Evaluation of design properties of electric and combustion cars based on eye tracking

We acknowledge the Journal of Engineering Design for the publication of 1 the article titled: Evaluation of design properties of electric and combustion 2 cars based on eye tracking. 3 The link to the final published version is provided below: 4 • https://doi.org/10.1080/09544828.2024.2355759 5 6 7 Evaluation of design properties of electric and combustion cars based 8 on eye tracking 9 Raúl Núñez a, Antonio Córdoba a, Ana de las Heras a*, Amalia Luque a 10 a Dpto. Ingeniería del Diseño. Escuela Politécnica Superior. Universidad de Sevilla. 11 Virgen de África, 7. 41011 Sevilla, Spain 12 *[email protected] 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 Evaluation of design properties of electric and combustion cars based 30 on eye tracking. 31 The current transition to the electric car by all users is advancing steadily, and 32 hence the design of these products occupies a prominent place in the car design 33 sector. Likewise, the user experience is one of the determining aspects for the 34 choice of a model. The main objective of this work is to determine, through eye 35 tracking, which areas of an image (in this case, images of the front of cars) are 36 more influential in the perception of the vehicle as either electric or combustion- 37 powered, by segmenting the most important areas of a vehicle and users' 38 perceptions thereof according to gender and other socio-demographic variables. 39 This work provides an experimental design and a way to analyze and draw 40 conclusions from an oculometry study for vehicles, as well as an initial design 41 briefing for the generation of the visual identity of electric vehicles based on the 42 perception of users for different elements of the vehicle. A methodology is 43 proposed that includes the use of tools such as ANOVAs, eye-tracking, heat 44 maps, tracking maps and fixation times. Through this data it is possible to 45 develop optimal designs aligned with the user's perception. 46 Keywords: Eye tracking; electric car; design elements; aesthetics; styling; 47 perceptual elements 48 1. Introduction 49 50 1. 1. Justification 51 The year 2021 was the fastest year of growth for electric cars worldwide. According to the 52 official data, global electric car sales amounted to 4.2 million units, 108% more than in 2020, 53 and 198% more than in 2019 (Jato Dynamics 2023). The promotion of the electric car 54 therefore provides one way to address the mounting environmental and sustainability 55 challenges facing transportation (Rifkin 2011). Seba (2014) talks extensively about the 56 transition to electric transport and how this can reduce carbon emissions and improve air 57 quality (Seba 2014). This transition to electric transport also constitutes a way to lower 58 dependence on oil and reduce greenhouse gas emissions (Smil 2016; Mattick, Williams, and 59 Allenby 2009). 60 Electrification of transport can play a major role in the transition to a more sustainable and 61 cleaner energy system (Lovins, Odum, and Rowe 2013). 62 Decarbonization in the transport sector, especially in private mobility, is one of the main 63 objectives of the European Union (EU) for the coming years. Battery electric vehicles represent 64 a promising solution to reduce pollution and emissions; However, its purchase price contributes 65 to slowing its spread. The results obtained show how the electricity consumption of electric 66 vehicles is low, even for larger vehicles, and this allows them to be considered less impactful 67 than vehicles with internal combustion engines (De Santis, Silvestri, and Forcina 2022). 68 Digalwar (Digalwar et al. 2022) shows how the Indian transportation sector powered by fossil 69 fuels is mainly responsible for creating serious problems such as greenhouse gas emissions, 70 dependence on foreign fuels, the economic burden and chronic health effects. To mitigate these 71 serious problems, electric vehicles are positioned as an alternative, green and clean technology, 72 which can potentially enable the efficient transition to a sustainable transportation system with 73 low carbon emissions and the preservation of scarce natural resources. 74 75 1. 2. Background 76 The study by Sperling (2018) explores how the combination of automated, shared, and electric 77 vehicles can enable a transition to a more sustainable, efficient, and accessible transport 78 system. Electrification of transport can significantly reduce greenhouse gas emissions and 79 improve air quality in cities, and these technologies can influence urban design and the way 80 people move around cities, with the aim of providing a holistic perspective on the transition to 81 a more sustainable and efficient mobility (Sperling 2018). 82 Given the exponential trend in the decision to purchase electric vehicles, certain factors help 83 consumers choose electric vehicles in preference to other types of vehicles, with 84 environmental, social, and economic factors all playing a role. 85 Husain et al. (2021) have studied how consumers perceive the electric car and how automotive 86 companies can design and market these vehicles to increase their market acceptance (Husain et 87 al. 2021). They analyse the evolution of electric mobility, trends in the electric vehicle market, 88 and companies' strategies to compete in this emerging market. Consumer perception of the 89 electric car is examined, including attitudes towards the technology, assessment of product 90 features, and barriers to electric vehicle adoption. Furthermore, topics such as charging 91 infrastructure, government policy, , and technological innovation in the automotive industry, are 92 also covered. 93 Giesler (2006) explores how consumers construct meanings around technology through a 94 process they call "consumer gift systems". In turn, manufacturers can use these meanings to 95 enhance market acceptance of their products. Gift systems are cultural patterns that involve the 96 transfer of material and symbolic goods between people, and that help shape the way 97 consumers perceive and use products. Using the example of the electric car, it shows how 98 consumers construct meanings around this product through a variety of gift systems, such as 99 innovation, sustainability, and belonging to a community (Giesler 2006). 