First Monday

Self-presentation in scholarly profiles: Characteristics of images and perceptions of professionalism and attractiveness on academic social networking sites by Andrew Tsou, Timothy D. Bowman, Thomas Sugimoto, Vincent Lariviere, and Cassidy R. Sugimoto



Abstract
Online self-presentation is of increasing importance in modern life, from establishing and maintaining personal relationships to forging professional identities. Academic scholars are no exception, and a host of social networking platforms designed specifically for scholars abound. This study used Amazon’s Mechanical Turk service to code 10,500 profile pictures used by scholars on three platforms — Mendeley, Microsoft Academic Search, and Google Scholar — in order to determine how academics are presenting themselves to their colleagues and to the public at large and how they are perceived — particularly in relation to professionalism and attractiveness. The majority of the individuals on Mendeley, Microsoft Academic Search, and Google Scholar were Caucasian, male, and perceived to be over the age of 35. Females and younger individuals were perceived as less professional than male and older individuals, while women were more likely to be perceived as “attractive.” In addition, the Mechanical Turk coders were susceptible to framing; the individuals in the profile pictures were considered more “professional” if they were identified as “scholars” rather than merely as “individuals.” The results have far-reaching implications for self-presentation and framing, both for scholars and for other professionals. In the academic realm, there are serious implications for hiring and the allocation of resources and rewards.

Contents

1. Introduction
2. Methods
3. Results
4. Discussion
5. Conclusion and future directions

 


 

1. Introduction

The adage “don’t judge a book by its cover” notwithstanding, sociological research has shown that people are indeed judged by their appearance. Outward manifestations matter more than ever in the digital age, and a person’s online presentation of visual self — be it via a formal photograph, a cartoon image, or a “selfie” taken on a camera phone — is now one of the ways in which they are judged, both socially and professionally. This can have “considerable real-world consequences” (Vernon, et al., 2014) in today’s “attention economy” (Davenport and Beck, 2001). For example, hiring decisions in academe may be affected by the types of self-images that applicants choose to make available to the public, just as potential collaborators may be influenced by the images of their peers that are encountered online.

Self-presentation is particularly important when it comes to conveying professionalism. How we look, dress, and frame ourselves typically influences how we are perceived. As Goffman observed, front-stage presentation of self, which can now include the use of self-portraits and “selfies”, projects the image we want to convey (Goffman, 1959). It is for the audience to decide whether the representation is accurate or not.

Whereas scholars once decorated their office doors with a miscellany of objects, decoupage, and memorabilia (Lang, 2014), many are now judged based on their online persona(s), which appear in a diverse array of venues: from official faculty homepages to social networking sites. In this paper we undertake an exploratory study to analyze how academics present themselves in profile pictures online and how they are perceived. In particular, we examine perceptions and predictors of professionalism, and how these are influenced by age, gender, and race. We examine differences amongst three different academic social networking platforms: Microsoft Academic Search, Mendeley, and Google Scholar. Furthermore, we analyze how framing individuals as “scholars” alters perceptions of professionalism. Specifically, we address the following research questions:

  1. How do scholars present themselves online? Do presentation characteristics vary by demographic characteristic (i.e., gender, age, or race)? Do presentation characteristics vary by platform (i.e., Mendeley, Google Scholar, Microsoft Academic Search)?

  2. What is the relationship between presentation characteristics and perceptions of professionalism? Does this vary by demographic characteristic (i.e., gender, age, race)?

  3. What is the relationship between presentation characteristics and perceptions of attractiveness? Does this vary by demographic characteristic (i.e., gender, age, race)?

  4. How does priming affect the framing of perceptions of professionalism? How does priming affect the framing of perceptions of attractiveness?

Our results provide a novel analysis of online scholarly personas and reveal the extent to which assessments based on appearance may reflect and perpetuate prevailing stereotypes.

1.1. Professional appearance

It may come as no surprise that a professional’s appearance is a key factor in how they are perceived. There is a link between clothing style and perceived competence (Anderson, et al., 1994), both by workers within a given domain and by others whom they encounter in their business dealings. Patients respond more positively to their physicians if the latter exhibit traditional attire (Gjerdingen, et al., 1987) and it has been found that people are comforted by a “ritual of verification” in the field of auditing, which includes a professional appearance on the part of the auditor (Carrington, 2010). Of course, even unchangeable appearance factors (such as height) have a substantial effect on how a person is perceived (Stulp, et al., 2013) and their “social esteem” (Judge and Cable, 2004).

Stereotypes of what certain professionals should look like are deeply rooted in our culture, to the degree that schoolchildren recognize the stereotypical image of scientists dressed in lab coats, surrounded by books and laboratory tools, and more often than not wearing glasses and/or sporting facial hair (Chambers, 1983). There is also a tendency to view scientists as “male[s] with funny hair, weird smile, wild eyes, [and] robot-like features and scars” [1], a somewhat comical image that is perhaps unique in that scientists are expected to present a rather “unprofessional” appearance. For other professions, visual appearance can be an important factor when evaluating the worth of the professional (Anderson-Gough, et al., 2002; Cooper and Robson, 2006).

There are also elements of racism and sexism at play; a recent study of politicians’ official headshot photographs found that there was a greater likelihood for the portraits to emphasize the person’s face when the politicians were “men and racial majorities,” and in some cases a person’s facial features can be indicative of their voting record [2]. Given the prevalent gender disparities in higher education (Barone, 2011; Larivière, et al., 2013; Peterson, 2012; Teelken and Deem, 2013), the degree to which self-presentation contributes to gendered notions of professionalism warrants analysis.

1.2. Presentation of self

Research investigating impression management and self-presentation has proliferated in sociology since the late 1950s, thanks to Goffman’s (1959) Presentation of self in everyday life. However, definitions can be traced back to Cooley, who argued that “the imaginations which people have of one another are the solid facts of society” (Cooley, 1902). There were others who wrote about impression management before Goffman (Tedeschi and Riess, 1981), but it was Goffman who created new interest in the phenomenon, spawning multiple investigations into self-presentation and impression management across a number of disciplines (Baumeister, 1982; Leary and Kowalski, 1990; Schlenker, 1980).

