Using HCA and TOPSIS approaches in personal digital assistant menu–icon interface design

International Journal of Industrial Ergonomics 39 (2009) 689–702

Contents lists available at ScienceDirect

International Journal of Industrial Ergonomics journal homepage: www.elsevier.com/locate/ergon

Using HCA and TOPSIS approaches in personal digital assistant menu–icon interface design Ming-Shi Chen a, Ming-Chyuan Lin a, *, Chen-Cheng Wang a, C. Alec Chang b a b

Department of Industrial Design, College of Planning and Design, National Cheng Kung University, Tainan 701, Taiwan Department of Industrial and Manufacturing Systems Engineering, College of Engineering, Univeristy of Missouri-Columbia, Columbia, MO 65203, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 24 October 2007 Received in revised form 2 January 2009 Accepted 19 January 2009 Available online 25 February 2009

Mobile electronic products have recently become very popular because of their portable convenience and versatile functions. The personal digital assistant (PDA) can even access the Internet. However, there is still plenty of room for improvement in the PDA interface. This research proposes a systematic approach to analyze, generate and evaluate a PDA integrated menu–icon interface design for the DigitHub Company based on customer requirements. Hierarchical clustering analysis (HCA) is incorporated with the analytic hierarchy process (AHP) to identify and categorize functional PDA menus and their corresponding icons. We evaluate the importance of each of the different functional menus and categories. We generate five PDA menu–icon interface design alternatives that meet the proposed design guidelines, and we evaluate each for their respective feasibilities. The technique for order preference by similarity to ideal solution (TOPSIS) method is applied to measure the overall operating performance of the five PDA menu–icon interface design alternatives. The evaluation results show that the preferred design is option PDA5, a hierarchical and separated menu–icon layout style that features a two-layer menu structure. We expect that the proposed development procedure for the generation and evaluation of PDA menu–icon design alternatives based on customer requirements will be of interest to interface designers who wish to focus on mobile products.

Keywords: Personal digital assistant Hierarchical clustering analysis Analytic hierarchy process Technique for order preference by similarity to ideal solution Interface design

Relevance to industry: This paper proposes an integrated procedure for designing PDA menu–icon interfaces. Our methodology should help in the creation and optimization of screen layouts for mobile phones, global positioning system (GPS) navigation devices, digital cameras and related screen interfaces.  2009 Elsevier B.V. All rights reserved.

1. Introduction Given recent advances in mobile technology, the handheld personal digital assistant (PDA) has become very popular owing to its small size, high portability and personal information management and communication capabilities. Most PDAs feature a fiveinch diagonal screen, stylus, and detachable or virtual keyboard. They can execute a limited set of tasks and applications. Hayhoe (2001) noted that most PDA screens feature a low resolution, low color depth, and unsatisfactory brightness with weak contrast. At present, leading PDA operating systems include Palm OS, Windows CE, Epoc and Linux. Like desktop or laptop computers, the aforementioned operating systems allow an interface designer to use

* Corresponding author. Tel.: þ886 6 2757575x54327; fax: þ886 6 2522609. E-mail addresses: [email protected] (M.-S. Chen), [email protected] (M.-C. Lin), [email protected] (C.-C. Wang), [email protected] (C.A. Chang). 0169-8141/$ – see front matter  2009 Elsevier B.V. All rights reserved. doi:10.1016/j.ergon.2009.01.010

components such as windows, icons, menus and pointers (WIMP) to design user-facing software. In addition, most PDAs typically feature graphical user interfaces (GUIs) that allow a user to directly manipulate icons and other graphical symbols on a screen (Stramler, 1993). The ubiquity of high-speed Integrated Services Digital Networks (ISDN) and wireless modem services can allow PDAs to access information from anywhere and at any time (Gessler and Kotulla, 1995). PDA menus are categorized as follows: hierarchical and scrolling menu interfaces with multiple layered menus (Wang et al., 2004). A hierarchical menu focuses on the functionmenu structure and the associations among functional icons. The scrolling menu is similar to a menu list, in the sense that a linear concept is used to search for the desired function icon. The user can interact with a PDA by tapping, tapping and holding, or dragging a stylus to manipulate objects and menus directly on the screen. Recently, designers encountered usability and psychological overload challenges in creating PDA software due to the limitations of screen menu interface frameworks. Kiger (1984) found that an individual’s operating time is proportional to the menu selection

690

M.-S. Chen et al. / International Journal of Industrial Ergonomics 39 (2009) 689–702

depth. Snowberry et al. (1985) suggested that increasing the number of menu layers leads to mental overload and loss of direction, lengthening search time and diminishing the advantages of hierarchical menus. Dillon et al. (1990) also argued that if the number of menu–icons exceeds the capacity of the PDA screen, the increased need for scrolling will decrease the effectiveness of any given layout. Since designing too many menu–icons will increase the amount of scrolling and reduce operational effectiveness, Norman (1991) suggested that for best results one should properly balance the depth and breadth of the menus. Moreover, the function-menu layout should address the sequence, functional group and use frequency of functional items. Sanders and McCormick (1993) pointed out that the LCD touch screen is effective for functional icon manipulation but has poor operational accuracy due to its high sensitivity, especially for users with poor vision. Gessler and Kotulla (1995) suggested that only about 50 (Palm OS) to 75 (Windows CE) words can be shown simultaneously on a PDA screen, with the result that designers must observe more conservative interface requirements than those for desktop platforms. Marcuse et al. (1998) revealed that standard criteria such as font size, color, resolution, and graphics capabilities have been challenged by small-screen size. Tang (2001) stressed that users need more time to learn how to use a PDA software product if it includes more hierarchical configuration levels. Independently, Albers and Kim (2002) noted that the small PDA screen is difficult for users to model mentally, leading to short-term memory overload. All of the aforementioned factors can affect navigation performance in search operations, particularly in a deeply structured menu–icon layout. Icon design principles that are applicable for large desktop computer displays are not suitable for small PDA screens. Huang et al. (2002) used principal components analysis to evaluate nineteen existing computer icon elements and suggested that meaningfulness and accessibility are two of the most important factors for icon design. In general, a performance measurement is crucial for evaluating usability goals and comparing competitive design alternatives in the context of the usability engineering lifecycle (Nielsen, 1993). Different interface styles may offer different operating performances. Treu (1994) stressed that a good interface design can decrease the user’s learning time and lower the probability of error in using the software. Chu (2001) proposed that interface design for tiny screens should aim to simplify content presentation to improve readability, help users navigate system information and avoid loss of direction. Lee and Yoon (2004) defined an adaptive menu that would display items that are most frequently used on the first screen for ease of task selection. Park et al. (2007) further examined the effectiveness of adaptable and adaptive menus for desktop computing scenario and concluded that the adaptable menu system was more efficient. Greater attention has recently been focused on the critical role that sound PDA interface design can play in securing competitive advantage. Unfortunately, many research efforts regarding PDA interface design are still limited to general design rules or guidelines. The integration of interface design with customer requirements is seldom considered. The objective of our research is to propose a useful interface design process that integrates hierarchical clustering analysis (HCA) (Lehmann, 1979) and the technique for order preference by similarity to ideal solution (TOPSIS) in the customeroriented menu and icon design process for PDA software. Note that the TOPSIS approach is one of the best-known multiple criteria decision making (MCDM) methods (Hwang and Yoon, 1981; Cheng et al., 2003). In this interface design process, the HCA incorporates an analytic hierarchy process (AHP) (Saaty, 1980) to identify and evaluate the critical attributes of the stated customer requirements. TOPSIS is used to perform the competitive benchmarking and to evaluate the resulting design alternatives.

