Keywords

1 Introduction

According with Orbis Research, the global market of the weight loss industry was USD$ 168.95 billion in 2016 and it is projected a growth to USD$ 278.95 in 2023 [1]. These companies have a large budget for advertising their weight loss products, although, as has been identified by the Federal Trade Commission (FTC), there are many deceptive ads within them [2,3,4].

Many consumers are aware and can identify deceptive weight loss ads. However, even the skeptic consumers are prone to buy weight loss products in Internet with the hope to lose weight [2]. For instance, adolescents, women with appearance concerns, and obese people, are the most vulnerable consumers for this kind of advertising [2, 5, 6]. Specifically, the most representative segment of consumers for these products are young adult women (below 30 years), representing up to 40% of all consumers [7]. In most of these cases, the consumers are buying these weight loss products because of their concerns with their body weight and shape.

These concerns with body weight and shape are studied under the term of body dissatisfaction, i.e., individual’s cognitive and affective negative evaluations of his/her own body and its characteristics [8]. Unfortunately, body dissatisfaction is considered an important predictor in the development of eating disorders as well as risky eating behaviors and patterns, such as extreme diets for weight loss [9].

As evidenced by previous studies, the effect of advertising on consumer’s diet and eating behavior is so large that it has been proposed a causal relationship between advertising and diet [10]. Particularly in the case of women with high levels of body dissatisfaction, it is important to highlight that the continuous exposure to this kind of advertising in Internet and other media, can promote unhealthy eating patterns and even the relapse in women recovering of the treatment for an eating disorder [10,11,12,13]. For that reason, it is necessary to scientifically examine the effect of the online advertising of diet and weight loss products in the most vulnerable individuals, such as young women with body dissatisfaction.

Several studies have explored the effect of online advertising on body image attitudes and eating behaviors to lose weight, in both cross-sectional and experimental studies [14]. However, to the best of our knowledge, there is not such study as the one we present here, which includes a biomarker, a behavioral marker, and experimental stimuli established by computer algorithms.

In this study, the biomarker is obtained by measuring facial muscles activity (physiological response), using facial electromyography (fEMG), whereas the behavioral marker is obtained by measuring eye movements (eye fixations), using eye tracking. fEMG has been used in media and advertising research, being considered a useful measure of emotional valence and arousal [15, 16]. Similarly, eye tracking has been used in studies of digital online advertising, and it is considered an objective measure of gaze patterns [17, 18]. Regarding computer algorithms, most online advertising is established by a series of rules (i.e., algorithms) that personalize the content for each Internet user. These algorithms collect information about our online identity and behavior (e.g., sociodemographic characteristics, online searches we do, online content we download, “liked” content, etc.), to personalize the content displayed while we browse the Internet [19]. Because of it, this algorithmic regulation has been described as a “filter bubble”, because most content we see on the Internet is actually reflecting our idiosyncratic interests, desires, and needs [20].

Following the algorithmic regulation paradigm, this study presents a research protocol to experimentally examine the effect of online weight loss advertising in the attention (eye fixation) and physiological response (electrodermal activity) of women with different levels of body dissatisfaction. The current protocol describes the methodology for: participants’ recruitment; collecting weight loss ads; and the experimental study, which includes the stimuli (ads) and the responses (eye fixations and fEMG).

2 Methodology

Regarding the research design, we propose an experimental design with a matched control group. In this scenario, participants are randomly assigned to either the control or experimental condition and the control group will match with the experimental group in their sociodemographic characteristics [21].

2.1 Participants

Sampling is non-random and participants (young women between 18 and 30 years old) are invited to participate in the experiment. The invitation is sent by email to university students and it is also posted on Facebook groups of students.

2.2 Instruments

Sociodemographic Questionnaire.

Participants are asked to report their age, highest educational attainment, current occupation, weight, and height. The latter two are used to calculate the body mass index (BMI).

Body Dissatisfaction.