100 Consumers make decisions regarding the purchase of electric vehicles and the knowledge of 101 how these decisions are made enables the design of policies and programmes to increase 102 the market uptake of electric vehicles. In this vein, Vandenbergh and Gilligan (2007) present a 103 critical analysis of the role of human behaviour in environmental and sustainability-related 104 decision-making and how this affects the uptake of electric vehicles. It is argued that 105 consumers do not always make rational decisions based on self-interest, but are influenced by 106 emotional, social, and cultural factors that can lead to unsustainable behaviour. Using the 107 example of the electric car, the authors suggest that market acceptance of these vehicles can be 108 increased through policies and programmes that promote accessibility, information, innovation, 109 and cooperation between the various stakeholders involved (Vandenbergh and Gilligan 2007). 110 Another study by Dudenhöffer (2012) analyses the opportunities and risks of electric vehicles 111 and provides a detailed assessment of the electric vehicle market. It uses a wide variety of 112 data and statistics to show how consumers perceive electric vehicles and how car 113 manufacturers can adapt to the needs of the market to increase their acceptance (Dudenhöffer 114 2012). 115 This study also confirms that the aesthetic properties of the vehicle can influence consumer 116 perceptions of whether it is electric or combustion-powered, and these perceptions can be 117 employed to improve the market acceptance of electric vehicles (Secinaro et al. 2022; Brase 118 2019). 119 The work developed by Rothenberger et al. (2012) investigates how the aesthetic properties 120 of electric vehicle design can affect consumer perception of electric vehicle range and market 121 acceptance. The study focuses on the design of the visual interface of the instrument panel of 122 an electric vehicle and assesses how different visual designs affect the perception of vehicle 123 range and consumer acceptance of electric vehicles. The results of the study suggest that a 124 carefully designed visual design can improve the user experience and market acceptance of 125 electric vehicles (Rothenberger et al. 2012). 126 There are numerous publications that have studied how the aesthetic properties of the car can 127 affect consumer perceptions of whether it is electric or combustion-powered, and how these 128 perceptions can be used to improve the market acceptance of electric vehicles. They address 129 various aspects of user experience and acceptance of electric vehicles, and propose 130 methodologies and approaches to improve usability, satisfaction, and adoption of electric 131 vehicles in different contexts and cultures (Cocron et al. 2011; Chun et al. 2019; Curtale, Liao, 132 and van der Waerden 2021; Burmann, Grzesiak, and Huang 2016; Liu, Yang, and Wu 2023). 133 Eye tracking (eye tracking or oculometry) is a technique used to measure and record the 134 movement of a person's eyes and gaze while observing visual stimuli (products, images, videos, 135 interfaces...). The main objective is to analyse how and where a person directs their visual 136 attention. Eye tracking is performed using devices called "eye trackers" that record the position 137 and movement of the eyes. These trackers can usually be stand-alone cameras or devices built 138 into special glasses. Some systems use infrared cameras to track the position of the pupils and 139 calculate where the gaze is directed (Wang et al. 2023). In our case study, an independent 140 infrared camera located in front of the respondent was used. In summary, eye tracking provides 141 detailed information about users' visual behaviour and attention. Figure 1 shows in detail what 142 the eye tracking methodology consists of. 143 144 Figure 1. Eyetracking methodology 145 146 The eye-tracking technique is applicable in various fields, including research in graphic design, 147 advertising, user interfaces, and other areas related to visual perception and attention (Punde, 148 Jadhav, and Manza 2017). 149 The book "Eye Tracking Methodology: Theory and Practice" discusses how eye tracking can be 150 employed to determine which areas of an image or design are most relevant or appealing to 151 users, and how this information can be utilised to improve visual communication and user 152 experience (Duchowski 2007). 153 There are computational models of visual attention based on the detection of saliency features 154 in the image, such as edges, contrasts, and colours. Itti, Koch and Niebur (1998) used eye- 155 tracking techniques to measure participants' eye movements while looking at natural and 156 artificial images. The results showed that the proposed visual attention model was able to 157 accurately predict the areas of the image that attracted the most visual attention of the 158 participants (Itti, Koch, and Niebur 1998). 159 In the research conducted by Pajares, Rivas-Perea and Cruz (2013), eye-tracking techniques 160 are employed to measure the distribution of participants' visual attention while observing 161 images in conditions of reduced visibility. These authors also developed a computational model 162 to predict visual perception and attention under various atmospheric conditions (Pajares, Rivas- 163 Perea, and Cruz 2013). 164 In Ramanuja (2017), an emotional design model based on emotional evaluation theory is 165 proposed. To validate the model, eye-tracking techniques were utilised to measure 166 participants' visual attention while they looked at different product designs and evaluated their 167 emotional appeal. The results showed that certain design features, such as symmetry and 168 complexity, attracted participants' visual attention more and contributed to their emotional 169 perception of design (Ramanujan 2017). 170 Eye tracking can be employed to analyse user interaction with complex visualisations, such as 171 network graphs and flowcharts. Blascheck and Ertl (2014) used eye-tracking techniques where 172 they measured participants' visual attention while exploring visualisations and developed a 173 model to determine which areas of the visualisation were most important to the task at hand 174 (Blascheck and Ertl 2014). 175 Finally, it is necessary to highlight how users' perceptions of the most important areas of a 176 vehicle may vary according to gender and other socio-demographic variables. It has been 177 studied how individual differences in perception and attention can affect the most important 178 areas of a vehicle, and how these findings can be utilised to improve design and user 179 experience. Endsley (2016) addresses the importance of understanding individual differences 180 in perception and attention when designing interactive systems, including vehicles, and 181 strategies are offered to optimise situational awareness and user experience. Furthermore, said 182 author discusses how technology, such as eye tracking, can help identify and understand these 183 individual differences (Endsley 2016). Nowadays, it is always a challenge to study and 184 interpret how gender differences can affect perception and attention in design in general and, 185 for the particular case of vehicles, to analyse how these findings can be applied in order to 186 improve the user experience since, depending on the gender of the person purchasing or 187 driving the vehicle, one can have completely different needs and user experience (K. Li, Li, 188 and Liu 2013). 