Goffman provided a framework for examining social interactions in everyday life, dissecting the details of face-to-face interaction to discuss the self and identity, cooperation, context, information flow and meaning, and impression management, which he described as the process of expressing certain information in order to impress certain ideas upon an audience during social interaction. Goffman utilized “dramaturgical” concepts to interpret roles performed by individuals during face-to-face interactions and to understand the social meanings recognized by the participants through these various roles. He also used terms such as “actors,” “teams,” and “audience” to describe social interaction (Goffman, 1959). According to Goffman, during any interaction an actor performs for an audience either as an individual or as part of a team. During this performance the actor both gives (e.g., verbal communication) and gives off (e.g., body language, gestures, movement, use of props, etc.) expressions through signs and signals, and also uses language, mannerisms, and props to facilitate impressions of the self for others to interpret. A person engaged in impression management can present a self that is true to the nature of the presentation or present a self that is embellished in some way so as to accommodate the presentation goal. The belief in a performance by an audience is dependent upon the ability of the performer(s) to maintain a consistent self (e.g., remaining consistent to a portrayed role) and to maintain control of the presented information.

In regard to technology, Goffman used examples from radio and television to discuss mediated impression management, stating that workers in those fields “keenly appreciate that the momentary impression they give will have an effect on the view a massive audience takes of them ... great care is taken to give the right impression and great anxiety is felt that the impression given might not be right” (Goffman, 1959). Goffman also introduced a model to describe how people ‘frame’ events in order to make sense of them (Goffman, 1974). When people interact or come upon an event, they apply frames to understand what they are viewing or experiencing. Frames can be adapted (what Goffman calls keyed or fabricated) in a specific context so that it alters the way people interpret the activity. This model has been used to examine areas such as social movement (Harlow, 2012; Pyles and Harding, 2012), media studies (Jarlenski and Barry, 2013; Lin and Sun, 2011), and online presentation (Miller, 1995).

As others have done before (Baym and boyd, 2012; boyd, 2007; Buffardi and Campbell, 2008; Meyrowitz, 1990; Murthy, 2013; Papacharissi, 2011), this work extends Goffman’s impression management framework to new media. We will extend this work by examining online scholarly presentation using the lens of impression management. Specifically, we will study profile pictures and the degree to which scholars’ professional identities are shaped by publicly shared personal images.

1.3. Social networking

There are many motivations to engage in online social networking activity, both personal and professional. It has been argued that the basic attraction of social media sites is derived from the continuous flow of new people and information, which creates a type of “social grooming” (i.e., a repetitive experience signaling need in an otherwise austere computer-mediated environment) (Donath, 2007). Although early research into Internet self-presentation indicated that some users were unconcerned with their online self simply because their online presence was intended to connect them with people whom they already knew (Walker, 2000), contemporary Internet use involves interacting with a vast invisible and unknown audience.

Modern Internet users can hardly be said to be shy about the degree to which their online self-presentation extends, particularly younger users (Nosko, et al., 2010). Of course, “the participants of digital environments may create multiple identities through digital appropriation and manipulation of text, images, icons, and hyperlinks to other Web sites” [3], a fact that has always been of concern to all but the least savvy of Internet users. Online social networking allows not only for multiple identities, but for a democratic system in which self-presentation to a public need not be limited to media personalities or those who work in the public sphere. In a mediated network where communication is lasting, discoverable, reproducible, and presented to potentially anyone with access, the presenter and audience are affected — and perhaps defined — in novel ways (boyd, 2011).

 

++++++++++

2. Methods

2.1. Data

The data were gathered from three online social networking sites for academics: Microsoft Academic Search, Google Scholar, and Mendeley. Ultimately, 10,000 profile pictures were sampled from each of these sites, for a total initial sampling frame of 30,000 images. All scraping was done according to the regulations of the respective platforms, and with direct consent of the staff from one venue (Mendeley). This phase of our research was judged to be exempt from IRB considerations by Indiana University’s Human Subjects Office. All sampled images and profiles are publicly available, such that the study is replicable by others wishing to use similar automatic scraping methods (see below) in order to harvest data from the three sites.

2.1.1. Microsoft Academic Search

A program was written using PHP and jQuery to automatically retrieve profile images from Microsoft Academic Search (MAS) profile pages. The images from MAS were gathered from the base domain of http://academic.research.microsoft.com/ between 27 December 2013 and 7 March 2014. To obtain a sample of 10,000 images, the program visited 242,000 profiles starting from the profile with an ID of “11.” If a profile had a generic profile image, the scholar was skipped. This meant that 4.1 percent of the examined profiles contained original images. It is worth mentioning that the IDs are a product of the MAS system and may reflect the order in which the profiles were created.

In addition to the profile images, the textual content of the MAS profiles was downloaded on 7 March 2014 from http://datamarket.azure.com/dataset/mrc/microsoftacademic. This data set contains the textual profile information associated with each scholar, which may include institutional affiliation, full name, e-mail address, homepage, LinkedIn URL, Wikipedia URL, Twitter username, and research interests.

2.1.2. Google Scholar

A bookmarklet script and harvesting application were written using PHP and jQuery in order to automatically retrieve profile images from Google Scholar (GS) profile pages. The images from GS were gathered from the base domain of http://scholar.google.com/citations between 2 January and 15 February 2014. To obtain a sample of 10,000 images, three researchers visited the GS page and began harvesting by searching GS for the letters “A”, “I”, and “R,” respectively, and continued through the alphabet until saturation in the sample. It is important to note that the GS search retrieves all profiles where first, middle, or last names start with the letter (e.g., searching “R” would retrieve “Robert General” and “Amy R. Fernandez”), so duplicates were identified during searching that were subsequently removed.

Each search page contained a maximum of 10 profiles. Because the search feature of GS is limited and different searches can contain duplicates, no total count of GS profiles can be reported. The bookmarklet ignored all profiles that contained a generic profile image. Slightly less than half (49.1 percent) of the examined profiles were missing images or had a generic image; therefore, more than 20,000 profiles had to be examined in order to identify 10,000 unique profiles with associated images. In addition to the profile images, the bookmarklet downloaded any textual content associated with the GS profiles. For any given scholar, this could include institutional affiliation, full name, citation count, and an e-mail verification notice.

2.1.3. Mendeley

The data from Mendeley were gathered from http://www.mendeley.com/directory/ in early February 2014. Unlike Google Scholar and MAS, the entire Mendeley directory was harvested. The directory classified profiles into 25 different disciplines. Profile pictures were gathered proportionally from those disciplines so that the collection of images would reflect the disciplinary breakdown of Mendeley users. For example, the discipline with the most pages was “Computer and information science,” with 1,166. This comprised roughly 14 percent of the total pages, and hence 1,404 images from this group (14 percent of 10,000) were randomly selected for inclusion in the final sample. In total, 70.4 percent of all Mendeley profiles contained images (of the 207,550 profiles on the site at the time of harvest, 146,096 had an associated image), a much higher proportion than found on Microsoft Academic Search and Google Scholar, suggesting that the user community behaves differently on this platform.