2. Procedure for PDA menu–icon interface design Menu–icon interface design is a creative process that integrates abstract image components into a complete set of physical menu– icon characteristic specifications that satisfy the stated user requirements. Based on this research objective, we created a conceptual framework for PDA menu–icon interface design as shown in Fig. 1. Fig. 1 illustrates the three-stage development process for PDA menu–icon interface design. The stages include (1) identification and analysis of customer requirements, (2) synthesis and generation of feasible menu–icon design alternatives, and (3) evaluation and recommendation of the most suitable menu–icon design option. Stage 1 identifies the critical customer requirements for PDA products by employing a user preference investigation and analysis framework. HCA is used to group the customer requirements as derived from questionnaires and interviews. The analytic hierarchical process is then used to judge the weights for the identified customer requirements. Stage 2 develops feasible menu– icon design alternatives and establishes online prototypes for experimentation and evaluation, consistent with the customer requirements identified in Stage 1. In Stage 3, an experimental design process is conducted to evaluate the operating time and number of errorful clicks for each of the menu–icon design alternatives. The TOPSIS approach is ultimately employed to evaluate the overall operating performance for each of the design options. The conceptual framework for the development process is shown in Fig. 1. 3. Identification and analysis of customer requirements Customer preferences must first be investigated and analyzed to aid the designer’s understanding of customer requirements in PDA menu design. The designer can then develop a desirable menu–icon system that will meet customer expectations. Possible PDA functional menus will first be collected from various sources including discussions, brainstorming, marketing information, the internet and PDA users. A semantic differential method will then be implemented using a questionnaire based on a five-point scale to assess the significance of the functional menus. Points 1, 2, 3, 4 and 5 correspond to very low, low, medium, high, and very high importance. Each tester will assess the functional menus based on six attributes, namely relevance to purpose, flexibility of system adjustment, efficiency of search or inquiry, length of operating time, practicality of functional menus and entertainment value. These attributes were originally derived using the same identification process as for the PDA functional menus. The survey results are then pooled to form a relationship matrix with rows that denote functional menus and columns that are based on these six attributes. Once the matrix has been created, HCA is employed to categorize the functional menus into a hierarchical tree structure. Note that the Euclidean distance and average linkage are used in this research to compute the distances among functional menus. 3.1. HCA to categorize customer requirements Let SFM denote the set of labels or names that identify semantic functional menus.

SFM ¼ fSFMa ja ¼ 1; 2; .; rg: Similarly, let ATT denote the set of labels or names that identify semantic attributes regarding PDA functional menus,

ATT ¼ fATTb jb ¼ 1; 2; .; sg:

M.-S. Chen et al. / International Journal of Industrial Ergonomics 39 (2009) 689–702

1

691

Identification and Analysis of Customer Requirements Collection of PDA Functional Menus

Hierarchical Clustering Analysis on Categorization of Customer Requirements

Weightings on PDA Functional Categories and Menus

Determination of Importance Degrees for Functional Menus

2 Synthesis and

3

Generation of Feasible Design Identification of Menu and Icon Design Guidelines

Development of Feasible Menu-Icon Design Alternatives

AHP Evaluation on PDA Functional Menus

Evaluation and recommendation of Menu-Icon Design Alternatives Experiment for Evaluating Operation Performance Subjects:Experienced and Inexperienced Groups Experimental Variables: Total Operating Time and Number of Error Clicks Experimental Process: PDA Typical Operating Tasks Performance

TOPSIS Evaluation on PDA Menu-Icon Desig Alternativesn

Recommendation of the Most Suitable Design Alternative Fig. 1. PDA menu–icon development and evaluation processes.

After the semantic functional menus and corresponding semantic attributes were identified, the selected testers were asked to evaluate the relationships between the functional menus and attributes. A rating scale was used to indicate their significance. Based on the assessments made by these testers, hierarchical cluster analysis was then employed to categorize the collected functional menus. The hierarchical cluster analysis in this research relied on the nearest neighbor method, incorporating the Euclidean distance method in the context of the Minkowski Metric approach (Lehmann, 1979; Kusiak, 2000) to categorize the functional menus. The nearest neighbor method can be expressed as (Lehmann, 1979; Kusiak, 2000):

  HCAðSFMc ; SFMd Þ ¼ min dm;n where HCA(SFMc, SFMd) denotes two collected functional menus SFMc and SFMd from the set of functional menus SFMs (c, d ˛ s) that are categorized according to their homogeneity or similarity, dm,n denotes the distance between two functional menus SFMm and SFMn, where m, n ˛ s. The formula for dm,n is expressed as:

dm;n ¼

s  X 2 SFMm;b  SFMn;b

!1=2 ;

b ¼ 1; 2; .; s;

(1)

b¼1

where SFMm,b and SFMn,b denote the number of testers who judge the bth attribute, where b ¼ 1, 2,., s for the functional menus SFMm and SFMn. In this research, r is 44 for the semantic functional menus and s is 6 for the semantic attributes. The hierarchical cluster analysis algorithm uses a critical distance measure at a combinational level to form each cluster. The choice of critical distance measure is based

on the functional menus that have the most homogeneous or most similar distinctions. The critical distance measure ultimately determines the number of clusters. According to HCA, the functional menus are categorized based on six attributes. Let the set of categories for the functional menus be denoted as [C1, C2,., Ce] and the set of functional menus for a specific category Ce be denoted as [CFMg,hg], where g ¼ 1, 2,., e corresponds to the categories C1, C2,., Ce; and hg ¼ 1, 2,., Fg. F1, F2,., Fe represent the number of possible functional menus for the categories C1, C2,., Ce, respectively. 3.2. AHP in the context of evaluating PDA functional menus After categorizing functional menus, the AHP must then evaluate the importance of the functional menus. The test subjects will judge the local priorities for functional menus within each category and the global priorities for each category. The global priority for each category distributes a weight to each local functional menu priority and this in turn is the basis of the importance weighting for the functional menu. The AHP method uses paired comparisons to weight the importance for attributes based on a hierarchical structure. We propose a three-step implementation procedure to assess the relationships between categories and functional menus. Step 1. A paired comparison matrix is necessary for the categories and corresponding functional menus. Pairwise comparisons are conducted in each category according to the functional menu hierarchical tree structure. Based on the pairwise comparisons, the relative degrees of importance are estimated for the categories and corresponding functional menus. This study used numerical values 1–9 and

692

M.-S. Chen et al. / International Journal of Industrial Ergonomics 39 (2009) 689–702

Table 1 The degrees of importance PDA functional menus. Categories Cg and corresponding degrees of importance

Functional menu–icons CFMg hg and corresponding relative importance

C1: Internet (0.128)