189 Numerous publications have studied how cultural and socio-economic differences can affect 190 the most important areas of a vehicle, and how these findings can be put to use to improve 191 the design and user experience (Luo et al. 2024). In Thorandt et al. (2018), mixed reality 192 technologies are utilised to design and evaluate a user-centred vehicle interior (Detjen et al. 193 2017). Sullivan et al. (2015) examine how demographic differences influence perceptions of 194 design features of electric and internal combustion vehicles (Sullivan et al. 2015). Bhise et al. 195 (2019) provide an overview of ergonomics and human factors in vehicle design, and highlight 196 opportunities and challenges for the future (Bhise et al. 2019). In Thorandt et al. (2018), a 197 methodology is presented for the design of user interfaces for electric vehicles that are adapted 198 to the needs and preferences of users (Detjen et al. 2017). In Morris and Stanton (2013), 199 the existing literature on designing vehicles for older drivers is reviewed and implications for 200 future research are highlighted (Morris and Stanton 2013). 201 Some recent works have addressed real-life applications. In (Erdaş et al. 2023) structural 202 optimization methods are used to carry out the optimal design of a seat support. As a result of 203 the topology optimization, a new support design concept was created and used to optimize the 204 shape. 205 Aye (Aye et al. 2019) seeks to design the shape and size of a crash box, to optimize resistance 206 to car accidents, while the objective function is to maximize the total absorption of subject 207 energy to a mass restriction. According to this study, the application of the proposed multiple 208 surrogate model is better than using a traditional surrogate model in particular, especially for 209 constrained optimization. 210 At (Yildirim et al. 2023) experimental and numerical collision analyzes are carried out to reach 211 an optimal design of bumper beam and energy absorber for a passenger car. The models that are 212 formed by using Taguchi tables are subjected to crash analysis and the answers are obtained to 213 find an optimal design. As a result of the numerical analysis, a new model of energy absorber 214 and bumper beam with better crash behavior and weight reduction. 215 The objective of this study of Yildiz (Yıldız 2020b) focuses on designing a new thin-walled 216 energy absorber to be used in the design of electric vehicles. Presents an application of the slime 217 mold algorithm to the optimal design of automotive components. Function evaluations are 218 carried out using finite element analysis and estimated using the kriging surrogate model. 219 In this research, (Yıldız 2020a), a new hybrid salp swarm-Nelder-Mead (HSSA-NM) algorithm 220 is developed to optimize electric vehicle components. Both the Latin hypercube sampling 221 methodology and the radial basis function surrogate modeling approach are used to derive 222 constraint equations and objectives used in shape optimization. As a result, a design problem is 223 solved using HSSA-NM. 224 225 1.3. Hypotheses and objectives 226 227 parameters. In this case, the incorporation of Orthogonal Fractional Factorial Designs is 375 proposed, which are utilised to reduce the number of profiles presented to users. 376 Therefore, a reduction of the sample space is proposed using orthogonal design. Figure 3 shows 377 the design process of an orthogonal experiment. 378 With the application of this technique, it is possible to obtain a representative and suitable 379 fraction of all possible combinations of attribute levels (Xu et al. 2009). To carry out the 380 fractional factorial design, the IBM SPSS Statistics v.26.0 software is applied using the 381 “Orthogonal Design” function. The result of this application is the attainment of 49 profiles. 382 383 Figure 3. Design process of an orthogonal experiment Image adapted from J. Zhou et al. 384 (Zhou et al. 2012) 385 386 387 388 2.3. Representing the sample space 389 The vehicle samples are created by graphic representation using images of real (commercial) 390 vehicles. In these graphical representations, the background of the image, brand logos and 391 badges, number plate data, and vehicle colour are removed, since these are elements that fall 392 outside the scope of the study that may influence the perception and opinion of the respondents, 393 thereby distorting the objective of the experiment. 394 395 2.4. Defining AOIs 396 For each stimulus (vehicle) to be analysed, areas of interest (AOIs) must be differentiated. The 397 AOIs are employed to divide the images that represent the stimuli and to obtain specific data of 398 zones or areas of the graphic representation that help to improve the results by providing a 399 greater level of detail. These areas must be clearly delimited in each of the images obtained. 400 401 2.5. Conducting experiments 402 The number of respondents has been based on a study on similar experiments, as detailed in the 403 discussion section entitled "Regarding the sample size used for the experiment ". It must be 404 considered that this is a pilot study carried out at the university level and at this time there is no 405 professional or business support that may entail extrapolating the experiment to a larger 406 population. For reasons of budget, space and time, a larger sample could not be included. 407 Numerous reference texts have been analysed and are found to seldom use a very large sample 408 size. Therefore, it is considered that the sample size used herein of 33 participants is highly 409 suitable for this type of experiment. 410 The results of eye tracking do not depend so much on the number of respondents but on the 411 population under study, that is, on the demographic characteristics, the participants' previous 412 experiences and interest in the product, since these can influence their gaze patterns. . For 413 example, age, gender, experience and hobby in relation to the automotive field can be relevant 414 variables. 