Keywords

1 Introduction

Checking news is a popular action for most people every day. In recent years the trend of using a smartphone while driving, and browsing the news as secondary task, is increasing. Any visual secondary task, i.e. browsing news, while driving is often done in spite of the potential risk of car accident. According to statistics from the National Safety Council, 26 % of crashes (1.5 million) involved Smartphone usage distraction in 2012.Footnote 1 Yu et al. revealed that driving is one of the top activities during which people used their smartphone as a secondary task, in addition to walking and waiting. It is worth noting that, in 64 % of these situations, the users were browsing the news [1]. These statistics show the practical need for a particular consideration of user interface (UI) design for browsing the news on a smartphone while driving.

Research showed that only 24 % of mobile news consumers reported that they use applications (apps) to check news [2]. Despite the publication of a variety of apps, many users still abstain from relying on them, and check news through following significant reasons:

  • Apps switch news websites’ layout, which can impede users’ browsing while driving

  • Apps may compound or exclude parts of their respective original websites; missing parts (e.g. pictures, old news) can be found by viewing original news websites

  • Smartphones’ lack of memory influences speed and installations

In addition to mobile apps, mobile web browser is another option for browsing the news. A survey was conducted to identify mobile users’ experiences using mobile web browsers. According to the result, 82 % of participants disliked this way of browsing due to, for example, bad content layout and slow interface [2]. Therefore, the mobile web browser needs a lot of improvement. This paper seeks to consider all of these challenges to obtain optimal UI news websites for use while driving.

In line with UI design importance, the interface style, which is dependent on language, should be carefully studied. A written language can influence browsing performance while driving because information perception and processing highly depend on the style of the written language in the information system [3]. In terms of writing, Persian, Arabic, Urdu, Pashto, and Dari languages share the same letters and altogether these speakers constitute 10 % of the world’s population [4]. The language of news considered in this study is Persian because the Persian language has different alphabets and style.

The objective of this article is to propose better website adaptation on smartphones for browsing news written in Persian while driving.

2 Literature Review

2.1 Page Adaptation for Smartphone

Different ways to adapt websites on smartphone screens are direct, manual, and automatic [5]. Direct adaptation is not a suitable method due to some small device limitations such as a small screen. Manual methods are expensive to develop for each particular website. For these reasons, this research utilizes automatic methods. Motamedi, et al. categorized the automatic adaptation in four sub methods [6].

Format Conversion. In this method, a website is splitted in different parts that connect with links to have better UI on small size screens [7]. But the weak point of this method is deep structure, which creates confusion for users who are familiar with a website’s original view [8].

Overview. At the outset, this method provides the original view of websites at the beginning. In this way, users are allowed to use their skills and memories to navigate the website even on their smartphones. B. Mackay et al. proposed the “Gateway” method in which users click or rollover the individual sections [7]. Then the selected section is expanded over the original view of the website [6].

Summarization. As you can guess from the name of this method, all contexts are summarized based on key words. Although this method can summarize the whole website on the small screen size of smartphone, it has two disadvantages: tables and pictures are eliminated from websites after summarization, and key words are developed manually for each website paragraph, which takes a lot of time and effort [9].

Linearity. In this method, website layouts are changed to a long linear list which adapts to the small screen size of a smartphone. This long linear list, a menu, is generated with all the extracted links and content appearing as menu items. By doing this, websites effectively separate navigation and action, making navigation simply a matter of selecting a link to follow from a list [10]. However, this method changes the original layout of the website, which can create confusion for users.

Among these categories, the overview method is less expensive and has a better user interface due to its allowing users to have the same layout of an original website seen on a desktop computer [7]. In this study, three overview methods were compared while driving.

2.2 Persian Language Complexity in Terms of UI Design

Persian language letters are different from English or Latin in many ways, for instance: starting from right to left for writing, connecting letters within a word (even in printed writing, see Fig. 1), twenty-eight letters having separate parts (as circled in Fig. 1), and requiring a larger optimum font size in terms of readability [11]. The Persian language, which shares letters with similar languages such as Arabic, Urdu, Pashto, and Dari, has thirty-two letters which have different shapes based on their position in words and sentences. Including all shapes and sizes, the Persian language has 115 effective letters [12]. All of these differences make Persian a complicated language in terms of UI design [13]. These differences influence the perceptions and performance of Persian users searching websites [6].