CFM11: Electronic Mail (0.034)

CFM12: Web Browser (0.041)

CFM13: Pocket MSN (0.035)

C2: Calculator (0.105)

CFM21: Calculation (0.049)

CFM22: Exchange Rate (0.045)

CFM23: Unit Converter (0.011)

C3: Time (0.103)

CFM31: World Time (0.048)

CFM32: Alarm Clock (0.046)

CFM33: Sounds & Notifications (0.009)

C4: Notepad (0.148)

CFM41: Word Processing (0.052) CFM45: Schedule Information (0.004)

CFM42: Tasks (0.038)

CFM43: Spreadsheet (0.036)

CFM44: Notes (0.018)

C5: Office (0.181)

CFM51: Calendar (0.067)

CFM52: Address Book (0.051)

CFM53: Memo (0.048)

CFM54: Dictionary (0.015)

C6: Multimedia (0.127)

CFM61: Video Player (0.027) CFM65: Pictures Player (0.014)

CFM62: Sound Recorder (0.025) CFM66: E-Book (0.008)

CFM63: Simple Drawing (0.024) CFM67: Map (0.007)

CFM64: Media Player (0.022)

C7: Games (0.109)

CFM71: Gomoku (0.035)

CFM72: Solitaire (0.040)

CFM73: Mine (0.034)

C8: System Settings (0.099)

CFM81: Function Menu (0.015) CFM85: Battery Power (0.006) CFM89: Input Method (0.005) CFM813: Memory (0.002)

CFM82: Function Keys (0.014) CFM86: File Explorer (0.006) CFM810: Search (0.004) CFM814: Backlight (0.002)

CFM83: Network Settings (0.014) CFM87: Input Panel (0.005) CFM811: Device Status (0.003) CFM815: Screen (0.002)

CFM14: Messaging (0.018)

CFM84: Handwriting Settings (0.013) CFM88: Synchronization (0.005) CFM812: Bluetooth (0.003)

Note: Values in () denote the relative importance of functional menus.

their reciprocals as the measurement scale. The values 1, 3, 5, 7, and 9 represent equal importance, weak importance, essential importance, demonstrated importance, and extreme importance between categories and corresponding functional menus, respectively. The values 2, 4, 6, and 8 represent the intermediate values of the adjoining scales. Let CMC represent an e  e pairwise comparison matrix for categories and let CMFM represent an Fg  Fg pairwise comparison matrix for the functional menus of a specific category. Note that e and Fg denote the number of categories and the number of functional menus in a specific category, respectively.

2

1 6 p21 CMC ¼ 6 4 . pe1 2

p12 1 . pe2

1 6 q21 CMFM ¼ 6 4 . qFg 1

. . . .

q12 1 . qFg 2

3 p1e p2e 7 7 .5 1

. . . .

3 q1Fg q2Fg 7 7 . 5 1

The diagonal elements in the matrix CMC and CMFM are derived from self-comparisons in the context of the categories and functional menus, respectively. Thus, pij ¼ 1 and quv ¼ 1,where i ¼ j, u ¼ v, i, j ¼ 1, 2, .,e, u, v ¼ 1, 2, ., Fg. The values on the left and right sides of the matrix diagonal represent the strength of the relative importance of the ith or uth element as compared to the jth or vth element. Let pij ¼ 1/pij and quv ¼ 1/quv, where pij > 0 and quv > 0, i s j and u s v.

Step 2. Calculate the category importance and the corresponding functional menus. The NGM (Normalization of the Geometric Mean) method is used to determine the importance of the category degrees and the corresponding functional menus. Let WCMCi denote the importance degree for the ith category, in which case

Q WCMCi ¼

e j¼1

Pe

i¼1

pij

1=e

Q

e j¼1

pij

1=e ;

i; j ¼ 1; 2; .; e

(2)

Similarly, let WCMFMu denote the importance of the uth functional menu of a specific category, then

Q WCMFMu ¼

Fg v¼1

PFg

u¼1

Q

qij

1=Fg

Fg v¼1

qij

1=Fg ;

u; v ¼ 1; 2; .; Fg

(3)

According to the category degrees of importance WCMC and the corresponding functional menus WCMFM, the relative degree of importance for the uth functional menu of a specific category Ci can be determined. Let RWFM denote the set of relative degrees of importance for the uth functional menu of a specific category Ci, with typical element RWFMi,u:

RWFMi;u ¼ WCMCi  WCMFMu ¼ 1; 2; .; Fg

i ¼ 1; 2; .; e and u (4)

Step 3. Test the consistency of the category degrees of importance and the corresponding functional menus. Perform a consistency check to ensure that the evaluation of the pairwise comparison matrix is reasonable and acceptable.

M.-S. Chen et al. / International Journal of Industrial Ergonomics 39 (2009) 689–702

693

Character Image Extract

Functional Image Extract

Images Composition Icon Images Generation

Image Background Selection

Image Dimensions Determination

Icon and Background Composition

3D Icon Modification

Design Concepts Generation Fig. 2. Icon image generation process for a PDA functional menu–icon design.

Let CCW denote an e dimensional column vector describing the sum of the weighted values for the category degrees of importance. We write

CCW ¼ ½CCWi e1 ¼ CMC,WCMCT

i ¼ 1; 2; .; e

where

2

1 p12 . 6 p21 1 . T 6 CMC,WCMC ¼ 4 . . . pe1 pe2 . 3 2 CCW1 6 CCW2 7 7 ¼6 4 . 5 CCWe

3 p1e p2e 7 7,½WCMC1 WCMC2 .WCMCe T .5 1

(5) Similarly, let CFMW denote an Fg dimensional column vector describing the sum of the weighted values for the functional menu degrees of importance in a specific category, then

CFMW ¼ ½CFMWu Fg1 ¼ CMFM,WCMFMT

u ¼ 1; 2; .; Fg

where

2

1 q12 6 q 1 6 21 CMFM,WCMFMT ¼ 6 6 . . 4 qFg 1 qFg 2

3 . q1Fg 7 . q2Fg 7 7 . . 7 5 . 1 2

CFMW1

3

iT 6 CFMW2 7 h 7 6 , WCMFM1 WCMFM2 .WCMFMFg ¼ 6 7 5 4 . CFMWFg

ð6Þ

Note that if the consistency ratio is less than the threshold 0.1, then the inconsistency is considered tolerable and the estimated results for the vectors of importance in respect of the degrees of the elements are reasonable. The designer can then plan layouts and design icons for functional menus based on the determined priorities as revealed through the AHP procedure. 4. Synthesis and generation of feasible menu–icon design alternatives In designing feasible menus and icons for PDA products, designers should consider certain functional factors, including avoiding software memory overload and user recall overload for a given forward or backward step. The instructions and data should also be appropriately tabulated in a menu or listed on the PDA screen. Consistent with the preliminary PDA functional analysis, menu design factors include hierarchical or scrolling menu types, menu styles, menu and icon layouts, and the number of layers. Note that combining the above factors in different ways will result in a range of different menu and icon user experiences. A few design guidelines are as follows (Shneiderman and Plaisant, 2005): (1) Functional menus should be developed in recognizably symbolic shapes. Both the hierarchical and scrolling menu approaches should feature the same categories. Moreover, the most frequently used menu functions should always appear in the uppermost menu layer. (2) Menu categories and the corresponding icons should be able to share the entire screen. (3) The icons should be sorted and arranged from top to bottom and from left to right based on their relative degrees of importance.