2.2. Sampling

As stated earlier, 10,000 images were sampled from each platform, for a total of 30,000 images. These images were coded by workers (known as Turkers) on Amazon’s Mechanical Turk (AMT) service. The Turkers were asked to identify all images that were photographs of a single adult human: that is, photographs that contained only a single person who appeared to be above 18 years of age (as opposed to objects, groups of people, photographs of adults with children, etc.). Only Mechanical Turk Masters were used, and each HIT (Human Intelligence Task) requested that the Turkers categorize six images. Turkers were paid one cent for each of these HITs.

The HITs were titled “Image categorization.” Their description read as follows: “You will be shown a series of images. You will be asked to identify those images that contain photographic representations of a single adult person; keywords: images, categorization, tagging.” In order to comply with the Indiana University Human Subjects Office’s IRB requirements, the following text was included with each HIT (this message was also included in subsequent phases of the study): “Thank you for agreeing to participate in our research. Before you begin, please note that the data you provide may be collected and used by Amazon as per its privacy agreement. Additionally, this research is for residents of the United States over the age of 18; if you are not a resident of the United States and/or under the age of 18, please do not complete this survey.”

Results were validated by a researcher who manually checked every image that had not been flagged as a photograph of a single adult. In total, 267 images were identified as incorrectly flagged; these images were added back into the pool of photographs that were eligible for analysis in a later round of AMT coding. It was later established that were in fact more false negatives than this (n = 76). However, the sampling frame described earlier made this point all but irrelevant.

In addition, the researcher manually checked roughly 2,000 images that had been coded as a “photograph of a single adult,” and none of these had been incorrectly coded. At the end of this process, 27,109 (90.4 percent) of the images were categorized as “photographs of a single adult.” That is, less than 10 percent of users provided an image that either contained multiple individuals or was not an image of a person. To make the analysis manageable, a random sample of 3,500 images from each platform was sampled, for a total sample size of 10,500.

2.3. Coding

2.3.1. Codebook development

The images were coded for descriptive (e.g., age, race, gender), objective (e.g., presence/absence of glasses, color of clothes), and subjective (i.e., attractiveness, professionalism) variables. The codebook was created inductively by having five different coders examine a random set of photos and identify themes amongst the images. Three researchers subsequently developed the codebook, which was informed by previous analyses in the literature and the themes identified in the first iteration of analysis. This codebook was released to three coders who tested it on 25 images. Inter-rater reliability was assessed and a discursive process was employed where each coder reflected on their experience with the codebook. The coding scheme was developed iteratively, with a pair of researchers testing the codebook at each iteration. Decisions as to which options to include, edit, or remove were made according to a variety of factors, including previous literature, the practicality of each question, and the relevance of questions to the research questions. The final codebook is provided in Appendix 1, along with justifications and motivations from the literature for each of the variables.

2.3.2. Coding

Each of the 10,500 images was posted as its own HIT on AMT, along with the codebook. After some initial price testing, 10 cents per HIT was offered. In addition, Turkers were required to have at least a 98 percent approval rating and have completed at least 5,000 HITs. In contrast with the initial phase of AMT coding described earlier, Turkers were not required to be Masters.

In order to ensure that the Turkers were reading the questions and not just randomly selecting variables, a validation question was added (“What is 1+1?”). In addition, the Turkers’ input was manually reviewed if the Turker in question had completed fewer than five HITs for the project. Afterwards, work done by Turkers who had previously provided reliable information was automatically approved.

The title for this HIT was “Hogwild for classifying images of people?” The title was chosen to be intentionally memorable, in the hope that Turkers would remember the title for the next round of coding. The description for this round of coding read as follows: “You will be shown a series of photographs, each depicting a person. You must respond to questions about the person’s appearance, objects in the photograph, etc. Keywords: images, tagging, evaluation.”

2.3.3. Priming analysis

In order to both validate the initial coding and investigate priming, another round of coding using the same 10,500 images was conducted using AMT Turkers. The title of this HIT was “Categorize photographs of scholars,” and the description read as follows: “You will be shown a series of photographs, each depicting a scholar. You must respond to questions about the academic’s appearance, objects in the photograph, etc. Keywords: images, scholars, categorization, evaluation.” Essentially, whereas the instructions given to the AMT Turkers in the previous round of coding referred to the people in the images as “individuals,” the second round of coding referred to the people in the images as “scholars,” both in the description and in the codebook (see Appendix 1). Otherwise, all of the variables remained the same — only the wording provided to the Turkers was changed. This allowed us to determine whether and to what extent framing (i.e., framing people in a particular way, viz., “scholars”) influenced how individuals were classified/rated. In Goffman’s terms, the profile photos became keyed (Goffman, 1974) in such a way that viewers reinterpreted them.

Critically, this phase of the study required that any Turker who had completed even a single HIT in the previous round of coding be considered ineligible to complete any of the HITs in the present round, as to do otherwise would undermine the entire point of using this phase as a validation/primer investigation. The HITs for this batch of images began with the following disclaimer: “IMPORTANT: Please do NOT complete this HIT if you completed one of our HITs entitled “Hogwild for classifying images of people?” These hits were available between April 6, 2014 and April 8, 2014. YOUR WORK WILL BE REJECTED IF THIS APPLIES TO YOU! If you are unsure whether or not you qualify to participate in this HIT, please contact us.” This disclaimer was added after having to reject over 1,000 HITs from ineligible Turkers and receiving a number of complaints. Turkers were offered 10 cents as was done previously.

In total, data for both rounds of coding of single adult images were gathered from 404 and 221 unique Turkers, respectively, while the initial coding of the 30,000 images was gathered from 28 unique Turkers (note that there may be overlap between these groups, as Turkers from round 1 were NOT prohibited from participating in rounds 2 or 3, so long as they did not participate in both).

2.4. Coding validation

Two types of inter-rater reliability were conducted. For one round, a random sample of 100 images coded by AMT Turkers was compared against coding done by one of the researchers. The second round of coding used the priming coding to assess reliability amongst Turkers.