415 It must be said that an eye tracking study can be influenced by the visual stimuli, the 416 environmental context, the eye tracking technology, the design of the experiment, the data 417 analyses used and the objectives of the study. The experiment for all participants has been 418 carried out under the same environmental context, with the same eye tracking technology and 419 with the same visual stimuli, following the same experimental procedure and sharing the same 420 objective. The participants are all legal age and are familiar with the vehicles. 421 At the beginning of the experiment, the researchers explain the test procedure as well as the 422 objectives of the study prior to the experiment. The users are also asked in writing for their 423 consent and informed that the data collected is anonymous and will only be used for this study 424 and that they can leave the test at any time. 425 426 2.5.1 User test 427 A user test is carried out based on the closed interview education technique: in this case a multi- 428 choice questionnaire with different single response alternatives is used. 429 The purpose of the test questions is twofold: 430 431 432 • P1 and P2: categorisation of the user based on gender and age. 433 • Q3 to Q7: analysis of the relationship between the respondent and the object of study. 434 Variables Categories Sex Female Male Age Between 18 and 21 Between 22 and 25 Over 25 Driving licence Yes In process Ecological choice Yes No Purchase preference for a 100%-electric vehicle Price Automatic Cargo volume Brand Power Desire to purchase a 100%-electric vehicle Yes No Ability to differentiate a 100%-electric vehicle Yes In doubt No Table 1. Variables and categories of the user test 435 436 Due to the anonymous nature of the survey, instead of directly asking the users their age, the 437 user test presents 3 age groups (between 18 and 21, between 22 and 25, and over 25). 438 On the other hand, the survey, carried out mainly with students in these age ranges, can give us 439 information on whether they are probably in the 1st or 2nd academic year (18 to 21 years old), 440 in the 3rd or 4th year (22 to 25 years old), or whether these are respondents who are probably 441 not students (over 25 years old). 442 Out of the 33 participants, 8 are between 18 and 21 years old, 9 are between 22 and 25 and 16 443 are over 25 years old (see Table 3). 444 2.5.2. Eye tracking 445 At the beginning of the eye-tracking experiment, prompts are provided to participants before 446 they view each vehicle stimuli. The instructions provided out loud to the participants are read by 447 the researcher and always provided in the same way. Specifically, the prompts are: 448 "In this part you will be presented with images, and you must respond out loud whether it is an 449 electric car for you. Respond YES or NO when the image appears so that your response can be 450 noted by the researcher". 451 This is followed by an eye-tracking experiment (Wang et al. 2023), in which respondents are 452 shown the images, consisting of 49 stimuli. 453 The experiment is carried out using the OGAMA software (Voßkühler 2009) and the 454 Gazepoint GP3 oculometry equipment (Sarraipa et al. 2019). The experiment settings have 455 been: 456 ▪ Screen size of eye-tracker monitor is "x size"=1024 pixels and "y size"=768 457 pixels. 458 ▪ Sampling rates = 60 Hz. 459 ▪ Fixation calculation parameters: 460 o Maximum distance in pixels that a point may vary from the 461 average fixation point and still be considered part of the fixation 462 = 60. 463 o Minimum number of samples that can be considered a fixation = 3. 464 ▪ Eliminate first fixation if is at the same place as the last fixation of the 465 foregoing trial and if it is shorter than 100ms. 466 Each stimulus (vehicle image) is displayed for 5000ms. Between stimuli, a blank screen 467 appears for 1000ms. 468 The total duration of the oculometry experiment is 293000ms (4.9 minutes). In 469 this experiment, the data collected includes: 470 • Fixations (count) for all AOIs 471 472 • Total Fixation Duration Mean (ms) for all AOIs 473 474 • Fixation duration mean at each AOI and Number of fixations (count) at 475 each AOI 476 • % of total area of total picture area covered by the area of interest (AOIs) 477 478 479 2.5.3. Vehicle-type identification 480 481 During the eye-tracking experiment, in which respondents are shown images of different 482 vehicles, they are asked to answer whether it appears to be an electric or combustion-powered 483 car. 484 In this experiment, the data collected includes: 485 486 • Hits and misses in vehicle recognition 487 488 2.6. Descriptive analysis of the results 489 For the user test, an analysis of response frequencies, a correlation analysis between study 490 variables, and ANOVAs between the variables that analyse the relationship between the 491 respondent and the object of study, and the categorisation variables are carried out. 492 From the eye-tracking experiment, in the second phase of the proposed method, the number of 493 fixations per AOI and the fixation duration per AOI are analysed, thereby obtaining the 494 mean value, standard deviation, and maximum value, and heat maps and eye-tracking maps. 495 Fixations are moments when the eyes are relatively still, focused on a specific area. In our case 496 study, a fixation was counted when the eye spent 300 ms in a specific area relatively still. The 497 number of fixations in a specific area is called "Fixation count". The image that is generated to 498 visualize the areas of the screen or image that receive more visual attention, that is, that receive 499 more numbers of fixations, is called "Heat map". 500 Correlations between parameters of the eye-tracking experiment are attained. 501 From the third proposed phase, an ANOVA is performed to study the relationship between the 502 numbers of hits and misses in the eye-tracking test and the type of vehicle (electric or 503 combustion-powered). 504 The Analysis of Variance (ANOVA) is based on the statistical assumptions of independence, 505 normality, and homoscedasticity of the experimental responses (Villanueva, Petenate, and Da 506 Silva 2000). It has been proven that the data meets ANOVA assumptions. 