694

M.-S. Chen et al. / International Journal of Industrial Ergonomics 39 (2009) 689–702

Table 2 Our icon designs for the functional menu–icons. Categories Cg

Functional Menu–icons CFMg hg

C1: Internet

CFM11: Electronic Mail

CFM12: Web Browser

CFM13: Pocket MSN

C2: Calculator

CFM21: Calculation

CFM22: Exchange Rate

CFM23: Unit Converter

C3: Time

CFM31: World Time

CFM32: Alarm Clock

CFM33: Sounds & Notifications

CFM41: Word Processing

CFM42: Tasks

CFM43:Spreadsheet

CFM44: Notes

CFM51: Calendar

CFM52: Address Book

CFM53: Memo

CFM54: Dictionary

CFM61: Video Player

CFM62: Sound Recorder

CFM63: Simple Drawing

CFM64: Media Player

CFM65: Pictures Player

CFM66: E-Book

CFM67: Map

CFM71: Gomoku

CFM72: Solitaire

CFM73: Mine

CFM14: Messaging

C4: Notepad

CFM45: Schedule Information

C5: Office

C6: Multimedia

C7: Games

M.-S. Chen et al. / International Journal of Industrial Ergonomics 39 (2009) 689–702

695

Table 2 (continued ) Categories Cg

Functional Menu–icons CFMg hg CFM81: Function Menu

CFM82: Function Keys

CFM83: Network Setting

CFM84: Handwriting Setting

CFM85: Battery Power

CFM86: File Explorer

CFM87: Input Panel

CFM88: Synchronization

CFM89: Input Method

CFM810: Search

CFM811: Device Status

CFM812: Bluetooth

CFM813: Memory

CFM814: Backlight

CFM815: Screen

C8: System Settings

The design of menu–icons is also complicated. Since an icon is an image, picture or symbol representing a functional concept, the design of icons should consider both popularity and ease of recognition, especially for a relatively small visual space like a mobile phone screen. Even though actual icons may differ from the user’s previous experience and knowledge, in this paper we will develop a metaphorical icon system to describe visual hints and relationships. Icon symbols will be collected from various sources as part of the first stage of design analysis to diversify the icon design options. Certain additional icon design guidelines are as follows (Horton, 1994):

with that of a desktop platform and download online prototypes using the ActiveSync software package. This allows experimental subjects to try out the online prototypes and evaluate their usability. 5. Identification of the most suitable menu–icon design option To evaluate the usability of several menu–icon design alternatives, we conducted an experiment to measure performance. 5.1. Choosing operating tasks for experiments

(1) The icon resolution should be at least 48  48 pixels. (2) Consider using an icon that features visual effects such as distinctive and subtle deformations, perspective distortion or shallow shadows. (3) The shape and color of metaphorical icons should be simple, clear and easy to understand. (4) The dominant color of an icon should contrast with that of the background so that the icon’s outline is clear. (5) If an icon features a two dimensional image, it will be more recognizable on a small screen than a three dimensional image would be. (6) A semantic noun icon may be used to represent a menu function. (7) The style and visual appearance of icons in the same category should be consistent and unified. (8) An icon’s image should be represented in a familiar manner to ensure the harmony and consistency of the icon family. (9) The icon should be clearly noticeable and distinct from the background. Based on the menu–icon design guidelines identified from the exploration of customer requirements, it is necessary to imagine several conceptual design alternatives. In this research, the conceptual design alternatives will be constructed as online prototypes using script-driven prototyping tools such as Macromedia Director or Adobe Flash. Using these tools, menu–icon components can be developed to interact with each other, and can be further configured on screen. Designers can synchronize a PDA application

Our experimental framework focused on user fatigue. A reasonable number of operating tasks using current PDA functions were considered for the experiment. Participant discussions, opinions and briefs helped to guide our experimental approach. Because operating tasks have different levels of importance and may have different operating sequences, our experiment featured predefined importance levels for the identified operating tasks. These included: (1) Calendar, with high importance, (2) Web Browser, with high importance, (3) Word Processing, with medium importance, (4) Video Player, with low importance and (5) Solitaire, with low importance. 5.2. Experimental method 5.2.1. Subjects The participants recruited for this experiment included 30 experienced (at least 2 years) and 30 inexperienced PDA users. Each group of subjects consisted of 15 males and 15 females, aged from 25 to 36 (Mean ¼ 29.7, SD ¼ 2.3). All subjects had normal or corrected-to-normal eyesight of 1.0 or better and normal color vision. 5.2.2. Apparatus/material Our online prototypes were constructed and tested on a desktop platform, then uploaded onto PDAs. The Flash Assist version 1.3 software package was used to run the online prototypes for usability testing and full-screen evaluations. The Acer n310 PDA, running Microsoft Windows Mobile 2003 for Pocket PC Premium Edition, and a desktop computer were used for our experiments.

696

M.-S. Chen et al. / International Journal of Industrial Ergonomics 39 (2009) 689–702

Table 3 Characteristics of menu interface design alternatives. Design alternative PDAj

Design layout

Design description

C PDA1

C

C

PDA2

C C

C

PDA3

C C

C PDA4

C

A hierarchical structure with all menu categories in the first layer and the corresponding menu–icons in the second layer. A menu category on the right of the screen must be clicked to open all the associated icons on the next screen. The desired icon is then clicked to execute a program.

This mixed-menus approach combines both hierarchical and scrolling menu types on the screen to use available screen space more effectively. The left part of the screen only displays partial menu categories, with the corresponding functional menus displayed on the right of the screen. When clicking a menu category, both the menu category and the corresponding functional menus are highlighted.

A scenario environmental image based on the user’s preference is displayed on the screen as a background. The icons of the menu categories are in locations that are appropriate to the background image. A hierarchical structure that features all the menu categories appears in the first layer, and the corresponding menu–icons appear in the second layer.

Menu categories with higher priorities are regarded as the most frequently used items and will be displayed in the first layer. When clicking the functional category and menu lists at the top of the screen, the hidden menu categories and the corresponding functional menus will be displayed simultaneously.

M.-S. Chen et al. / International Journal of Industrial Ergonomics 39 (2009) 689–702

697

Table 3 (continued ) Design alternative PDAj

Design layout

Design description

C

PDA5

C C

The design is a pure hierarchical menu that features two layers. A user can click the icon of the menu category in the first layer to open the associated menu–icons in the second layers. A user can then click the desired menu–icon in the second layer to execute a program. By closing the programs or clicking the ‘Window Close-Down’ icon in the top-right corner, the menu category icons are redisplayed in the first layer.

