20 participants, 10 current Apple iPhone owners, and. 10 non-iPhone owners were asked to use a web-based IGA tool to design touchscreen and non- ...
PROCEEDINGS of the HUMAN FACTORS and ERGONOMICS SOCIETY 55th ANNUAL MEETING - 2011
1666
Determining the Effect of Users’ Mobile Phone on Design Preference via Interactive Genetic Algorithms Dan Nathan-Roberts1, Jarod C. Kelly2, Yili Liu1 Industrial and Operations Engineering, University of Michigan, 2School of Natural Resources and Environment, University of Michigan
1
This study uses an Interactive Genetic Algorithm (IGA), a design space searching method, to determine the degree to which a user‟s current mobile phone impacts their design preference, and how features in a product can change preference. IGAs mimic evolution by iteratively converging towards a design while exploring a design space through random mutations. 20 participants, 10 current Apple iPhone owners, and 10 non-iPhone owners were asked to use a web-based IGA tool to design touchscreen and non-touchscreen phones for dialing use only. Similar to other IGA mobile phone work (Nathan-Roberts & Liu 2010), the IGA varied screen size, button spacing, and phone radius independently. Results showed iPhone users, and non-iPhone users did have different design preferences, but that there was a bigger difference between touchscreen phone owners (iPhone and non-iPhone touchscreen phones), and non-touchscreen phone owners. Overall participants had significantly different preferences for touchscreen and non-touchscreen designs for all variables except for the vertical button spacing, and phone radius. This work is part of a larger research study of aesthetic ergonomics of mobile phones, specifically looking at usability, and the capacity of users to combine multiple goals in design. Future research needs are discussed, including further testing the effect of non-iPhone touchscreen phone ownership.
Copyright 2011 by Human Factors and Ergonomics Society, Inc. All rights reserved DOI 10.1177/1071181311551347
INTRODUCTION Mobile phones have become ubiquitous in modern society with 4.6 billion mobile phone contracts globally (ITU, 2010). Despite this high degree of ownership, people may have difficulty when selecting their phones because they have multiple goals for the phone (e.g. aesthetics, function, durability, etc.), and they may be affected by what they already know and use. While people may not be able to explicitly state all of their preferences they can make selections that reveal their preferences (Green, 1990). Using an Interactive Genetic Algorithm (IGA) it may be possible to have users make a series of choices between a set of designs and determine what their preference is for each goal (Kelly, 2008). Methods of assessing user preference have been studied widely within the marketing, design, and engineering communities (Kelly, 2008). These methods include both mathematically rigorous preference-modeling tools, such as conjoint analysis (Louviere, 2000), and less formal methods such as focus groups (Otto, 2001). With unguided prompts, it is nearly impossible for users to saliently describe their ideal design for a given product type. This is because of the essentially infinite design space presented to them when such prompts are given. Users can more easily express their preferences within a constrained design space. Phone manufacturers must maintain aggressive product release cycles, competing on a number of factors for users who have unique previous experiences and multiple goals such as aesthetics, function, and battery life. Studies have shown that aesthetics has a large impact on product assessment (Jordan, 1998). Additionally, studies have shown that the size, and shape of the phone, and the individual components have the largest affect on mobile phone aesthetics (Seva, Duh, & Helander, 2006). Within size and shape,
people vary in preference along several key dimensions. A more detailed investigation of the size and shape factors leading to preference indicates that changes in screen width and screen height account for most of the variability between peoples‟ aesthetic preference for mobile phones (NathanRoberts & Liu, 2010). As phones replace more tools in our lives, reliance on phones, and their prevalence grows. This further fuels the phone manufacturing market to become more competitive thereby reducing product lifecycle and reducing each phone‟s time to market. This shorter timeframe further increases the pressure on designers to understand what users want and incorporate those features and styles in to new designs. To be successful, designers must be able to design functional, and aesthetically pleasing phones quickly. To do this, designers often rely on heuristics, experience, and intuition, which is common to most fields of product design (Otto, 2001). This experience can be enhanced with preference assessment tools and quantitative data related to the design space of interest. One such preference assessment tool is the IGA, which can help designers or users quickly explore a design space and identify optimal design areas (Kelly, 2008). Genetic Algorithms (GAs) are useful tools for solving both optimization and search problems. They are based upon the principles of biological genetics. In nature, populations grow and change based on the ability of population members to survive. This survival of the fittest leads to a population that is well tuned to its environment. This concept can be applied to optimization and search problems in the same way. By describing a potential set of solutions to a problem in terms of a genetic code, an evolutionary process can occur in which subsequent populations of solutions can be propagated based on the „fitness‟ of the previous population. This evolution should eventually lead to a population that is well tuned to its environment. GAs have been used successfully for many applications: from optimizing the strength to weight ratio of a
PROCEEDINGS of the HUMAN FACTORS and ERGONOMICS SOCIETY 55th ANNUAL MEETING - 2011
bridge, to determining the least wasteful way to fold a box. GAs were first introduced by Holland (Holland, 1975), and there have since been many improvements in their application and implementation. For a GA to be implemented there must be a way to evaluate the fitness of each individual, or design, within a population. In the traditional GA this is done with some type of mathematical function, or „black box‟ analysis. For an IGA this is done through user input (Dawkins, 1986). IGAs have been for both visual aesthetic purposes (Kelly, 2008; Cho, 2002), as well as auditory purposes in the academic literature (Durant, 2004). Within the visual domain, IGAs have successfully been used to determine shape aesthetic preference (Kelly, 2011; Nathan-Roberts & Liu, 2010). The principles of the IGA have also been used commercially. The most recognizable instance is the Pandora Radio website, used for suggesting music that a user may like based on his or her evaluations of other music. This experiment is part of a larger set of studies focused on testing user ability to distinguish between multiple goals that will be presented in greater detail elsewhere. This portion looks specifically at how changes in the technology, or feature-set of a product can impact user design preference, while also testing the impact user biases can have on the final designs. The goal of this article is to build on previous mobile phone IGA work based on aesthetics. This article will extend the literature by testing users‟ internal biases based on multiple goals and previous experiences. To do this, the study tested user ability to select one goal (phone number dialing), and tested the effect of users‟ current phone and bias on their design preference for two phone types (touchscreen and nontouchscreen phones).