Fig. 1.
figure 1

Comparing the complexity of english and persian languages

Full size image

3 Methodology

3.1 Design

The experiment was a 3 × 1 variables design (browsing methods: Pop-up, Full-screen, and Auto zooming) for analyses of efficiency, effectiveness, and satisfaction.

Browsing Methods. This paper utilized the same browsing methods introduced by Mohamed et al. [6]. As the study concluded, the overview method is a better method for smartphones due to using the same layout as the website’s original appearance [6]. The innovation of this study is the use of three different overview methods which are compared while driving, as a secondary task instead of a first task.

Many screen adaptations for smartphones such as apps and mobile browsers exclude or compound some part of news websites. However, these browsing methods avoid do so and try not change layout of news websites by considering the lack of space. Moreover, by cutting zooming action during checking news, these methods improve UI. In detail, we used JavaScript functions to detect all clicks and capture the background based on a predefined method. Figure 2 illustrates the screen shot of the first look at the website, which is same in all three methods used in this study, and the original website. After clicking on the screen each method acted differently (see Fig. 3).

Fig. 2.
figure 2

The first screen shot of all three methods

Full size image
Fig. 3.
figure 3

Screen shot of all three methods after one click

Full size image

First method: Pop-up. In this method, the selected and clicked news would Pop-up in a bigger font size on the website; this is expanded and superimposed on top of the overview [6]. Figure 3(a) shows the example of a Pop-up method after one click on the desired news.

Second method: Full-screen. In this method, after having an overview of the news website, a user can read the whole selected news in full-screen mode only using one click [6, 7]. You can see the screen shot of this method after one click in Fig. 3(b).

Third method: Auto-zooming. In this method, after having the overview, a user is able to automatically zoom in and out of any of the selected news on the website; however, this is not a concern of our study. In order to read the whole news content, a user needs to double click [6]. In this case, the user can read in full screen mode, which is similar to Full-screen method. This method is depicted in Fig. 3(c).

3.2 Participants

This study compared these methods to obtain the best news website adaptation for smartphones via an experiment performed by 12 people ranging from 25–35 years old. All participants were randomly selected from the local Persian population. The participants had annual driving experience of at least 15,000 miles. These drivers have held a valid license for at least two years and count as active drivers (i.e., they have driven at least 30,000 miles in their lifetimes). Persian is the mother tongue of all participants. Inexperienced and elderly drivers are excluded in this study due to low driving performance. The participants reported normal or corrected-to-normal vision.

3.3 Apparatus

A virtual-reality driving simulator was employed in the experiment, specifically the TranSim VS IV driving simulator produced by the L3 Corporation. The simulator provides high-fidelity, real-world driving environments that can be customized for various applications. It is a fixed-base simulator consisting of a regular driving module and three channel plasma monitors in an immersive driving environment that combines the look and feel of a real vehicle. Participants interact with the simulator using a sedan steering wheel and pedals that provide real-time feedback. As Wang et al. [14] pointed out, driving simulations can be safe and effective environments to evaluate drivers’ electronic device interfaces.

A separate computer was used to run and control the simulator through the Operator Console (OPCON) software. Another software, “Scenario Builder”, was used to create desired test conditions for various scenarios. These driving scenarios were designed to be predictive of driving performance in naturalistic settings and standardized road tests. Simulated “real-world” driving events were included that placed relatively greater or lesser demands on the effective utilization of: (1) executive control processes, (2) visual search mechanisms, and (3) selective attention mechanisms underlying the appropriate processing of both relevant and irrelevant orienting cues (Fig. 4).