2.4.1. Turker-researcher reliability

One of the authors of this study selected 100 random images and coded them using the same scheme used by Turkers. These new data were then compared with the data collected by the pair of Turkers who coded each image. Agreement rates were generally high for the “objective” questions, while the more subjective questions (attractiveness, professionalism) predictably generated more disagreement (Appendix 2 presents the percentage of exact agreement between the author and the Turkers). This indicates not only that AMT is useful as a reliable tool for collecting data, but also that our codebook was successfully interpreted by Turkers. We also used this to inform our analysis categories — in particular, we found that there was poor inter-rater reliability for highly nuanced categories around age and race, but that coders generally agreed on larger categories: i.e., white vs. non-white and under 35 vs. above 35. Therefore, we binned these data for further analysis.

2.4.2. Turker-turker reliability

In addition, we conducted a comparison between the Turkers who coded the “individual” data and those who coded the “scholar” data (i.e., the same images, but with slight changes in question formulation) using all of the coded data. Appendix 3 shows the percentage of images that were coded as exact matches between the two Turkers. The majority of images were coded as exact matches for most “objective” variables (e.g., glasses, animal, etc.). Again, this indicates that AMT is useful as a tool for collecting data and that the codebook was successfully interpreted by Turkers.

Because the present study focuses on perceptions, particularly those of attractiveness and professionalism, there is no need for exact matches to occur. What matters is for Turkers to honestly report their own perceptions of the images in question. If perceptions of age affect perceptions of professionalism, then the objective age matters less than the perceived age. To further illustrate this point, it is less important to know that a person depicted in a photograph is wearing glasses than it is to know that the glasses are visible to the viewer of the photograph. For example, the person might be so far removed from the camera that their glasses are all but invisible to the casual viewer. For the purposes of self-presentation, then, in such a case it could be said that the person is not presenting themselves as a glasses wearer, either consciously or subconsciously.

2.5. Gender validation

One potential bias in the data is that women may be less likely to put a photo online than men. To test whether or not the data were systematically biased in this regard, we took a random sample of 350 profiles without photos from each of the three data sources — Microsoft Academic Search, Google Scholar, and Mendeley — and automatically assigned a gender to these profiles using the name-gender matching process developed by Larivière, et al. (2013).

While most of the name data contained camel cased strings, which facilitated automated analysis, some names did not. (A camel cased string is one in which words are delimited by capital letters rather than by spaces. For example, “Albert Einstein” could be written as “AlbertEinstein” or, more commonly, “albertEinstein.”) Accordingly, not all 1,050 names were classified. Table 1 presents the percentages of classed names that were female, male, or unisex, while the final row (“TOTAL Classed”) displays the percentage of names that were able to be classed from each platform.

 

Table 1: Gender classifications for accounts that did not have original profile images.
GenderMicrosoft Academic SearchGoogle ScholarMendeley
Female14.8 percent
(n=39)
11.9 percent
(n=38)
25.8 percent
(n=68)
Male(n=218)(n=270)(n=184)
Unisex2.7 percent
(n=7)
3.8 percent
(n=12)
4.6 percent
(n=12)
Total classed(n=264)(n=320)(n=264)

 

Of those images that could be classed either as male or female, women comprised 15 percent of MAS, 12 percent of Google Scholar, and 26 percent of Mendeley. This is closely related to the proportions coded by AMT Turkers, who coded 12 percent of MAS, 15 percent of Google Scholar, and 32 percent of Mendeley as female (this counts the proportion of those coded either as male or female and discounts missing data, as well as those data coded as “unable to tell”). A chi-square test comparing the proportion of females in the dataset to the proportion without pictures shows there is no significant difference in the gender distribution of the datasets between those with photos and those missing photos — that is, one gender is not particularly more likely to avoid posting an actual photo (chi-square = 0.009391, df=2; p>0.05). This also reinforces the validity of the AMT coding for gender.

2.6. Analysis

We used SPSS (version 21) for the analysis. Differences in presentation by gender, race, and age were detected using chi-square tests for categorical variables and Kolmogorov-Smirnov tests for ordinal variables. These variables served as independent variables. Other presentation variables served as dependent variables (a full list may be found in Appendix 4). Given the large number of variables, we used a critical value of 0.001 for reporting statistical significance. Differences that arose due to the scholar/individual variable were detected using paired sample t-tests, also with a critical value of 0.001. Differences in presentation by platform were tested using chi-square tests for categorical variables and Kruskal Wallis tests for ordinal variables. Platform served as the independent variable. The same presentation variables noted above served as dependent variables. Post hoc Tukey tests were used to detect differences between each platform, with a critical value of 0.001.

Partial effects of race, gender, and age on professionalism and attractiveness were tested using ordinary least squares (OLS) regression (results in Tables 3 and 4). A baseline model with all other variables was specified. Each variable of interest (e.g., race, gender, and age) was individually added to the baseline model (for three additional models). A final model included the baseline model, as well as race, gender, and age (simultaneously). Significant results were noted by statistical significance of the variables of interest as well as statistically significant changes in the explained variation (change in R2).

We performed additional tests to verify the robustness of the main results (shown in Tables 3 and 4). Additional model specifications were run in which variable types were introduced in the order in which they were considered to be theoretically important (more important variables introduced earlier). We began with variables of interest (age, gender, race) and added control variables (e.g., attire) surrounding presentation later. This process checks for the robustness of the results for age, gender, and race and may be found in Appendices 7 and 8.

 

++++++++++

3. Results

The following infographic (Figure 1) presents the dominant characteristics of the entire dataset (i.e., “The Typical Scholar”) and then displays those characteristics that statistically differentiate each of the binary groups (i.e., men from women, under 35 from over 35, and white from non-white). For example, men were statistically (p<.001) more likely to wear glasses than women; 39 percent of men wore glasses, compared to 21 percent of women. We present, for each demographic type, those categories which were significantly different for each group type. The percentages represent the percentage of those in that group exhibiting this behavior.

 

Main characteristics of profile pictures found on Google Scholar, Mendeley, and Microsoft Academic Search
 
Figure 1: Main characteristics of profile pictures found on Google Scholar, Mendeley, and Microsoft Academic Search.

 

3.1. Demographic differences

The sample was coded as almost 80 percent male, reflecting the gendered nature of the scientific workforce (Larivière, et al., 2013; United Nations Educational, Scientific and Cultural Organization, 2007). The typical person was white, male, and over the age of 35. It should be noted that for the remainder of this article, all discussions of characteristics refer to perceptions. That is, when we say that “the typical person was white, male, and over the age of 35,” this refers to the manner in which the scholars were perceived by the coders, and not necessarily to “objective” fact.