5.2.3. Experimental design The experiment was designed to evaluate the impact of conceptual PDA menu–icon alternatives on user operating performance. The user’s preference for performing different operational tasks was also investigated. During the experiment, each subject was required to randomly perform and complete the assigned operating tasks. The experiment measured the total amount of time required to complete an assigned task and the number of errorful clicks. Note that if we encountered a bug during an experiment, the operating time for troubleshooting was not recorded for that experiment and the bug was coded as an unnecessary user click instead of an operating error. Before the start of the experiment, each subject was informed of the procedures and assigned operating tasks. An existing PDA model menu interface was prepared for practice trials to allow the subjects to become familiar with how to operate a PDA operation. The subjects also needed to feel completely comfortable with the experimental tasks and ambient lighting conditions before the experiment could be conducted. To prevent fatigue, a rest period was scheduled during each experimental session. 5.2.4. Analysis of experiments As mentioned above, our experiment considers the total operating time and number of errorful clicks as two dependent variables. Our analysis will explore the nature of certain common PDA operating tasks for both the experienced and inexperienced users. The experiment was analyzed using Statistica software (version 6) and specifically the analysis of variance (ANOVA) functionality. p < 0.05 was considered to indicate significance. Our results may serve as a reference to designers regarding PDA menu–icon design usability. If the outcome indicates that the operating performance achieved by experienced testers is better than that of inexperienced testers, the design should be considered a candidate for development. Otherwise, the design needs further modifications. Our design criterion is based on the assumption that experienced testers will exhibit the same behavior while using the design alternatives as they would when ultimately using the marketed products.

results, the net operating PDA task performance metrics using each of the menu–icon design alternatives were ranked with TOPSIS. In TOPSIS, the chosen alternative must have the shortest distance from the positive-ideal solution and also the farthest distance from the negative-ideal solution in some geometrical sense (Zeleny, 1982; Hall, 1989). Let OT denote the set of operating tasks as identified by unique labels or names, OT ¼ {OTiji ¼ 1, 2,., f}, and let PDA denote the set of labels or names that identify the menu–icon design alternatives, namely PDA ¼ {PDAjjj ¼ 1, 2, ., t}. Meanwhile, let PDAOT be an f  t multiple criteria decision matrix with typical element PDAOij representing the performance rating metric for each operating task OTi in the context of each alternative PDAj. The implementation steps for the TOPSIS method in the PDA menu–icon design example are described as follows (Sen and Yang, 1998; Tong et al., 2004; Yang and Chou, 2005): Step 1. Establish a normalization performance matrix. A normalization performance matrix representing the set of normalized assessments of the relative performance of the competitive PDA menu alternatives is defined as PDAN ¼ [PDANij]ft, with typical element PDANij,

PDAOTij PDANij ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; Pf PDAOT2ij i¼1

(7) Step 2. Generate a weighted normalization performance matrix. A vector of importance criteria weightings for WNOTV ¼ (WNOTV1, WNOTV2,., WNOTVf) can be obtained through the HCA process. Note that the typical element of WNOTVi, where i ¼ 1, 2,., f, is expressed as:

Q WNOTVi ¼

f i¼1

Pt

j¼1

Q

After the user operating performance comparisons between the menu–icon design alternatives had generated satisfactory

pij

1=f

f i¼1

¼ 1; 2; .; t 5.3. TOPSIS evaluation for the generated PDA menu–icon design alternatives

i ¼ 1; 2; .; f ; j ¼ 1; 2; .; t

pij

1=f ;

i ¼ 1; 2; .; .; f ; j (8)

The vector WNOTV is then transformed into a matrix for ease of operational usage. Let WNOTM be an f  f matrix with the diagonal cells assigned values of WNOTV from the first row to the last row, respectively. The remaining elements in the

698

M.-S. Chen et al. / International Journal of Industrial Ergonomics 39 (2009) 689–702

WNOTM matrix are set equal to ‘‘0’’. Further, let WNPM be an ft weighted normalization performance matrix, WNPM ¼ [vij]ft, with typical element WNPMij ¼ WNOTMii  PDANij, i ¼ 1, 2,., f and j ¼ 1, 2,., t. Step 3. Determine the positive-ideal solution (PIS) and negativeideal solution (NIS) based on the weighted normalization performance matrix (WNPM). The PIS and NIS are defined as

PIS ¼ ¼ NIS ¼ ¼

 

    max vi;j i˛J or min vi;j i˛J 0 jj ¼ 1; 2; .; t j

vþ 1;

 

vþ 2 ; .;

vþ t



j

(9)

    min vi;j i˛J or max vi;j i˛J 0 jj ¼ 1; 2; .; t j

v 1;

v 2 ; .;

v t



j

(10)

where J ¼ {i ¼ 1, 2,., f and i is associated with the beneficial customer requirements of vi,j}, and J0 ¼ {i ¼ 1, 2,., f and i is associated with the cost-effective customer requirements of vi,j}. Step 4. Use the Euclidean distance method to determine the separation distance for each alternative based on the positive and negative-ideal solutions shown in Step 3. The distances are calculated, respectively, as

dþ j

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u f  2 uX vi;j  vþ ¼ t ; i

j ¼ 1; 2; .; t

(11)

j ¼ 1; 2; .; t

(12)

i¼1

d j

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u f  2 uX vi;j  v ¼ t ; i i¼1

 where dþ j and dj represent the distance of the jth alternative from the PIS and NIS, respectively. Step 5. Calculate the relative closeness or similarity to the ideal solution for each competing PDA menu–icon design alternative. Let SD denote an n dimensional column vector describing the coefficient of relative closeness to the ideal solution for the competing design alternatives, with typical element SDj,

SDj ¼

d j dþ þ d j j

displays. We wished to demonstrate that their product can greatly improve the efficiency of small display mobile software. 6.1. HCA categorization and AHP evaluation for identified customer requirements We identified the relevant customer requirements by exploring functional menu–icon systems from several modern commercial PDA platforms. A total of 74 specific menu–icon approaches were collected. We used a semantic differential method incorporating a Likert scale questionnaire to determine the more important functional menu–icons. The five-point scale represented very low, low, medium, high and very high degrees of importance. The questionnaire was available both on paper and online. The subjects included 43 males and 25 females, all of whom had experience using PDAs. Their ages ranged from 20 to 36 (Mean ¼ 28.4, SD ¼ 2.1). After a preliminary evaluation, the research focused on 44 of the menu–icons that were associated with higher cumulative scores for our subsequent cluster analysis. We used the hierarchical cluster analysis to categorize the highly homogeneous stimuli into one cluster. An assessment survey was conducted for the 44 menu items in terms of six attributes, namely relevance to purpose, flexibility of system adjustment, efficiency of search or inquiry, length of operating time, practicality of functional menus and entertainment value. The five-point rating scale denoted not at all, slightly, somewhat, mostly and completely. The survey was available on paper and online. The subjects included 23 males and 18 females who had experience with PDAs and ranged in age from 20 to 30 (Mean ¼ 25.6, SD ¼ 1.7). According to our clustering results, the functional menu–icons were split into eight categories: notepad, calculator, internet, office, games, multimedia, time, and system settings. To evaluate the relative importance of each category, the AHP approach was then employed. A decision support software package, namely Expert Choice version 2000, was used to help implement the AHP evaluation. Table 1 illustrates the AHP evaluation results. In Table 1, the numerical values in brackets denote the relative importance of the categories and the corresponding menu–icons. The consistency ratios for all of the pairwise comparison matrices and the overall consistency ratio were calculated to be less than 0.1, thus meeting the consistency requirement. 6.2. PDA functional menu–icon design