In this study we will use three measures of body dissatisfaction. The first measure is the Photographic Figure Rating Scale (PFRS), which consists of pictorial stimuli composed by gray scale photographic figures of women [22]. Participants are asked to choose an image that represents their ideal body, and an image that represents their actual body. Thus, this instrument measures the discrepancy between the ideal body and the perceived body (i.e., body dissatisfaction), using culturally neutral stimuli. The second instrument includes a single-item measure of global body dissatisfaction (“How satisfied are you with the physical appearance of your body?”) that uses a 5-point Likert scale (from “Not satisfied at all” = 1 to “Absolutely satisfied” = 5). Moreover, this instrument asks the same to participants but regarding different body parts (e.g., stomach, buttocks, etc.). The third instrument is the Body Dissatisfaction subscale of the Eating Disorders Inventory - 3, which is a renowned instrument used in hundreds of studies [23].

Eye-Tracking.

To measure eye movements, we propose the use of Pupil Labs binocular glasses (https://pupil-labs.com), which have good accuracy and precision [24]. Moreover, this eye tracker has a sampling frequency of 200 Hz, which is an acceptable sampling rate for available algorithms to detect eye fixations [25]. Although most studies apply mathematical and statistical procedures to handle eye tracking data [26], for this protocol we propose using also visualization techniques that allow, for example, to identify areas of interest [27]. The protocol of data acquisition follows recommended guidelines for eye tracking research [28].

Facial Electromyography (fEMG).

fEMG data is obtained by locating gold cup electrodes in specific facial muscles (zygomaticus major, corrugator supercilia, and levator labii), and a ground electrode on the left mastoid, according to recommended guidelines [16, 29, 30].

2.3 Procedure

First, the research follows ethical guidelines for the study of human subjects [31], which implies the approval of this protocol by the Institutional Review Board (IRB) of our university. This also includes asking participants to sign an informed consent, in which we explain them about the experiment and all possible consequences of it. Moreover, all information gathered from participants is anonymized, to take care of the privacy of participants.

Collection of Experimental Stimuli.

The procedure to collect weight loss advertising images from the Internet is presented in Fig. 1. First, seed terms are identified from previous studies [32,33,34,35], and are used for the acquisition of other terms [36] in Google trends, YouTube, Instagram, Twitter, Pinterest, and Facebook, through an iterative process. For example, the terms “diet” and “weight loss” are seed terms found in previous literature, whereas “intermittent fasting” is a term that can be found in Google trends while searching for “diet” in the geographical location of our interest. The iterative process finishes by considering keyword grouping relevant to the original seed keywords, a technique used in sponsored search advertising [37, 38].

Fig. 1.
figure 1

Visualization of the methodology to obtain weight loss advertising images.

Full size image

Next, to perform the Internet searches, four virtual machines are configured, and four digital personae are created. A digital persona refers here to the digital data consolidation regarding the characteristics and attributes of a particular consumer [39]. In our case, we propose assigning the following sociodemographic characteristics to our digital personae: young women between 18 and 30 years, born and living in Lima, Peru. We propose four digital personae (one for each virtual machine), considering that word grouping in topic modeling is parsimonious between two and five (more could be considered overclustering) [40]. Moreover, four groups of keywords are also an appropriate number in online advertising. In other words, one group of keywords is assigned to each of the four digital personae.

A virtual machine is an operating system (virtually) installed within a host (native) operating system [41]. Considering the popularity of Microsoft Windows and Ubuntu Linux distribution, we propose configuring several virtual machines with Microsoft Windows (one for each digital persona), installed within a host Ubuntu operating system. Each virtual machine is configured and used in a way that has a unique Internet protocol (IP) address, and a unique media access control (MAC) address. Moreover, each virtual machine has a web browser configured by default. Google Chrome and Mozilla Firefox are the chosen web browsers given their popularity in desktop computers running Microsoft Windows.