The photos were usually limited to the scholar’s head and shoulders, and taken indoors. Black was the most frequently worn color, and casual clothing the most common style of dress. We examined each of the primary variables of interest (gender, race, and age) for statistically significant differences. Hereinafter, if we make a claim as to the perceived characteristics of a scholar, it is implied that the difference is statistically significant at the p<.001 level. These differences can be seen in Figure 1. Full data are provided in Appendix 4.

3.1.1. Gender

Men were perceived as older and heavier than women. They were more likely to wear glasses and have serious facial expressions. Their photos were more likely to include only head and shoulders and were taken in a work environment. Women were more likely than men to reveal skin on their arms, legs, and chest and to employ full-body photographs. It should be noted that many of these frequencies are very small — for example, only two percent of women displayed their legs in photographs, and fewer than one percent of men did (yet this difference was statistically significant). Men were more likely to wear black, while women were more likely to wear red.

3.1.2. Age

A number of age-related differences were found. Scholars under 35 weighed less and were more likely to be female and non-white, whereas those over 35 were more likely to be male and white. Distinctive items in the photos for the under-35 age group were animals and cars; for those over the age of 35, books and computers were more prevalent. Scholars who were under 35 were more likely to show skin on their arms, chest, and legs. They were also more likely to be portrayed from the waist up, whereas those over 35 were more likely to feature just their head and shoulders. Scholars under 35 were more likely to display a comical face and be obviously taking a “selfie,” whereas scholars in the older group were more likely to be wearing business casual clothing and be indoors. The younger group was more likely to wear green when compared to the older group, and the older group was more likely to wear white.

3.1.3. Race

There were fewer statistically significant differences between white and non-white scholars than amongst the previous two variables. White scholars were likely to be older, wearing business casual clothing, and be located in a work environment. Non-white scholars were more likely to be depicted from the waist up. Non-white scholars were also more likely to be wearing white clothing, while white scholars favored blue clothing.

3.2. Platform differences

Significant differences across platforms were tested using chi-square tests. This showed significant differences (at p<.001) for nearly all variables. Therefore, t-tests were conducted between all pairs to further investigate differences. This section reports only those categories for which one platform significantly differed (at p<.001) from both of the other two platforms (Figure 2). For example, it is significantly more likely to observe a female in Mendeley profile pictures compared to both MAS and GS. We did not report characteristics where the platform differed significantly from one other platform, but not from both other platforms. Presented in Figure 2 is the percentage of users on that platform exhibiting specific characteristics.

 

Differences in profile pictures across platforms
 
Figure 2: Differences in profile pictures across platforms.

 

3.2.1. Mendeley

Individuals on Mendeley were seen as significantly less professional than members of the other two venues. They were younger and less likely to wear glasses. They wore casual clothing, and were significantly less likely to wear formal wear, business wear, or business casual attire. They showed significantly more body parts and skin; had photos including legs, arms, and stomach; and were seen from the waist-up. They were less likely to be pictured indoors, and more likely to feature a car in the photo. They were the most likely to be female and the least likely to be Caucasian.

3.2.2. Microsoft Academic Search (MAS)

Individuals on MAS were seen as significantly less attractive. They were older and more likely to wear glasses. They were less likely to wear orange, red, or blue. They were most likely to present a headshot and significantly less likely to show waist or full body; in fact, they showed fewer body parts and less skin (chest and arms, in particular) than the other two groups.

3.2.3. Google Scholar (GS)

Individuals on GS were heavier set and less likely to be associated with a comical visage. They wore black and blue more than the individuals featured on the other platforms. They were the most likely to take a photo from the shoulders-up and wear business and business casual. They were also the most likely to be Caucasian.

3.3. Professionalism and attractiveness

Statistically significant differences between subpopulations (e.g., by gender, race, and age) were found for the subjective variables (i.e., professionalism and attractiveness) using Kolmogorov-Smirnov tests. Figure 3 shows the average rating for each subgroup, relative to the overall mean. Units are expressed in standard deviation units: for example, men were rated as .198 standard deviation units higher on professionalism than women. Women were rated as less professional and more attractive than men (it is worth mentioning again that previous studies of AMT have found that women comprise a majority of Turkers [Ross, et al., 2010]). Similarly, the under-35s (who were more likely to be female) were rated as less professional and more attractive than the over-35s. Non-white individuals were rated as slightly more attractive than white individuals. There were small differences in professionalism ratings between non-white and white individuals, but the differences were not statistically significant at the 0.001 level.

 

Professionalism and attractiveness - average differences
 
Figure 3: Professionalism and attractiveness — average differences.

 

3.3.1. Attractiveness

Table 3 shows the abbreviated results of the regression analysis (full results are available in Appendix 5). The baseline model controls for other factors related to physical appearance (i.e., weight) and photographic choice (i.e., the amount of skin that was shown, how much of the scholar’s body was visible, photo settings, facial expression). Models I-III show the parameter estimates when adding race, gender, and age individually to the baseline model. Model IV shows the results when adding all three jointly to the baseline model. The F statistic tests whether the independent variables (i.e., race, gender, and age) explain additional variation.

Even after controlling for other factors related to physical appearance (i.e., race, gender, age, weight) and photographic choice (i.e., the amount of skin that was shown, how much of the scholar’s body was visible, photo setting, facial expression), females were still generally considered more attractive than males, as shown by the statistically significant parameter estimate and change in explained variation between the baseline model and models II and IV. The association between gender and age on attractiveness remains robust under different specifications (models I-VII can be found in Appendix 7).

Perhaps predictably, older scholars were seen as less attractive than younger scholars, as shown by the statistically significant estimate and change in explained variation between the baseline model and models III and IV. The result is robust to various specifications (Appendix 7). Although non-whites were perceived as more attractive than non-minorities when using more parsimonious models, the difference fails to reach statistical significance under full specification (that is, when all variables are included in model IV). This suggests that while the typical non-white scholar is perceived as being more attractive, it may be due to factors related to presentation (e.g., facial expression or platform) rather than race per se.