507 Specifically, the independence condition means that the observations in each group are 508 independent of each other and the observations within groups were obtained by a random 509 sample. Since, to the best of the authors’ knowledge, there is no formal test available to verify 510 that the observations in each group are independent and that they were obtained by a random 511 sample, then the only way this assumption can be satisfied is if a randomised design is used. In 512 our work, independence refers to the fact that a sample of fixations of one individual is 513 independent of that of another, and that a randomised design has been used. 514 The ANOVA analysis were carried out using SPSS, with the one-factor ANOVA analysis 515 module. The database has 32 case studies, each of them corresponding to the respondents (one 516 of them was discarded due to an error in the measurement of one of the eyetracking parameters). 517 There are two groups of variables: 518 1) Qualitative or nominal variables: user test questions 519 2) Quantitative variables or scale: measurements obtained from eyetracking. 520 To treat the quantitative variables, it has been worked with the average of the responses 521 provided by each respondent. Working with the means of the responses is positive to guarantee 522 the independence of the data entered in the ANOVA analysis. 523 When carrying out the one-factor ANOVA, the quantitative variables obtained from eyetracking 524 were used as dependent variables, such as the number of fixations and fixation time. The 525 qualitative variables obtained from the user test were used as factor variables, such as sex, 526 purchase preference, etc. 527 2.7. Study of the relative importance of AOIs 528 To analyse the relative importance of AOIs, the number of fixations and the average fixation 529 time are analysed based on the size or extent of the AOIs, using two indicators. These indicators 530 are our own proposal based on the study by Hyun et al. (2017) (Hyun, Lee, and Kim 2017): 531 𝐼𝑛𝑑𝑒𝑥 𝐴 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑓𝑖𝑥𝑎𝑡𝑖𝑜𝑛𝑠 %𝐴𝑟𝑒𝑎 (1) 532 𝐼𝑛𝑑𝑒𝑥 𝐵 = 𝑇𝑖𝑚𝑒 𝑜𝑓 𝑓𝑖𝑥𝑎𝑡𝑖𝑜𝑛𝑠 %𝐴𝑟𝑒𝑎 (2) 533 Index A and Index B are indices that analyse the importance of the AOIs based on the number 534 of fixations they receive (Index A) and based on the total duration of the fixations (Index B). 535 The two indices are directly related, since the greater the number of fixations an AOI receives, 536 the greater the gaze retention time at that AOI. These indices serve to measure the probabilities 537 of fixations per area and thus determine the meaning and visual importance of each area. They 538 are useful for making direct comparisons of one AOI with another and establishing rankings. 539 3. Results 540 The main results obtained are shown below, following the methodology described in the 541 previous section and in Figure 2. 542 3.1. Experiment design 543 In order to graphically represent the proposed attributes and levels, the morphological chart 544 creativity technique has been applied, since it enables the graphical analysis of the attributes of 545 the product, its variants, and their possible combinations. 546 The attributes and levels of the proposed product have been defined. An analysis is 547 subsequently carried out to decide which properties (areas of the vehicle) are to be studied, 548 while taking into account that the analysis is based on the front view of the vehicle. For each 549 view, the set of different levels to be considered are determined. 550 In this case, a sample space is obtained that comprises the different levels of the following 551 properties: 552 • Headlights shape (5 levels) 553 • Headlight separation (2 levels) 554 • Grille shape (6 levels) 555 • Frontal shape (5 levels) 556 • Bonnet shape (3 levels) 557 • Bonnet nerve (3 levels) 558 The combination of these properties and levels constitutes a sample space of 2,700 in size. 559 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑠 = 5 × 2 × 6 × 5 × 3 × 3 = 2,700. 560 The sample space obtained is too large to be tackled in a survey. The reduction of the sample 561 space is carried out (through orthogonal design, performed using the IBM SPSS Statistics v.26 562 software). 563 The orthogonal design experiment carried out with the SPSS software proposed 49 564 combinations of attribute levels based on the proposed morphological chart. The design team 565 decided to have an equal representation of electric (n=24) and fossil-fuel (n=25) vehicles in the 566 proposed sample space (N=49). 567 568 Variable Categories n % Vehicle propulsion system Electric 24 49.0 Fossil fuel 25 51.0 Table 2. Vehicle sample characteristics (N=49). 569 The graphical representation of the 49 samples obtained is then generated, and for each stimulus 570 (vehicle) the areas of interest (AOIs) are differentiated. For the analysis of the front view of the 571 vehicle, six AOIs have been defined: windscreen, bonnet, right headlight, left headlight, grille, 572 and frontal. These areas are clearly delimited in each of the images obtained. An example of a 573 sample represented graphically is shown in Figure 4, with its delimited AOIs. There are 49 574 samples like this. 575 576 Figure 4. Definition of AOIs on one of the images used in the experiment frontal. 577 3.2. Experiments and descriptive analysis of the results 578 At the beginning of the experiment, the researchers explain the test procedure and the objectives 579 of the study. 580 Three data collections from the participants have been carried out and therefore three types of 581 results are obtained. In addition to the analysis of each experiment, analysis is carried out to 582 study possible relationships between the data from the three experiments. 583 3.2.1 User test 584 In the proposed experiment, 33 respondents of legal age from the university environment have 585 participated, with representatives of both sexes and of various age ranges. 586 The results obtained from the 7 questions asked in the user test survey are reflected in Table 3 587 below. 588 Variables Categories n % Sex Female 17 51.5 Male 16 48.5 Age Between 18 and 21 8 24.2 Between 22 and 25 9 27.3 More than 25 16 48.5 Driving licence Yes 30 90.9 In process 3 9.1 Ecological choice Yes 19 57.6 No 14 42.4 Purchase preference for a 100%-electric vehicle Price 6 18.2 Automatic 19 57.6 Cargo volume 5 15.2 Brand 1 3.0 Power 2 6.1 Desire to purchase a 100%-electric vehicle Yes 19 57.6 No 14 42.4 Ability to differentiate a 100%-electric vehicle Yes 7 21.2 In doubt 17 51.5 No 9 27.3 Table 3. Participant characteristics (N=33) 589 A correlation analysis has been carried out between the variables of the user test and it has been 590 established that there is a slightly statistically significant correlation at the 0.05 level (bilateral) 591 between: 592 • Age (P2) and possession of the driving licence B1 (P3): r2,3=-0.352; sig2,3=0.045. 593 • Purchase preference for a 100%-electric vehicle (P5) and the ability to differentiate a 594 100%-electric vehicle (P7): r5,7=-0.351; sig5,7=0.046. 