Fig. 4.
figure 4

Driving simulator

Full size image

Furthermore, in this study drivers were equipped with a smartphone. Due to smartphones’ different sizes, this study used one common small device, the Google Nexus 5 Android Phone. The size of this Smartphone is 5.4” × 2.7” × 0.3”, and the main display’s resolution is 1920 × 1080 pixels. This smartphone was equipped with a framework that can adapt a Persian news website with these methods. In order to avoid Internet speed effects this experiment conducted offline.

3.4 Procedure

Firstly, an orientation video was played to explicitly explain the experiment to the participants. Then, participants were asked to perform a ten minutes warm-up run, followed by the experiment. In total, a participant went through three scenarios in a random manner. At the beginning of each scenario, the participant was informed that what kind of browsing methods (Pop-up, Full-screen, and Auto-zooming) should be used.

In each scenario, a participant drove about one mile on a city roadway, which took approximately two to three minutes. The participant was asked to keep their speed in the 25–35 mile per hour range. In order to simulate the city environment, five challenging events occurred in each scenario. For example, challenging events included events where other drivers or pedestrians emerged suddenly, thus provoking collisions if not avoided. By demanding active action from the driver, we were able to obtain an assessment about the driver’s performance and response. In order to avoid the learning curve effect, three similar scenarios were designed. These scenarios share similar road environments, urban roadways, and routes. However, they differ in terms of objects — the people and cars used in the scenarios.

Lengths of selected news, randomly assigned to each method, were almost the same (M = 87.16, SD = 1.16). It is worth noting that since the news locations were randomly changed when participants browsed the websites, they were not able to use their memory to guess the location of the news.

After a warm-up run in which participants familiarized themselves with the scenarios and browsing methods, the experiment began. Participants were asked to complete two tasks. In the first task, participants, while driving in the scenarios, browsed the news, which meant that they needed to find, read aloud (heard by the researchers), and understand the browsed news using the methods. Time of task is one of our responses and was measured through captured video using a timer on the screen of the smartphones. This data objectively measured the efficiency of the methods. Moreover, participant driving performance was recorded and monitored by two researchers in order to evaluate the participant performance in scenarios based on a grading system, which will be explained in following section. The overall score of driving performances is the response for measuring effectiveness of each method.

In the second task, participants were asked to complete a questionnaire. In the first part of the questionnaire, participants answered multiple questions about the news they read during the scenarios. If they were not able to identify the correct option, their data was excluded. In the other part of the questionnaire, participants rated (from 1 = not satisfied to 5 = very satisfied) the ease-to-use (satisfaction level) of each method for checking the news.

3.5 Variables

There are a wide variety of usability measurements. Usability is counted as a primary factor in UI and the consequent success of any kind of adaptations. Users’ satisfaction, time efficiency, and effectiveness are the elements of usability based on ISO 9241. Therefore, this study evaluated usability of these methods by measuring time completion of the tasks, driving performance, and subjects’ opinion about ease-to use of methods.

The independent variables were the browsing methods while the dependent variables measured were effectiveness, efficiency, and satisfaction (ease-to use) level of methods.

As mentioned above, effectiveness of methods measured by driving performance was assessed by a grading system. This system calculated the overall score (from 100) of participants’ driving performance while browsing the news. Ten categories with different weights are included in this grading system (Table 1) [15].

Table 1. Categories and weights
Full size table

The other dependent variable is efficiency of methods, which was measured with task completion times. The time was recorded precisely via capturing a screenshot from a smartphone’s screen during the experiment. Moreover, experienced subjects were asked to rate the browsing methods according to satisfaction (ease-to-use).

One-way analysis of variance (ANOVA) with the significant level of 0.05 was utilized for within-subject analysis. Three methods, as independent variables, analyzed three dependent variables: efficiency, effectiveness, and satisfaction.

4 Results

Two of the 12 drivers who participated in this experiment are excluded from the experiment results. These two drivers could not correctly answer the first part of questionnaire, which was about the news that they read during the experiment. Next, the normality assumption of user performance data obtained from the captured video and driving simulation were checked. All data passed the normality test (with α = 0.05), which ensured the validity of the comparison analyses. Then, the results were analyzed using one-way ANOVA (with 95 % confidence level) procedure (see Table 2).