;

j ¼ 1; 2; .; t and 0  SDj  1

(13)

Step 6. Rank the preference order of the competing PDA menu– icon design alternatives according to their relative closeness SD to the ideal solution. Greater values of relative closeness imply a higher ranking order among the competing PDA menu–icon design alternatives. The option with the greatest value is the recommended design alternative. 6. Functional menu–icon design strategies for PDAs Based on the procedure developed above, we can then design and evaluate various PDA menu–icon alternatives. Our research into PDA functional menu–icon design was sponsored by the DigitHub International Corp. (Taiwan & USA), a software consulting company. DigitHub has successfully developed an efficient window-driven interface for small display products, such as mobile phones, digital personal assistants (PDAs) and vehicle operation

The PDA functional menu–icon design process is used to create appropriate icons. Adobe Photoshop and AWicons Pro were used in our study to build the icons and make necessary modifications. The icon image dimension was set at 100  100 pixels with 72 dpi resolution. The eventual finished icon must be reduced to 48  48 pixels to test the minimum PDA resolution requirement. Five graduate students in our industrial design department formed an icon design team. Based on the identified icon design guidelines and questionnaire results, four icon design concepts were generated and evaluated. According to the results of our questionnaire analysis, the design team decided to recommend the circular image as the icon background for the purpose of our experiment. The background image and the generated icon image can be combined to form a PDA functional icon. The icon image generation process for PDA functional menus is shown in Fig. 2. In Fig. 2, the last step for the generated icon image involves modifications using the laws of perspective and the addition of shadows to produce a virtual 3D image. Note that the design team used a circular image as the icon background to follow the unifying design principle that states one should avoid inconsistency

M.-S. Chen et al. / International Journal of Industrial Ergonomics 39 (2009) 689–702

699

Table 4 Operating performance comparisons for both experienced and inexperienced users. Design alternative PDAj

Operation task OTi

PDA1

PDA2 F

PDA3

PDA4 F

PDA5

F

p value

p value

F

p value

OT1

Average total operation time Average number of errorful clicks

123.32 0.73

0.00* 0.40

36.30 0.64

0.00* 0.43

27.04 0.00

0.00* 1.00

24.16 0.21

p value 0.00* 0.65

F 82.80 1.16

p value 0.00* 0.29

OT2

Average total operation time Average number of errorful clicks

94.49 0.73

0.00* 0.40

123.32 1.06

0.00* 0.20

56.79 0.00

0.00* 1.00

53.08 0.21

0.00* 0.65

270.06 3.34

0.00* 0.07

OT3

Average total operation time Average number of errorful clicks

55.87 0.73

0.00* 0.40

26.51 1.08

0.00* 0.30

7.47 2.07

0.01* 0.16

31.79 1.50

0.00* 0.13

21.71 1.00

0.00* 0.32

OT4

Average total operation time Average number of errorful clicks

20.73 1.44

0.00* 0.24

14.31 2.32

0.00* 0.13

78.36 1.16

0.00* 0.29

8.40 4.46

0.01* 0.04*

72.11 1.00

0.00* 0.32

OT5

Average total operation time Average number of errorful clicks

13.82 2.32

0.00* 0.13

27.24 1.83

0.00* 0.18

65.70 1.92

0.00* 0.17

181.99 4.46

0.00* 0.04*

80.46 3.34

0.00* 0.07

Note: ‘‘*’’ Denotes p < 0.05.

between the new icon and existing PDA icons. We expect that creating virtual 3D icon images should enhance icon contrast without negatively impacting PDA memory or processing load. Table 2 illustrates the icon image generation process for the functional PDA menu–icons. Icon image design involves an interface design process for the functional menu style, layout and layer. In this research project, the script-driven prototyping tool Flash MX 2005, was employed to develop five functional menus (PDA1– PDA5), as shown in Table 3. Note that the size of the graphic was set at 240  320 pixels to match the normal screen resolution of the Pocket PC, and the frame rate was set at 12 frames per second (fps). 6.3. Using TOPSIS for PDA operational performance evaluations To explore the usability of the resulting design, we conducted a performance assessment. We used a semi-functional online prototype for this purpose. Five typical PDA operating tasks were chosen for this experimental design consistent with the most important types of PDA activity. These tasks included: (1) OT1: Calendar, (2) OT2: Web Browser, (3) OT3: Word Processing, (4) OT4: Video Player and (5) OT5: Solitaire. To minimize learning effects, all of the 25 (5 PDA design alternatives  5 typical operation tasks) experimental conditions were assigned randomly. In the experiment, each subject used a stylus to select a functional menu–icon for the assigned task and then returned to the main menu for the next assigned task. The total operating time and number of errorful clicks were considered as the two dependent variables. Table 4 shows operating performance comparisons between the experienced and inexperienced users who tested the design

alternatives with the aforementioned typical operating tasks. The results in Table 4 imply that the operating performance for the experienced group was better than that for the inexperienced group. The number of errorful clicks was not significantly different between the experienced and inexperienced groups. This might explain why the inexperienced testers required a longer time than the experienced testers, but maintained accuracy. Because these users had never previously used a PDA, they would be considered a potential market for the development of new PDA user interfaces. The experimental data for average total operating time and average number of errorful clicks from the inexperienced group of users is listed in Table 5. This experimental data can be used to evaluate the five menu–icon design alternatives. Table 5 shows that each operating task OTi has two attributes: (1) average total operating time and (2) average number of errorful clicks. Since the measurement units for the average total time and average number of errorful clicks for each operating task were not consistent, the utility theory concept will be applied so that the data can be aggregated to a single value (Olson, 1996). Note that a default linear function will be used with the base value of 1.0 (best score) and 0 (worst score) determined based on the experimental data available. A shorter average total operating time and a smaller number of average errorful clicks were considered to represent better user operating performance. Table 6 illustrates the utility scores for the experimental data. In Table 6, we assign a value of 1.0 for an average total operation time of 6.109 s and a value of 0 to an average total operation time of 13.096 s. Similarly, a value of 1.0 was assigned for an average errorful click rate of 0.033 and a value of 0 for an average rate of 0.333.