595 It can be affirmed that all the other variables of the user test are independent of each other. 596 3.2.2. Eye tracking 597 An eye-tracking experiment was subsequently carried out, in which the images were shown to 598 the respondents. One of the respondents has been eliminated from the analysis due to errors in 599 the data collection from the eye trackers, thereby leaving 32 respondents: 16 men and 16 600 women. The oculometry experiment consists of the visualisation of 49 stimuli. 601 The following variables are obtained from this experiment: 602 • ANOVA between mean duration of fixations (ms) in AOIs vs. P4: Ecological choice 720 ANOVA: the mean fixation time in the AOIs presents significant differences in all the areas of 721 interest except in the “Frontal” AOI, with a confidence level of 95%, with respect to ecological 722 awareness (P4). 723 As a general conclusion, those surveyed who are not ecologically aware when purchasing 724 vehicles have higher average fixation times. 725 • ANOVA between mean duration of fixations (ms) in AOIs vs. P5: Purchase preference 726 for a 100%-electric vehicle. 727 ANOVA: the mean fixation time in the AOIs presents significant differences in all the areas of 728 interest with the exception of the AOI "Windscreen", with a confidence level of 95%, with 729 respect to purchase preference (P5). 730 Users whose purchase preference is the "Brand" have had a longer average fixation time in the 731 AOIs “Right headlight”, “Left headlight”, and “Bonnet”. For users whose preference is "Price" 732 they have had a greater fixation for the AOI "Grille". 733 • ANOVA between mean duration of fixations (ms) in AOIs vs. P6: Desire to purchase a 734 100%-electric vehicle. 735 ANOVA: the mean fixation time in the AOIs presents significant differences in all the areas of 736 interest with the exception of the AOI “Right headlight” and “Grille”, with a confidence level of 737 95%, compared to the purchase of a 100%-electric vehicle (P6). 738 As a general conclusion, respondents who plan to purchase a 100%-electric car have higher 739 average fixation times. 740 • ANOVA between mean duration of fixations (ms) in AOIs vs. P7: Ability to 741 differentiate a 100%-electric vehicle. 742 ANOVA: the mean fixation time in the AOIs presents significant differences in all areas of 743 interest, with a confidence level of 95%, with respect to the ability to differentiate a 100%- 744 electric vehicle (P7). 745 As a general conclusion, those surveyed who have answered that they are able to ascertain 746 whether a vehicle is electric or not by its aesthetic-formal appearance have higher mean fixation 747 times. Respondents who have answered that they are unable to ascertain whether a vehicle is 748 electric or not due to its aesthetic-formal appearance have lower average fixation times. 749 3.3.3. Vehicle-type identification 750 During the eye-tracking experiment, the results of which are described in the previous section, 751 respondents were asked to answer whether they think it is an electric or fossil-fuel car. In this 752 experiment, the data collected relate to hits and misses in vehicle recognition. 753 Summary Mean SD Max. Min. Hits 33.7 4.07 40 22 Misses 15.3 4.07 27 9 Table 8. Summary of hits and misses per user 754 Only one respondent has a hit rate below 50%. 755 756 757 758 Table 9. Summary of hits per vehicle propulsion system 759 Having carried out a series of ANOVAs, the following results are obtained: 760 • The number of hits and misses in the eye-tracking test is independent of the type of 761 vehicle (electric or fossil fuel). The average number of hits in fossil-fuel vehicles is 762 slightly higher than the average number of hits in electric vehicles. 763 • The number of correct answers presents significant differences, with a confidence level 764 of 95%, regarding the factor Ability to differentiate a 100%-electric vehicle (P7). The 765 average number of correct answers is higher in those surveyed who have answered that 766 they are able to recognise an electric vehicle. The average number of correct answers is 767 Vehicle propulsion system n Mean SD Min. Max. Electric 24 21.00 8.612 3 31 Fossil fuel 25 24.32 6.095 11 33 Total 49 22.69 7.545 3 33 less for the respondents who have answered that they are unable to recognise an electric 768 vehicle. This variable (P7) is the only variable related to the number of hits, which is an 769 indicator that users have answered honestly regarding their recognition ability. 770 • The number of hits presents significant differences, with a confidence level of 95%, 771 with respect to the number of fixations on the grille (Number of fixations in AOI: 772 Grille). The mean number of fixations in the AOI: Grille is significantly higher in 773 vehicles that have missed than in vehicles that have hit (see Figure 7). 774 775 Figure 7. Number of fixations in AOI: Grille 776 3.4. Study of the relative importance of AOIs 777 As indicated in the methodology section, in order to analyse the relative importance of the 778 AOIs, the number of fixations and the average fixation time based on the size or extension of 779 the AOIs are analysed through two indicators, which provide the main innovation of this work. 780 To this end, firstly, a correlation analysis has been carried out: 781 Variables to correlate: 782 • SIZE_AOI 783 • Complete fixation time in AOI (ms). 784 • Number of fixations in AOI. 785 • Fixation duration mean in AOI. 786 787 All correlations are significant, especially that of the number of fixations in the AOI with the 788 size of the AOI. Taking into account the size of the AOI, it should be noted that this variable 789 has a correlation coefficient of 0.707 with the number of fixations in the AOI, which is 790 an indicator that the number of fixations, as well as their average duration, is highly 791 correlated with the AOI size. 792 793 *All the following=0.00 794 Table 10. Correlation matrix using Pearson’s correlation coefficient. 795 Figures 8 and 9 below show the results of the proposed indices A and B for the various AOIs. 796 797 Figure 8. Cumulative number of fixations for each AOI 798 The AOIs that accumulate the greatest number of fixations per area size (Figure 7) are: 799 SIZE_AOI Complete fixation time in AOI (ms) Number of fixations in AOI Fixation duration mean in AOI Correlation SIZE_AOI 1.000 0.640 0.707 0.334 Complete fixation time in AOI (ms) 0.640 1.000 0.984 0.744 Number of fixations in AOI 0.707 0.984 1.000 0.679 Fixation duration mean in AOI 0.334 0.744 0.679 1.000 1. Grille 800 2. Left headlight (F2) 801 3. Right headlight (F1) 802 Likewise, the AOIs that accumulate the longest fixation time per area size (Figure 8) are: 803 1. Left headlight (F2) 804 2. Grille 805 3. Right headlight (F1) 806 It can be determined that these areas are the most important for users when deciding whether a 807 vehicle is electric or combustion-powered. 