 

Table 3: Models for attractiveness.
Notes: ***p<0.001; **p<0.01; *p<0.05.
1. Baseline model contains all covariates, except for race, gender, and age. A full list of covariates may be found in Appendix 5.
 Baseline1IIIIIIIV
n19,04919,04919,04919,04919,049
R2.269.269.274.283.287
Adjusted R2.266.267.271.280.284
White race -0.058***  -0.028
Female  0.226*** 0.201***
Age   -0.184***-0.174***
Change in F from baseline14.399137.584366.146160.107
p value <0.001<0.001<0.001<0.001
Partial R2.001.001.019.024

 

3.3.2. Professionalism

Age remains significantly related to professionalism under all specifications. Controlling for other physical characteristics and photographic choices, older scholars tend to be perceived as more professional than younger scholars (Table 4; full results in Appendix 6). Females tend to be perceived as less professional than males under most specifications, although the difference is not statistically significant using a 0.001 critical value when including age and race (Table 4) (it is, however, significant at the 0.05 level). Non-white scholars tend to be perceived as more professional than white scholars in the fully specified model (model IV, Table 4). However, results are not robust to specifications which omit dress and photographic characteristics (models III-V, Appendix 8) and the absence of statistical significance would seem to suggest that much of the difference in notions of professionalism by race may be explained by the dress and presentation variables.

 

Table 4: Models for professionalism.
Notes: ***p<0.001; **p<0.01; *p<0.05.
1. Baseline model contains all covariates, except for race, gender, and age. A full list of covariates may be found in Appendix 6.
 Baseline1IIIIIIIV
n19,04919,04919,04919,04919,049
R2.346.346.346.348.349
Adjusted R2.343.344.344.346.347
White race -0.039**  -0.055***
Female  -0.061** -0.046*
Age   0.088***0.091***
Change in F from baseline6.81810.03684.21934.777
p value 0.0090.002<0.001<0.001
Partial R2<.001.001.004.005

 

3.3.3. Predictors of professionalism

Figure 4 presents the parameter estimates from the fully specified models in standard deviation units (standardized by the professionalism mean and standard deviation). Each bar shows the relative relationship between that variable and professionalism. Nominal characteristics show the impact relative to the omitted characteristic (e.g., smiling shows the relative difference between smiling and not smiling on professionalism). For example, the professionalism rating for the average image with a smiling scholar was 0.169 standard deviations higher (i.e., more professional) than the average unsmiling image.

 

Parameter estimates in standard deviation units
 
Figure 4: Parameter estimates in standard deviation units.

 

As one might expect, wearing business clothing, business casual, or formal wear is strongly associated with professionalism, as is being seen in a work environment. Colors such as black, brown, purple, white, blue, and grey are all associated positively with professionalism. Both wearing glasses and being older are also strongly associated with professionalism. On the other hand, wearing a lab coat, graduation cap and gown, or athletic wear is likely to be negatively associated with professionalism. Obviously taking a “selfie” was also negatively associated with professionalism.

3.3.4. Contextualizing professionalism

We found that context, or expectations, could be influential in shaping coders’ ratings. Although inter-coder agreement on descriptive and objective variables was fairly high between the two rounds (Appendix 3), there were statistically significant differences (p<.001) with respect to professionalism when coders were told in advance that they would be shown images of “individuals” rather than “scholars” (specifically, “scholars” were perceived as being more professional than “individuals”). This suggests that verbal framing, keying (Goffman, 1974), or priming may in certain circumstances significantly influence perceptions. There were no significant differences in perceived attractiveness across the two rounds of coding, suggesting that the variations in perceived attractiveness are due to innate qualities of the presented images rather than researcher priming.

 

++++++++++

4. Discussion

4.1. Profiles

Our study sets a baseline for examining social media profile pictures and differences in use by various demographic variables. We found that most people employed a picture of a single individual (presumably themselves). Our exploratory analysis captured common demographics of the professoriate — men, those over the age of 35, and Caucasians were more highly represented in the profile pictures than other demographic characteristics. Photographs were most likely to be a head-and-shoulder shot and to be taken indoors. Casual clothing was typical and the most frequently worn color was black. The former is of particular interest when one considers a study by Karl, et al. that found that “respondents felt more competent and authoritative when wearing either formal business or business casual, more trustworthy and productive when wearing business casual, and least friendly and creative when wearing formal business attire” (Karl, et al., 2013). This may suggest a relationship between norms of academic dress and perceived notions of creativity. Significant differences by gender, race, and age suggest heterogeneous use of profile images.

4.2. Platform

Our data collection and analysis reveal several differences by social media platform. In constructing our initial database, we noted that the proportion of profiles with images varies dramatically, with Mendeley having a much higher proportion than Google Scholar or Microsoft Academic Search, suggesting a difference in user behavior. Our initial analysis also revealed differences by gender, with Mendeley having the largest proportion of women and Google Scholar the lowest. Other significant differences revealed that Mendeley had the youngest users, as well as the ones most likely to be associated with non-professional variables — reinforcing earlier research on the demographics of Mendeley users (Haustein, et al., 2014). Google Scholar academics were perceived as being associated with more professional variables — e.g., wearing blue or black, wearing business and business casual, and employing a traditional head-and-shoulders shot. MAS users, who were the most likely to be older and depicted wearing glasses, were seen as the least attractive. These findings have several implications for sociological and scientific studies using social media data. For example, contemporary scientometric studies often use data from Mendeley or other sources to describe the impact of research (Haustein, et al., 2014). However, users of this platform are significantly different demographically from Google Scholar or MAS, in ways that can have implications for the results of such studies.

4.3. Professionalism

The results suggest that stereotypical notions of gender (Anderson, et al., 1994; Cooper and Robson, 2006; Huber and Burton, 1995) continue to apply in academe, with women being perceived as less professional. In addition, younger professionals (or at least those who appear younger) are perceived as less professional than their older peers, even when accounting for differences in clothing. Thus, sex and age variables still carry an effect. There is also a priming effect at work, in that individuals were perceived as more “professional” when they were explicitly identified as “scholars.” Thus, independently of the degree to which one adheres to professional norms, a measure of perceived “professionalism” can be obtained simply by identifying as a scholar.

A number of factors contribute to perceptions of professionalism, including style of dress, location of photo, color of clothing, facial expression, age, and style of photo. It has been suggested that workers “have an interest in expressing themselves within the cultural codes of their time ... [but] in many contexts, the available (constructed) meanings do not leave room for the expression of what an individual perceives as her ‘self’” [4]; however, the wide variety witnessed in the sampled profile pictures suggests that academics, at least on social networking sites, do manage to express some perception of “self” despite largely appealing to professional norms.