Table 5 Experimental operating performance data for inexperienced users. Design alternative PDAj

Operation task OTi

PDA1

PDA2

OT1

Average total operation time (in seconds) Average number of errorful clicks

11.381 0.133

8.332 0.100

PDA3 8.080 0.067

PDA4 6.109 0.033

PDA5 11.473 0.200

OT2

Average total operation time (in seconds) Average number of errorful clicks

11.227 0.133

8.181 0.100

8.131 0.067

6.228 0.033

11.428 0.267

OT3

Average total operation time (in seconds) Average number of errorful clicks

8.587 0.133

7.614 0.100

6.941 0.067

11.585 0.200

9.488 0.167

OT4

Average total operation time (in seconds) Average number of errorful clicks

8.818 0.167

7.642 0.133

10.326 0.200

13.096 0.233

10.319 0.200

OT5

Average total operation time (in seconds) Average number of errorful clicks

8.594 0.200

7.906 0.133

10.178 0.233

13.043 0.333

9.947 0.233

700

M.-S. Chen et al. / International Journal of Industrial Ergonomics 39 (2009) 689–702

The utility scores for both average total operating time and average number of errorful clicks in performing the operating tasks were then used to generate an aggregated score, as shown in Table 6. In this research, the performance rating PDAOT is identified as:

2

0:912 6 0:934 6 PDAOT ¼ 6 6 1:312 4 1:165 1:087

1:459 1:480 1:562 1:448 1:410

1:605 1:598 1:768 0:839 0:751

2:000 1:974 0:659 0:333 0:008

3 0:675 0:459 7 7 1:069 7 7 0:840 5 0:341

The normalization performance matrix PDAN for the relative performance of the competing menu–icon designs can then be calculated as:

2

0:373 6 0:382 6 PDAN ¼ 6 6 0:537 4 0:477 0:445

0:443 0:449 0:474 0:440 0:428

0:520 0:518 0:573 0:272 0:243

0:688 0:679 0:227 0:115 0:003

3 0:416 0:283 7 7 0:659 7 7 0:518 5 0:210

According to the HCA process and AHP results, the importance vector weights of the five typical operation tasks WNOTV are identified as:

WNOTV ¼ ½ 0:073

0:070

0:061

0:045

0:040 

The weighted normalization performance matrix, WNPM, can then be calculated as:

WNPM ¼ WNOTM,PDAN 2 3 0:073 0 0 0 0 6 0 0:070 0 0 0 7 6 7 ¼ 6 0 0:061 0 0 7 6 0 7 4 0 0 0 0:045 0 5 0 0 0 0 0:040 2 3 0:373 0:443 0:520 0:688 0:416 6 0:382 0:449 0:518 0:679 0:283 7 6 7 7 6 6 0:537 0:474 0:573 0:227 0:659 7 4 0:477 0:440 0:272 0:115 0:518 5 0:445 0:428 0:243 0:003 0:210 3 2 0:027 0:032 0:038 0:050 0:030 6 0:027 0:031 0:036 0:048 0:020 7 7 6 7 ¼ 6 6 0:033 0:029 0:035 0:014 0:040 7 4 0:021 0:020 0:012 0:005 0:032 5 0:018 0:017 0:010 0:000 0:001 The weighted normalization performance matrix, WNPM, results are used to determine the positive-ideal solution (PIS) and negative-ideal solution (NIS) values. Based on the PIS and NIS definitions, the PIS and NIS values are determined as:

PIS ¼ ð 0:033

0:032

0:038

0:050

0:040 Þ

NIS ¼ ð 0:018

0:017

0:010

0:000

0:001 Þ

After determining the PIS and NIS values, the separation  distances dþ j and dj of each PDA menu–icon design alternative from the PIS and NIS can be calculated using the Euclidean distance  method. The values of dþ j and dj for the PDA menu–icon design alternatives are calculated and expressed as:

dþ ¼ ð 0:021 j

0:019

d j

0:024

¼ ð 0:020

0:038 0:046

0:076 0:071

 The dþ j and dj values can be used to derive the similarity degrees (SD) and relative closeness to the ideal solution for each PDA menu–icon design alternative. The SD values are calculated and expressed as:

SD ¼ ð 0:4878

0:5581

0:5476

0:4830

0:5701 Þ

Based on the SD values in the context of the ideal solution, the designer can rank the preference order of the PDA menu–icon designs. As mentioned before, the design alternative with a greater value of SD should be ranked higher. According to the SD values, the preference orders for the corresponding menu–icon design alternatives PDA1 through PDA5 are ranked as 2, 4, 3, 1, and 5, respectively. Because the preference order for the menu–icon design alternative PDA5 is 5, which means the highest rank, PDA5 is considered the final design recommendation. It appears that the recommended PDA5 design alternative uses larger icons and features wider spaces between icons, allowing the users to access the menus more easily.

7. Conclusions The objective of this research was to apply HCA and AHP to group and evaluate consumer requirements. We used the TOPSIS method to evaluate a set of menu–icon design alternatives for a PDA. We developed and tested the menu–icon designs as follows: (1) identify customer requirements with the HCA and AHP methodologies, (2) create PDA menu–icon design alternatives, and (3) evaluate the resulting design alternatives using the TOPSIS approach. During the development process, 74 types of current PDA products were assessed. 44 PDA functional menus were explored and classified under 8 categories. Metaphorical icons for the 8 PDA menu categories were developed consistent with published function-menu design guidelines. The resulting menu–icons met customer requirements and greatly enhanced the visual cognition and the relationships between the commands and system objects, thereby decreasing user operating load. For this research project, we also designed five types of functional menu interfaces that were combined with the menu–icons that we developed. These five PDA menu–icon design alternatives were then evaluated using experimental and TOPSIS analyses. The group of experienced testers showed better total operating time performance than the group of inexperienced testers. This suggests that our PDA menu–icon design options met the design guidelines in terms of allowing experienced users swiftly grasp the concept of operation. Operating performance in terms of the average number of errorful clicks was not worse in the inexperienced group than in the experienced group. Our TOPSIS evaluation results indicated that design option PDA5 is the best. PDA5 is a type of hierarchical menu system with two layers in the menu structure. PDA5 may benefit from larger icons as well as wider spaces. Further investigation and experiments should be conducted to yield insight into this issue. The second most preferable design alternative PDA2 uses a familiar layout that positions the partial menu categories adjacent to their functional menus. Our proposed development process may provide designers with an integrated procedure for PDA menu–icon interface design. Our procedure may be valuable for small-screen interface design in general. We would summarize our findings as follows:

0:046 Þ 0:061 Þ

(1) A mixed PDA menu–icon style that combines both hierarchical and scrolling menu types might perform better than a style with only hierarchical or only scrolling menu modalities.

M.-S. Chen et al. / International Journal of Industrial Ergonomics 39 (2009) 689–702