808 809 Figure 9. Accumulated fixation time for each AOI 810 4. Discussion 811 4.1. Regarding the sample size used for the experiment. 812 This section presents a discussion based on certain reference texts on the topic under study. 813 Regarding the design of the eye-tracking experiment, 10 texts have been analysed (see Table 814 11). 815 Reference Content Sample size and composition (Luo et al. 2022) Research on the aesthetic preferences of users for the design of the shape of the rear lights of vehicles through eye tracking. n= 46 (23 women, 23 men) (Hyun, Lee, and Kim 2017) Evaluation of automobile exterior designs through eye tracking to obtain shape similarities and novel design elements. n= not specified (Lin, Guo, and Xu 2019) Investigation of the perceived experiences through the formal aesthetic design of intelligent vehicles through eye tracking and semantic analysis. n= 89 (Wickman et al. 2014) Eye-tracking analysis of the influence of assembly lines between bodywork elements on the perception of a vehicle. n= 31 (5 women, 26 men) (Lu, Xu, and Cheng 2018) Objective evaluation by users on the formal design of vehicles through eye tracking for the extraction of properties. n= 11 (7 men, 4 women) (Moon et al. 2021) Analysis by eye tracking and electroencephalography of the visual impression of vehicle fronts. n= 12 (10 men, 2 women) (Purucker, Sprott, and Herrmann 2014) Investigation through eye tracking of user reactions to vehicle fronts with "aggressive" ("threatening") design. n= 39 (13 women, 26 men) (Chang, Chu, and Ma 2013) Eye-tracking analysis of the front appearance of cars to analyse the main areas of interest. n= 20 (Köhler, Falk, and Schmitt 2015) Product design by Kansei Engineering with eye-tracking techniques to evaluate the perception and areas of interest of users. n= not specified (Köhler, Falk, and Schmitt 2013) Analysis and identification of relevant components of the product and linking with semantic concepts using eye tracking. n= 11 Table 11. References 816 No consensus exists on the exposure time of each stimulus or image, although it can be 817 established as varying between 3 and 20 seconds. Several texts include a blank image between 818 images for usually between 0.1 and 5 seconds. Concerning the definition of how long a fixation 819 lasts is included in only one text, which establishes it as being 100ms. There are studies in 820 which the respondent must make a decision as to whether the stimulus they see is "type A" or 821 "type B", for which two letters on the keyboard are used. 822 In order to prevent the respondent from looking at the keyboard when they press a key, thereby 823 losing information in the eye tracking, our experiment commands the researcher to write down 824 what they think without looking up from the monitor. 825 Regarding the sample size, the texts analysed seldom use a very large size, and vary between 826 approximately 11 and 89 participants, the majority being at about 25 respondents. We therefore 827 consider the sample size used (33 participants) to be highly suitable for this type of experiment. 828 4.2. Suggestions to develop optimal designs aligned with user perception. 829 According to the capital value pyramid of Keller's customer‐based brand equity model 830 (see Figure 10) (Keller 2008), brand prominence measures the awareness thereof; for 831 example, how often and how easily the brand is evoked in different situations or 832 circumstances. Brand awareness refers to the customer's ability to remember and 833 recognise the brand under a variety of conditions, with certain associations in memory. In 834 particular, building brand awareness helps the consumer understand the product or 835 service category in which it competes and what products or services are sold under its 836 name. It also ensures that customers know which of their “needs” (through these 837 products) the brand is designed to satisfy. 838 839 Figure 10. Capital value pyramid of Keller’s customer‐based brand equity model. Adapted 840 from (Keller 2008) 841 Brand awareness gives the product an identity by linking its elements with a product 842 category, associated purchase, and situations of use or consumption. The depth of brand 843 awareness measures how likely a brand element is to come to mind, and how easily it 844 does so. Breadth in turn measures the variety of purchase and use situations in which the 845 brand element comes to mind and further deepens the brand’s organisation and product 846 knowledge in memory (Branch 2001). 847 848 This work aims to provide ideas and suggestions to develop optimal designs aligned with 849 user perception. As argued by Branch (2011), building brand awareness helps the 850 consumer understand the category of the product or service in which it competes and 851 what products or services are sold under its name, thereby granting the product an 852 identity by linking its elements with a category of products, associated purchase, and 853 situations of use or consumption. Therefore, the identity of the electric vehicle must focus 854 on enhancing those properties that give the electric car a differentiating character. 855 856 In the previous section it is shown that the AOIs that accumulate the greatest number of 857 fixations per area size and the longest fixation time are the Grille, the Left headlight, and 858 the Right headlight. These areas are therefore the most important for users when deciding 859 whether a vehicle is electric or combustion-powered. 860 861 The suggestions to develop optimal designs aligned with user perception are aimed at 862 focusing on enhancing those properties that give the electric car a differentiating 863 character, that is, they focus the design of electric cars on those areas most important for 864 users when deciding whether a vehicle is electric or combustion-powered: the Grille, the 865 Left headlight, and the Right headlight. 866 867 5. Conclusions 868 Promoting the electric car is one way to address the environmental and sustainability challenges 869 facing transport, since the electrification of transport can play a major role in the transition 870 towards a more sustainable and cleaner energy system. 871 There are numerous publications that have studied how the aesthetic properties of the car can 872 affect consumer perception of whether it is electric or combustion-powered, and how these 873 perceptions can be utilised to improve market acceptance of electric vehicles (Hypothesis 1). 