After controlling for dozens of objective variables, perceptual differences remained based on age (p<.001) and, to a lesser extent, gender (statistically significant at p<.05). It should be noted that effect sizes, as measured by partial R2, were low (less than 0.01 when adding age, race, and gender to the other variables). Perceptions of race did not result in statistically significant perceptions of professionalism. The differences in professionalism by age and gender have implications for academe, which already suffers from gender disparities and a “leaky pipeline” where women are underrepresented in the senior ranks (Day, 2013). Human resource and workforce policies should take into consideration the fact that appearance serves as another micromechanism that perpetuates bias in the scientific workforce.

4.4. Limitations

The data are limited in that these are the perceptions of AMT Turkers, of whom we know very little. However, AMT regulations at the time prohibited us from gathering demographic data on Turkers, which would be useful for interpreting coding (particularly in terms of perceived attractiveness and other subjective variables). AMT did not always include this provision, and thus there exist past studies regarding the service’s demographics. It was found that Turkers tend to be younger than the average U.S. citizen and are predominantly female (Ipeirotis, 2010; Ross, et al., 2010). However, these demographics would be expected to change over time, particularly given the evolution of Amazon’s rules regarding Turker eligibility.

Another limitation is that the results are not necessarily generalizable to all scholars (or even to their online presences). However, they do provide general insights into scholars’ self-presentation behaviors on academic social networking sites. Similarly, Turkers may not be representative of the typical viewer on these Web sites.

Finally, we do not know when the profile pictures were added (or last updated), or even if the accounts have been active in recent memory. Accordingly, it is possible that scholars’ self-presentation on academic social networking sites has changed over time, even though our data would not reflect this.

 

++++++++++

5. Conclusion and future directions

The data presented in this paper showed multiple significant differences amongst scholars, in terms of how they were perceived. Men were more likely to be viewed as professional, whereas women were more likely to rate highly on the “attractive” variable. In addition, demographics changed considerably between platforms, with Mendeley users more likely to be young and/or female. Norms of personal dress seemed to be biased towards appearances traditionally associated with men, thus putting women at a disadvantage in terms of the “professionalism” variable when they strayed from these norms. Tellingly, the “professionalism” variable was significantly susceptible to priming (or keying) — people were perceived as being more professional if they were primed with the frame “scholar” rather than “individual.” Scholars looking to manage their self-presentation in these contexts should understand how impressions given through photographs affect the way the vast invisible audience interpret them and (potentially) their work. Both online and off-line norms and rules form the basis for frames which are applied by audience members to the presentations available online, which has an effect on the way that the scholar is perceived (Goffman, 1974). There is a blurring of the boundaries between personal and professional self-presentations taking place in social media contexts, thus it is important that scholars think about the images they use in these types of publically available profiles.

It is important to note that these norms are rarely codified; more specifically, “dress and appearance codes are often seen as trivial, both because they seem to fit within our notions of how people ought to behave” [5]. Rather, oral tradition prevails in terms of normative appearance (Bartlett, 1994). Indeed, visual appearance is a critical component of perceived professionalism (Anderson-Gough, et al., 2002; Cooper and Robson, 2006). Recent research has indicated that women actually have an advantage over men when it comes to hiring decisions in academe’s STEM fields (Williams and Ceci, 2015). Nevertheless, “a stark gender disparity persists within academic science,” due at least partially to “faculty gender bias” (Moss-Racusin, et al., 2012). This is in line with previous research on TED Talks presenters (Tsou, et al., 2014), which found that “commenters were more likely to discuss the presenter if she was female. Furthermore, there were significant differences in the sentiment of the comments when the speaker was discussed: comments tended to be more emotional when discussing a female presenter (significantly more positive and negative)” — that is, comments concerned the presenter’s appearance or speaking style with more sentiment (either positive or negative) than was evident when the presenter was a male. Gender biases, then, are applicable to scholars across a variety of platforms, from social media sites to conference-style presentations aimed at a mass audience, and scholars should be aware of how their publicly presented image affects the manner in which they and their work are perceived. As Goffman suggests, these norms have been used to reinforce cultural values and to construct a frame of professionalism that can be employed to make sense of certain situations (Goffman, 1974).

Future research might study profile pictures from a wider variety of sources — for example, images on official faculty homepages might be considered alongside academic social networking images (and perhaps even personal social networking images), if the latter could be identified and linked to the scholars in question. Another avenue of inquiry could contrast “intent” and “perception.” However, because this would involve direct contact with each of the scholars in question (which could extend to lengthy interviews in some cases), the sample size would necessarily be restricted and could result in severe response bias. Finally, it would be interesting to compare the content of scholars’ social networking profile images with those of non-scholars, as well as to plot the observed demographics of academic social networks (or at least the demographics of those who use a profile picture that is presumably a photograph of themselves) against the known demographics of academics. This could be particularly interesting with the Mendeley dataset, as the scholar’s associated discipline is known, which would allow another variable to be considered in the analysis. End of article

 

About the authors

Andrew Tsou is a graduate student in the School of Informatics and Computing at Indiana University Bloomington.

Timothy D. Bowman is a postdoctoral researcher at the Research Unit for the Sociology of Education (RUSE) at the University of Turku, Finland.

Thomas Sugimoto is Evaluation Coordinator at the Center for Evaluation and Education Policy at Indiana University Bloomington.

Vincent Larivière is Professeur agrégé in the École de bibliothéconomie et des sciences de l’information at the Université de Montréal.

Cassidy R. Sugimoto is Associate Professor of Informatics in the School of Informatics and Computing at Indiana University Bloomington.
Corresponding author, E-mail: sugimoto [at] indiana [dot] edu

 

Acknowledgements

This work was funded in part by the Alfred P. Sloan Foundation (grant #G–2014–3–25). We would like to thank Blaise Cronin for discussions and edits on earlier drafts of this manuscript.

 

Notes

1. Huber and Burton, 1995, p. 373.

2. Konrath and Schwarz, 2007, p. 445.

3. Schau and Gilly, 2003, p. 386.

4. Bartlett, 1994, p. 2,550.

5. Brower, 2013, p. 491.

 

References

John C. Anderson, Eric N. Johnson, and Philip M.J. Reckers, 1994. “Perceived effects of gender, family structure, and physical appearance on career progression in public accounting: A research note,” Accounting, Organizations and Society, volume 19, number 6, pp. 483–491.
doi: http://dx.doi.org/10.1016/0361-3682(94)90019-1, accessed 14 March 2016.