Variables Within-subjects, the experiment used the IGA variables shown in Figure 1 and the user goals shown in Table 1. Between-subjects, phone type owned by the participant; an Apple iPhone, a non- iPhone touchscreen phone, or a nontouchscreen phone, changed as well. The independent variables in Figure 1 were calculated inside the web browser of the participants‟ computers, and therefore cannot be given in actual dimensions, but instead as pixels in a browser window, or as ratios.
Figure 1. Independent variables changing within the IGA. Table 1. Prompts used for experiment two in IGA software. Goal Practice
METHOD This study tested the ability of users to distinguish between multiple goals by asking all participants to use an IGA to design touchscreen and non-touchscreen phones to be used for the same task. Between-participants, the study looked at how the kind of phone they owned affected their design preference. Participants were either Apple iPhone owners, or non-Apple iPhone/iPod users.
Participants Twenty college students, ten males and ten females participated remotely in this study. The mean age was 22 years, with a standard deviation of 2.6 years. Participants had one of three types of phones: an iPhone (10), another touchscreen phone (3), or a non-touchscreen phone (7). The inclusion criteria were: ownership or extensive use of a mobile phone; access to a computer with a high-speed internet connection, and a 19 inch to 21 inch computer monitor; and a U.S. address where payment could be mailed.
1667
Touchscreen
Non-touchscreen
Prompt “This is the first trial, and is a practice trial. The software tries to learn from your previous selections. Please experiment with the software by picking mobile phone designs that you like and see how the program responds using your selection to give you new options in the next step.” “This trial will be a touchscreen trial. On every iteration please select the phone design(s) which you find most aesthetically pleasing as a touchscreen phone for dialing phone numbers.” “This trial will be a non-touchscreen trial. On every iteration please select the phone design(s) which you find most aesthetically pleasing as a non-touchscreen phone for dialing phone numbers.”
Interactive Genetic Algorithm configuration The software used in experiment two evolved designs in a similar manner as in Nathan-Roberts & Liu, 2010. The improved interface is shown in Figure 2Error! Reference source not found.. The web tool allowed participants to flow through the procedure by navigating the study using the lefthand bar shown in Figure 2Error! Reference source not found.. In each IGA trial participants used the IGA for 10 iterations before completing a summary and moving on.
PROCEEDINGS of the HUMAN FACTORS and ERGONOMICS SOCIETY 55th ANNUAL MEETING - 2011
Procedure Participants were directed to the experiment website via a recruitment email to university department list-serves. Interested participants were qualified via a screening form on the experiment website. Once qualified, participants received information, filled out a consent form, received training, and completed the study remotely. After consent, participants were given instructions about the tool, and how to access it via email. Participants were instructed on the importance of making their choices based “only on using the phone for number dialing”, and to consider “only the current prompt.” After accessing the study website, participants received further training on the software, and the importance of following the prompt was again highlighted. After training, participants completed a series of trials with a short questionnaire after each trial and a longer debriefing survey after the last trial. Participants completed nine trials; one practice, four nontouchscreen phone selection, and four touchscreen phone selection trials. Participants first completed a practice trial and then alternated between touchscreen and non-touchscreen phone selection. Half of the participants had a touchscreen selection trial after the practice, and half had a nontouchscreen trial as their second trial. Participants were not instructed how to interpret the designs; e.g., whether the keys were physical or part of the touchscreen display. Data were collected from the server MYSQL databases and analyzed locally using Microsoft Excel and R. Pairwise wilcoxon comparisons were performed between phone prompt
1668
types as well between iPhone and non-iPhone users and touchscreen and non-touchscreen users.
RESULTS As a test for convergence, participants were asked if the phones ever started to look the same, and if so, at what generation that happened. All participants indicated that the phones started looking the same for at least seven of their nine trials. Overall, 97% of the post-trial questionnaires (n=187) said the designs started looking the same. Of the trials where participants said it converged, the reported mean generation was 5.4 (SD 1.9). Although not presented here due to space constraints, the standard deviation of the population, and the standard deviation of the selections decreased steadily, leveling off between generations five to seven as well. A pairwise wilcoxon comparison between design prompts (Table 2Error! Reference source not found.) shows that participants had significantly different preferences for touchscreen, and non-touchscreen designs for all variables except the vertical button spacing, and phone radius. The mean of design of each group can be seen in Figure 3. Table 2. p-values from pairwise wilcoxon comparisons between touchscreen and non-touchscreen design trials. Button spacing Horizontal Vertical P-value 0.003 0.039 (α = 0.01, corrected with Bonferoni)
Screen size Vertical Horizontal 0.005