Table 2. Summary of ANOVA results
Full size table

According to the results, there was no significant method in terms of effectiveness (F-value = 0.70, p-value = 0.504). As noted above, the driving performance was considered as a response for effectiveness. With 95 % confidence, the mean of the driving performance for all three methods was between 73.00 and 79.86. Figure 5 illustrates the 95 % confidence interval of each method particularly for mean and median.

Fig. 5.
figure 5

Confidence interval for methods

Full size image

In terms of efficiency, methods had significant impact (F-value = 5.05, p-value = 0.014). Figure 6-a illustrates the main effect of this significant factor. In this figure, it is clear that subjects needed less time to complete the browsing task with method 3 (Auto-zooming).

Fig. 6.
figure 6

Main effect plot: (A) efficiency (B) satisfaction

Full size image

Based on subjects’ points of view, methods significantly affected their satisfaction (F-value = 6.84, p-value = 0.004). According to the main effect plot (see Fig. 6-b), the best method in terms of subjects’ satisfaction was method 2 (Full-screen) and with roughly 0.2 difference with method 3 (Auto –zooming), the second best method.

5 Discussion

In contrast to previous research [7, 9] that compared direct and/or linear website adaptation for smartphones, this research considered three methods from the overview category. In addition, this comparison was conducted in a driving simulator to measure the effectiveness, efficiency, and satisfaction of these methods as a secondary task while driving. The other innovation of this study was utilizing Persian as the news language due to its different shape and letter complexity.

Results revealed method 3 was the most efficient method. Method 3 provided the Full-screen of the news after two clicks on the smartphone’s screen. In this way, a user did not need to zoom in on the content because the letters became bigger and more readable [6]. Furthermore, method 3 is extremely similar to the way users on desktop computers browse the news (double clicking to open the news). Thus, in method 3, users spend less time (more efficient) browsing (= finding + reading) news.

Regarding subjects’ rating, method 2 was the most satisfying method among the three methods. As previously mentioned, this method provided the full-screen of the news, which provided more room and improved letters’ readability after only one click. As result, subjects were satisfied with method 2 for browsing tasks. However, considering the actual time that they spent completing the tasks (browsing news), users invested significantly more time (lowest efficiency) on this method. This is because of the limited capacity of the human brain, which cannot properly process the secondary task. Drivers clicked two times instead of once as they became acclimated to the task. In other words, the reason for the low efficiency of this method was not being familiar enough with needing only one click to open the desired news while driving (as secondary task), whereas users had better efficiency as primer task [6].

Effectiveness was not significant factor based on the analysis. However the confidence interval still revealed interesting information about this factor. Method 1 had the lowest mean for driving performance compared to the other methods. As a result, method 1 could be the most distracting method for drivers. This distraction could be the result of the smaller window method 1 provided after one click.

Comparing driving performance while browsing news with driving performance while hands-on texting [15], this finding can lead to the conclusion that browsing news distracts drivers more than texting. Longer context and smaller font could cause this increased distraction.

6 Conclusion

This study identified the usability of three website adaptation methods — Pop-up, Full-screen and Auto-zooming — for browsing news while driving in a city environment. Through the driving simulator, driver performances were examined under the effect of browsing news. Moreover, the efficiency of these methods was measured by recording completion times of the browsing task. Users’ satisfaction was assessed subjectively immediately after the driving test. As a result, Auto-zooming and Full-screen methods were found to be the most efficient and satisfying methods, respectively, for browsing the news while driving. In addition, the Pop-up method was found to be the most distracting browsing method for drivers. The results gained from the study support the notion that the more complex the visual tasks, the more unsafe distractions are while driving.

Although this research employed a high fidelity simulator with a high level of experimental control, in real-life driving settings, such as naturalistic studies, an expanded number of participants are needed in order to ensure the validity of the findings. In future studies, other factors such as weather condition, traffic density, visual conditions (day/night), and different languages should be considered.