701

Table 6 Utility scores for inexperienced users. Operation task OTi

Design alternative PDAj PDA1

PDA2

PDA3

PDA4

PDA5

Utility score

Aggregated score

Utility score

Aggregated score

Utility score

Aggregated score

Utility score

Aggregated score

Utility score

Aggregated score

OT1

Average total operation time Average number of errorful clicks

0.245 0.667

0.912

0.682 0.777

1.459

0.718 0.887

1.605

1.000 1.000

2.000

0.232 0.443

0.675

OT2

Average Total Operation Time Average Number of Errorful Clicks

0.267 0.667

0.934

0.703 0.777

1.480

0.711 0.887

1.598

0.974 1.000

1.974

0.239 0.220

0.459

OT3

Average Total Operation Time Average Number of Errorful Clicks

0.645 0.667

1.312

0.785 0.777

1.562

0.881 0.887

1.768

0.216 0.443

0.659

0.516 0.553

1.069

OT4

Average Total Operation Time Average Number of Errorful Clicks

0.612 0.553

1.165

0.781 0.667

1.448

0.396 0.443

0.839

0 0.333

0.333

0.397 0.443

0.840

OT5

Average Total Operation Time Average Number of Errorful Clicks

0.644 0.443

1.087

0.743 0.667

1.410

0.418 0.333

0.751

0.008 0

0.008

0.451 0.333

0.341

(2) The menu–icon category that is associated with the lowest priority will often be located at a deeper functional menu layer. Users will require more search time to find this menu item and may suffer from poor operating performance as a consequence. (3) In general, the hierarchical menu will perform better than the scrolling menu. However, a scrolling menu with two layers might perform better than a hierarchical menu with three layers. (4) A PDA interface that matches existing desktop shortcut menus will enable users to search for a typical high-importance task within a relatively short time. However, users will need a substantially longer time to search for a typical task that is less important and is therefore located in a deeper function menu layer. Acknowledgements The authors are grateful to the National Science Council, Taiwan and DigitHub International Corporation for supporting this research under grant number NSC92-2622-E006-122-CC3. Gratitude is also extended to the reviewers and the Editor for their valuable comments. References Albers, M., Kim, L., 2002. An overview of web design issues for personal digital assistants. Technical Communication 49, 45–60. Cheng, S., Chan, C.W., Huang, G.H., 2003. An integrated multi-criteria decision analysis and inexact mixed integer linear programming approach for solid waste management. Engineering Applications of Artificial Intelligence 16, 543–554. Chu, K.K.W., 2001. Universal usability in practice: users with small screens – less that 640  480. http://www.otal.umd.edu/uupractice/small_screen. Dillon, A., Richardson, J., McKnight, C., 1990. The effect of display size and text splitting on reading lengthy text from the screen. Behavior and Information Technology 9 (3), 215–227. Gessler, S., Kotulla, A., 1995. PDAs as mobile WWW browsers. Computer Networks and ISDN Systems 28 (1–2), 53–59. Hall, A.D., 1989. Metasystems Methodology: A New Synthesis and Unification. Pergamon Press, Oxford. Hayhoe, G.F., 2001. From desktop to palmtop: creating usable online documents for wireless and handheld devicesProfessional Communication Conference, IPCC 2001. IEEE International, 1–11.

Horton, W., 1994. The Icon Book: Visual Symbols for Computer Systems and Documentation. John Wiley & Sons, New York. Huang, S.-M., Shieh, K.-K., Chi, C.-F., 2002. Factors affecting the design of computer icons. International Journal of Industrial Ergonomics 29 (4), 211–218. Hwang, C.L., Yoon, K., 1981. Multiple Attribute Decision Making: Methods and Applications. Springer-Verlag, New York. Kiger, J.L., 1984. The depth/breadth trade off in the design of menu-driven user interface. International Journal of Man Machine Studies 20, 201–213. Kusiak, A., 2000. Computational Intelligence in Design and Manufacturing. John Wiley & Sons, New York. Lee, D.-S., Yoon, W.C., 2004. Quantitative results assessing design issues of selection-supportive menus. International Journal of Industrial Ergonomics 33 (1), 41–52. Lehmann, D.R., 1979. Market Research and Analysis. Richard D. Irwin, Homewood, Illinois. Marcuse, A., Ferrante, J.V., Kinnunen, T., Kuutti, K. and Sparre, E., 1998. Baby faces: user-interface design for small displays. In: Proceedings of the ACM Conference, CHI’98 April, pp. 18–23. Nielsen, J., 1993. Usability Engineering. AP Professional, New York. Norman, L.K., 1991. The Psychology of Menu Selection: Designing Cognitive Control at the Human/Computer Interface. Abex Publishing Corporation, New Jersey. Olson, D.L., 1996. Decision Aids for Selection Problems. Springer-Verlag, New York. Park, J., Han, S.H., Park, Y.S., Cho, Y., 2007. Adaptable versus adaptive menus on the desktop: performance and user satisfaction. International Journal of Industrial Ergonomics 37 (8), 675–684. Saaty, T.L., 1980. The Analytic Hierarchy Process. McGraw-Hill, New York. Sanders, M.S., McCormick, E.J., 1993. Human Factors in Engineer and Design. McGraw-Hill, New York. Sen, P., Yang, J.B., 1998. Multiple Criteria Decision Support in Engineering Design. Springer-Verlag, London. Shneiderman, B., Plaisant, G., 2005. Designing the User Interface, fourth ed. Pearson Education, Boston, MA. Snowberry, K., Parkinson, S.R., Sisson, N., 1985. Effects of help fields on navigating through hierarchical menu structures. International Journal of Man-Machine Studies 22, 479–491. Stramler, J.H., 1993. The Dictionary for Human Factors/Ergonomics. CRC Press, London. Tang, K.H., 2001. Menu design with visual momentum for compact smart products. Human Factors 43 (2), 267–277. Tong, L.I., Wang, C.H., Chen, C.C., Chen, C.T., 2004. Dynamic multiple responses by ideal solution analysis. European Journal of Operational Research 156, 433–444. Treu, S., 1994. User Interface Design: A Structured Approach. Plenum Press, New York. Wang, A.H., Lai, Y.Y., Sun, C.T., 2004. Effect of PDA scrolling- and hierarchy-menu design on users’ operating performance. Displays 25, 109–114. Yang, T., Chou, P., 2005. Solving a multi-response simulation-optimization problem with discrete variables using a multiple-attribute decision-making method. Mathematics and Computers in Simulation 68, 9–21. Zeleny, M., 1982. Multiple Criteria Decision Making. McGraw-Hill, New York.

702

M.-S. Chen et al. / International Journal of Industrial Ergonomics 39 (2009) 689–702

Ming-Shi Chen is currently a PhD candidate in the Department of Industrial Design at National Cheng Kung University, Taiwan. He is also a senior lecturer in the Department of Commercial Technology Management at Transworld Institute of Technology, Taiwan. He received his BS and MS in Industrial Design from National Cheng Kung University, Taiwan, respectively. His research interests are computer-aided design, technical issues for e-commerce, multimedia design for web page, product design and ergonomics.

Chen-Cheng Wang is currently a PhD candidate in the Department of Industrial Design at National Cheng Kung University, Taiwan. He is also a senior lecturer in the Department of Product Design at Fortune Institute of Technology, Taiwan. He received his BSE in Industrial Design from National Cheng Kung University, Taiwan and MS in Computer Science from Stevens Institute of Technology, Hoboken, NJ, USA. His research interests are computer graphics, computer-aided design, product design, Kansei engineering and ergonomics.

Ming-Chyuan Lin is currently a professor in the Department of Industrial Design at National Cheng Kung University, Taiwan. He received his BSE in Industrial Design from National Cheng Kung University and MS and PhD in Industrial Engineering from the University of Missouri-Columbia, USA, respectively. His research interests are human factors engineering and computer-integrated design and manufacturing systems.

Dr. C. Alec Chang is an associate professor of the Department of Industrial and Manufacturing Systems Engineering, University of Missouri-Columbia, USA. He has published more than 80 papers related to quality enhancement using experimental design, engineering data analysis, feature based design retrieval systems and general database design and ERP systems.