Then, within each virtual machine (one at a time), we launch the web browser to create email and social media accounts (Google Gmail, Facebook, Instagram, Pinterest, Twitter), using the sociodemographic information that better describes our participants (young women from Lima, Peru). Next, being logged in in all these websites, we start searching the Internet using the groups of search terms we previously found. Similarly, we start to “like” the web pages devoted to weight loss and dieting that we previously identified as the most popular in social media websites. Finally, we start to collect diet and weight loss advertising images that we encounter during our web surfing in the previous step. Considering that consumers tend to prefer skyscrapers [42], only this type of advertising is selected. This process finishes after clicking on all of the first five non-sponsored links, because 90% of all clicks usually occur there [43], or until we have an appropriate number of ads to fit into a 1920 × 1080 pixels screen.

To evaluate the saliency of these pictorial stimuli, we will do a quasi-experimental pilot study with 20 university students (young women between 18 to 30 years). The participants will be in front of a projector screen displaying all collected ads which shuffle randomly every 10 s to avoid the confounding effect of their position (see preview available online: https://cybermind.ai/eye-tracking-stimuli-one). Saliency will be evaluated evaluating visual fixation time (with eye tracking) and a brief interview asking the participant “which advertising image was the most pleasant for you?”. At the end, only one weight loss advertising image is going to be selected for the final experimental procedure.

Experimental Procedure. First, an online survey containing the sociodemographic questionnaire and the instruments assessing body dissatisfaction, is created in a dedicated website. The online survey is designed to automatically classify participants in “satisfied” or “no satisfied” based on the responses, and randomly assigns participants to either the experimental or control condition. The experimental and control conditions consist on a website page displaying an email message and next to it: an empty ad (control condition), or the selected weight loss advertising image (experimental condition) (see preview available online: https://cybermind.ai/eye-tracking-stimuli-two). Reading and replying an email was chosen because it resembles a natural context that most Internet users are familiar with [44].

Next, each participant is invited to the experimental room and asked to sit on a comfortable chair in front of a desktop computer. Once the informed consent and the research debriefing is provided, the calibration of devices (eye tracker glasses and electrodes for fEMG) is performed. Finished this step, the participant is asked to start the experiment by following the instructions provided on the computer screen. First, participants must provide the responses to the online survey containing the sociodemographic questionnaire and the instruments assessing body dissatisfaction. For the experimental/control condition, participants are given the following instruction: “Please, read the email message and reply to the sender writing at least three lines of text”.

Previous studies have identified weight-related attentional bias in people with body dissatisfaction [45]. Therefore, it is hypothesized that participants classified as dissatisfied with their bodies will display a longer total fixation time on the weight loss ad compared with their matched counterparts in the control group.

On the other hand, previous studies in young women have found that images of overweight bodies elicit facial muscle activation indicating disgust (corrugator supercilii and levator labii), whereas images of thin bodies do not [46]. Therefore, we hypothesize that participants exposed to the weight loss advertising image will display facial muscle activation indicating pleasantness.

Once collected the data, it will be analyzed with R Statistical Software 3.6.3 and Python 3.8.2.

3 Discussion

The current study was aimed to present an experimental protocol for the study of body-related attentional bias within the algorithmic regulation paradigm of online advertising. The weight loss industry devotes a large amount of money in online advertising. However, deceptive marketing strategies and weight loss products can have hazardous effects on vulnerable people such as young women with body dissatisfaction. The current study taps into this problem by providing a detailed procedure to gather weight loss ads, and an experimental protocol to examine the effect of weight loss advertising in young women with different levels of body dissatisfaction.

Although this study has strengths, is not exempt of limitations. The largest limitation is that this protocol has not been proved yet, and therefore we expect in the near future to provide further improvements along with the final results of the experiment. Meanwhile, future studies can improve this methodology by adding other neurophysiological measures and self-reports. For example, a multimodal experimental study can use galvanic skin response (electrodermal activity), facial recognition software, and self-reports of attention and bias.