874 In this work, we have studied how individual differences in perception and attention can affect 875 the most significant areas of a vehicle, and how these findings can be employed to improve the 876 design and user experience. Eye tracking has been utilised to analyse user interaction with 877 complex visualisations, such as network graphs and flowcharts. 878 Through the application of eye tracking, it has been determined which areas of an image (in this 879 case images of vehicles) are more influential in the perception of it as either being electric or 880 combustion-powered (Hypothesis 2) 881 It has been analysed how the most important areas of a vehicle and the perceptions of users 882 regarding said vehicle can differ depending on gender and other socio-demographic variables 883 (Hypothesis 3). 884 Numerous reference texts have been analysed and are found to seldom use a very large sample 885 size. Therefore, it is considered that the sample size used herein of 33 participants is highly 886 suitable for this type of experiment. 887 To determine the relative importance of the AOIs, the number of fixations and the average 888 fixation time based on the size or extent of the AOIs are analysed through two indicators. These 889 indicators provide the main innovation of this work. 890 The AOIs that accumulate the most fixations per area size are the grille and the two headlights. 891 It can therefore be determined that these areas are the most important for users when deciding 892 on whether a vehicle is either an electric or a combustion-powered car. 893 This work provides an experimental design and a way to analyze and draw conclusions from an 894 oculometry study for vehicles, as well as an initial design briefing for the generation of the 895 visual identity of electric vehicles based on the perception of user for different elements of the 896 vehicle. 897 Acknowledgements 898 This work has been supported by the GOYA – Antonio Unanue in Engineering in the Agrifood 899 Industry Chair of the University of Seville (Spain) and the VII Research and Transfer Plan 2023 900 of the University of Seville (Projects 2023/00000378 and 2023/00000390). 901 References 902 Aye, Cho Mar, Nantiwat Pholdee, Ali R. Yildiz, Sujin Bureerat, and Sadiq M. Sait. 2019. 903 “Multi-Surrogate-Assisted Metaheuristics for Crashworthiness Optimisation.” 904 International Journal of Vehicle Design 80 (2/3/4): 223. 905 https://doi.org/10.1504/IJVD.2019.109866. 906 Baroroh, Dawi Karomati, Mya Amalia, and Nur Puji Lestari. 2019. “Kansei Engineering 907 Sperling, Daniel. 2018. Three Revolutions. Washington, DC: Island Press/Center for Resource 1077 Economics. https://doi.org/10.5822/978-1-61091-906-7. 1078 Sullivan, John P., Shannon Chance, Jing Zhang, Michael J. Currano, and K. Jo Min. 2015. 1079 “Perceptions of Traditional and Electric Vehicle Design Features: The Influence of 1080 Gender, Age and Ethnicity.” SAE International Journal of Passenger Cars - Mechanical 1081 Systems 8 (2). 1082 Sun, J. 2007. Research on Method of Image-Based Produce Form Design. Wuhan University of 1083 Technology. Wuhan: Spirnger. 1084 Suparmadi, Yoko, Slamet Riyadi, and Deny Willy Junaidy. 2021. “Indonesian Consumer 1085 Preference on Electric Motorcycle Design with Kansei Engineering Approach.” Journal of 1086 Visual Art and Design 13 (1): 1–17. https://doi.org/10.5614/j.vad.2021.13.1.1. 1087 Vandenbergh, Michael P., and Jonathan M. Gilligan. 2007. “Consumer Choice and the 1088 Environment: A Behavioral Perspective on the Tragedy of the Common.” University of 1089 Chicago Law Review 74 (4): 1207–49. 1090 Villanueva, Nilda D.M, Ademir J Petenate, and Maria A.A.P Da Silva. 2000. “Performance of 1091 Three Affective Methods and Diagnosis of the ANOVA Model.” Food Quality and 1092 Preference 11 (5): 363–70. https://doi.org/10.1016/S0950-3293(00)00006-9. 1093 Voßkühler, A. 2009. OGAMA Description (for Version 2.5). A Software to Record, Analyze and 1094 Visualize Gaze and Mouse Movements in Screen Based Environments. Berlin, Germany: 1095 Freie Universität. http://www.ogama.net/sites/default/files/pdf/OGAMA- 1096 DescriptionV25.pdf. 1097 Wang, Yong, Fanghao Song, Yan Liu, Yaying Li, Xiaomin Ma, and Weihao Wang. 2023. 1098 “Research on the Correlation Mechanism between Eye-Tracking Data and Aesthetic 1099 Ratings in Product Aesthetic Evaluation.” Journal of Engineering Design 34 (1): 55–80. 1100 https://doi.org/10.1080/09544828.2023.2172662. 1101 Wickman, Casper, Ola Wagersten, Karin Forslund, and Rikard Söderberg. 2014. “Influence of 1102 Rigid and Non-Rigid Variation Simulations When Assessing Perceived Quality of Split- 1103 Lines.” Journal of Engineering Design 25 (1–3): 1–24. 1104 https://doi.org/10.1080/09544828.2013.875522. 1105 Xi, Le, Jianxin Cheng, Yixiang Wu, Junnan Ye, and Wangqun Xiao. 2017. “Research on Form 1106 Attractiveness of Electric Vehicle.” In Transdisciplinary Engineering: A Paradigm Shift, 1107 766–73. IOS Press book. https://doi.org/10.3233/978-1-61499-779-5-766. 1108 Xu, Qianli, Roger J. Jiao, Xi Yang, Martin Helander, Halimahtun M. Khalid, and Anders 1109 Opperud. 2009. “An Analytical Kano Model for Customer Need Analysis.” Design Studies 1110 30 (1): 87–110. https://doi.org/10.1016/j.destud.2008.07.001. 1111 Yildirim, Ahmet, Emre Demirci, Selçuk Karagöz, Şevket Özcan, and Ali Riza Yildiz. 2023. 1112 “Experimental and Numerical Investigation of Crashworthiness Performance for Optimal 1113 Automobile Structures Using Response Surface Methodology and Oppositional Based 1114 Learning Differential Evolution Algorithm.” Materials Testing 65 (3): 346–63. 1115 https://doi.org/10.1515/mt-2022-0304. 1116 Yıldız, Betl Sultan. 2020a. “Robust Design of Electric Vehicle Components Using a New 1117 Hybrid Salp Swarm Algorithm and Radial Basis Function-Based Approach.” International 1118 Journal of Vehicle Design 83 (1): 38. https://doi.org/10.1504/IJVD.2020.114779. 1119 ———. 2020b. “Slime Mould Algorithm and Kriging Surrogate Model-Based Approach for 1120 Enhanced Crashworthiness of Electric Vehicles.” International Journal of Vehicle Design 1121 83 (1): 54. https://doi.org/10.1504/IJVD.2020.114786. 1122 Zhang, Fan, and Jian Wang. 2013. “Application of Kansei Engineering in Electric Car Design.” 1123 Applied Mechanics and Materials 437 (October): 985–89. 1124 https://doi.org/10.4028/www.scientific.net/AMM.437.985. 1125 Zhou, Ji, Ran An, Heng Zhang, and Yongqiong Liu. 2012. “Orthogonal Design of 1126 Pharmaceutical Experiment Based on SPSS.” In , 552–60. https://doi.org/10.1007/978-3- 1127 642-34041-3_77. 1128 1129