI and BI and a between-participants factor of video and eBook applications. 10 out 20 participants invoked the commands in the video application by choosing TI or BI. And remaining 10 participants invoked the commands in the eBook application by choosing TI or BI.

Experiment Procedure.

The experiment participants viewed an instruction video that illustrated the 18 interactions. We allowed time for the participants to adapt to these interactions. The participants attempted to control the paper prototype. Following the adaptation time, the participants invoked the commands specified on the video and eBook applications using either BI or TI according to their intuitive preference. A secondary task was required when the participant executed a command using TI. These participants had to repeat the commands using BI intentionally. This time, they could use other bending shapes as well as the offered BIs. We called this interaction “intentional BI.” Participant behavior was recorded during the experiment. Upon completion of the experiment process, the participants completed a five-point-scale questionnaire for “usability” of the interaction they selected. And interviews were conducted as shown in Fig. 3.

Fig. 3.
figure 3

Experiment procedure

Full size image

Materials and Participants.

A total of 40 participants between the ages of 20 to 35 years were invited to participate in the experiment. The experiment proceeded after KIST institutional review board approval (IRB number: 2014–012) and the task-completion time per participant took about 30 min. In experiment, a low-fidelity prototype was used. The prototype made by putting an OHP film and a paper together in order to have a similar curvature to a bendable display currently in development. The prototype was about the size of a fifth generation iPad (240 * 170). Moreover, in order to give a feeling of working prototype, we provided the visual feedbacks for each command using animation function of PowerPoint presentations. When the participants performed certain command, the visual feedback was provided at the same time by researcher.

3 Results and Discussion

To analyze preference and usability between TI and BI, we performed the Wilcoxon’s signed rank test of nonparametric statistics because the data did not assume normality.

3.1 Frequency of BI Use

The frequency of BI use for the video application was significantly lower than that of TI (Z = −2.742, p < 0.05, N = 20): TIs were mostly used for executing seven commands in the video contents. However, for the eBook content application, there were no significant differences between frequency of use of the two interaction types (Z = −1.064, p = 0.287, N = 20): When the participants executed seven commands for the eBook content application, BI and TI were distributed among the commands at an almost equal frequency.

3.2 Commands with Potential BI Use

In both applications, the frequency of BI use was not significantly higher than TI. However, we found potential BI use for certain commands through frequency analysis, as shown in Table 1.

Table 1. Interaction frequency in each application
Full size table

Command 1 for User Friendly Bending: ‘Infrequently’ Used Command.

Participants preferred BI use for bookmark command included in both applications. The participants indicated that BI felt more intuitive, but required more muscle movement because BI is a direct physical movement compared with touch. Thus, participants wanted to use BIs for infrequently used bookmark commands.

Command 2 for User Friendly Bending: ‘Continuously’ Used Command.

In addition, for fast-forward and rewind commands included in both applications, the participants preferred BI, or used BI at nearly the same rate at TI. The commands were required for moving a considerable range in whole range of data. BI allowed the participants continuous control with one movement. However, the participants did not recognize volume control as a continuous command, although volume ranged from high to low. Rather, the participants preferred to control the volume quantitatively with several movements using TI.

3.3 Property of Preferred BIs

To understand detailed use context for BIs, we analyzed the preferred bending shape and area for bookmark, fast-forward, and rewind commands as indicated in Table 2.

Table 2. Bending shape with high preference rate
Full size table

User Friendly Bending Method 1: BI Applying “Analog Metaphor”.

The preferred bending shapes were similar for both applications. When the participants executed the bookmark, fast-forward, and rewind commands, they used BIs with appropriate “analog metaphors” for the commands: an analog metaphor of page tagging was applied to the bookmark command. And an analog metaphor of leafing through the pages was applied to forward and rewind commands. In particular, the preference rates for BIs in the fast-forward/rewind commands were higher for eBook contents than for video contents. The participants indicated that analog metaphors for leafing through the pages were easier to fit into the commands for eBook contents because the paper prototype felt similar to a page in a book.

User Friendly Bending Method 2: BI in Corner and Edge area.

The shapes in edge and corner area were mostly used for the commands: for the fast-forward/rewind commands, BIs in edge areas, numbered 16 and 17 in Fig. 2, were used most frequently. For the bookmark command, BI in a top corner area, numbered 11 in Fig. 2, was used most frequently. Through an interview, the participants indicated that they preferred BIs in corner and edge area to avoid screen distortion by bending.

However, in bookmark command of the eBook contents, BI in a center area, numbered 10 in Fig. 2, was preferred, as well as BI in a top corner area, numbered 11 in Fig. 2. In the case of the center area, the participants folded the paper prototype in half, as if closing a book. The participants indicated that the static application of eBook contents was less affected by screen distortion, compared with the dynamic application of video contents. The reason is that the focus area for eBook contents is smaller than for video contents, as indicated in Fig. 4 (b): when users read eBook contents, they only need to focus on the “reading area”, with the exception of the “passed area.” Therefore the range of flexibility use was wider in static application than in dynamic application.

Fig. 4.
figure 4

Focus area (red dotted square) (Color figure online)

Full size image

3.4 User Experience of TI and BI

Intuitive TI vs. Intuitive BI.

We analyzed the usability difference between intuitive BI and intuitive TI. For both applications, there were no significant differences (video application: Z = −1.680, p = 0.093, N = 20; eBook application: Z = −0.311, p = 0.756, N = 20). The participants were comfortable with both intuitive TI and BI (Usability average values of TI and BI in video contents: 4.307 and 4.007, respectively; Usability average values of TI and BI in eBook contents: 4.28 and 4.335, respectively).

Intuitive TI vs. Intentional BI.

For the commands where TI was used more often than BI, we compared the “usability levels” between intuitive TI and intentional BI obtained by the secondary task. Our analysis results indicate that the usability levels for intuitive TI were statistically higher than for intentional BI in both applications, as listed in Fig. 5. The participants indicated that BI for these commands was not accepted as an intuitive interaction because it is hard to apply an analog metaphor into the commands.

Fig. 5.
figure 5

Usability comparison between intuitive TI and intentional BI

Full size image

In particular, we found from the participant interview a common reason for not using BI with zoom in/out commands. When the participants executed these commands using BI, they could not accurately control the zoom because of difficulties selecting a target on the display. Interestingly, the participants indicated that TI was also uncomfortable for these commands. For the commands, the touch gestures numbered 6 and 5 in Fig. 2 were used most frequently. However, when the participants used the TIs, the paper prototype was bent unintentionally because of its high flexibility. Thus, the participants wanted a combined use of both BI and TI for zoom commands.

4 Conclusion and Future Works

In this study, we found a “User-Friendly” bending method for video and eBook applications through comparative analysis of TI. Through observation, we found the reason that BI does not apply to all commands in a flexible display. BI can give intuitive feeling, but it can induce screen distortion and requires significant muscle movement. Thus, the following conditions are required in order to become “User-Friendly” BI.

  • Which of commands in application are used ‘infrequently’ or ‘continuously’?

  • Is there a BI applied an analog metaphor for the commands? Or is it easy to apply an analog metaphor into the commands?

  • Is the BI in corner or edge area?

Of these conditions, influence of the third condition can become weak depending on application. Thus, we propose BI using analog metaphors when users execute a “saving task” that is used less frequently, but that is important, and for executing a “moving task” in a wide range of data. There was also an opinion that combined interaction between BI and TI is suitable for commands that require detailed control. Thus, we will propose a new interaction that uses TI for target selection and BI for target control. And it is necessary to verify the optimum conditions using a working prototype in future work.