Fiona Anderson-Gough, Christopher Grey, and Keith Robson, 2002. “Accounting professionals and the accounting profession: Linking conduct and context,” Accounting and Business Research, volume 32, number 1, pp. 41–56.
doi: http://dx.doi.org/10.1080/00014788.2002.9728953, accessed 14 March 2016.

Carlo Barone, 2011. “Some things never change: Gender segregation in higher education across eight nations and three decades,” Sociology of Education, volume 84, number 2, pp. 157–176.
doi: http://dx.doi.org/10.1177/0038040711402099, accessed 14 March 2016.

Katharine T. Bartlett, 1994. “Only girls wear barrettes: Dress and appearance standards, community norms, and workplace equality,” Michigan Law Review, volume 92, number 8, pp. 2,541–2,582.
doi: http://dx.doi.org/10.2307/1290002, accessed 14 March 2016.

Roy F. Baumeister, 1982. “A self-presentational view of social phenomena,” Psychological Bulletin, volume 91, number 1, pp. 3–26.
doi: http://dx.doi.org/10.1037/0033-2909.91.1.3, accessed 14 March 2016.

Nancy K. Baym and danah boyd, 2012. “Socially mediated publicness: An introduction,” Journal of Broadcasting & Electronic Media, volume 56, number 3, pp. 320–329.
doi: http://dx.doi.org/10.1080/08838151.2012.705200, accessed 14 March 2016.

danah boyd, 2011. “Social network sites as networked publics: Affordances, dynamics, and implications,” In: Zizi Papacharissi (editor). A networked self: Identity, community, and culture on social network sites. New York: Routledge, pp. 39–58.

danah boyd, 2007. “None of this is real: Identity and participation in Friendster,” In: Joe Karaganis (editor). Structures of participation in digital culture. New York: Social Science Research Council, pp. 132–157.

Todd Brower, 2013. “What’s in the closet: Dress and appearance codes and lessons from sexual orientation,” Equality, Diversity and Inclusion, volume 32, number 5, pp. 491–502.
doi: http://dx.doi.org/10.1108/EDI-02-2013-0006, accessed 14 March 2016.

Laura E. Buffardi and W. Keith Campbell, 2008. “Narcissism and social networking Web sites,” Personality and Social Psychology Bulletin, volume 34, number 10, pp. 1,303–1,314.
doi: http://dx.doi.org/10.1177/0146167208320061, accessed 14 March 2016.

Thomas Carrington, 2010. “An analysis of the demands on a sufficient audit: Professional appearance is what counts!” Critical Perspectives on Accounting, volume 21, number 8, pp. 669–682.
doi: http://dx.doi.org/10.1016/j.cpa.2010.03.004, accessed 14 March 2016.

David Wade Chambers, 1983. “Stereotypic images of the scientist: The draw–a–scientist test,” Science Education, volume 67, number 2, pp. 255–265.
doi: http://dx.doi.org/10.1002/sce.3730670213, accessed 14 March 2016.

Charles Horton Cooley, 1902. Human nature and the social order. New York: C. Scribner’s Sons.

David J. Cooper and Keith Robson, 2006. “Accounting, professions and regulation: Locating the sites of professionalization,” Accounting, Organizations and Society, volume 31, numbers 4–5, pp. 415–444.
doi: http://dx.doi.org/10.1016/j.aos.2006.03.003, accessed 14 March 2016.

Thomas H. Davenport and John C. Beck, 2001. The attention economy: Understanding the new currency of business. Boston Harvard Business School Press.

Elizabeth Day, 2013. “How selfies became a global phenomenon,” Guardian (13 July), at http://www.theguardian.com/technology/2013/jul/14/how-selfies-became-a-global-phenomenon, accessed 14 March 2016.

Judith Donath, 2007. “Signals in social supernets,” Journal of Computer–Mediated Communication, volume 13, number 1, pp. 231–251.
doi: http://dx.doi.org/10.1111/j.1083-6101.2007.00394.x, accessed 14 March 2016.

Dwenda K. Gjerdingen, Deborah E. Simpson, and Sandra L. Titus, 1987. “Patients’ and physicians’ attitudes regarding the physician’s professional appearance,” Archives of Internal Medicine, volume 147, number 7, pp. 1,209–21,212.

Erving Goffman, 1974. Frame analysis: An essay on the organization of experience. Cambridge, Mass.: Harvard University Press.

Erving Goffman, 1959. The presentation of self in everyday life. Garden City, N.Y.: Doubleday.

Summer Harlow, 2012. “Social media and social movements: Facebook and an online Guatemalan justice movement that moved offline,” New Media & Society, volume 14, number 2, pp. 225–243.
doi: http://dx.doi.org/10.1177/1461444811410408, accessed 14 March 2016.

Stefanie Haustein, Vincent Larivière, Mike Thelwall, Didier Amyot, and Isabella Peters, 2014. “Tweets vs. Mendeley readers: How do these two social media metrics differ?” it-Information Technology, volume 56, number 5, pp. 207–215.
doi: http://dx.doi.org/10.1515/itit-2014-1048, accessed 14 March 2016.

Richard A. Huber and Grace M. Burton, 1995. “What do students think scientists look like?” School Science and Mathematics, volume 95, number 7, pp. 371–376.
doi: http://dx.doi.org/10.1111/j.1949-8594.1995.tb15804.x, accessed 14 March 2016.

Panagiotis G. Ipeirotis, “Demographics of mechanical turk,” at http://archive.nyu.edu/fda/bitstream/2451/29585/2/CeDER-10-01.pdf, accessed 14 March 2016.

Marian Jarlenski and Colleen L. Barry, 2013. “News media coverage of trans fat: Health risks and policy responses,” Health Communication, volume 28, number 3, pp. 209–216.
doi: http://dx.doi.org/10.1080/10410236.2012.669670, accessed 14 March 2016.

Timothy A. Judge and Daniel M. Cable, 2004. “The effect of physical height on workplace success and income: Preliminary test of a theoretical model,” Journal of Applied Psychology, volume 89, number 3, pp. 428–441.
doi: http://dx.doi.org/10.1037/0021-9010.89.3.428, accessed 14 March 2016.

Katherine A. Karl, Leda McIntyre Hall, and Joy V. Peluchette, 2013. “City employee perceptions of the impact of dress and appearance: You are what you wear,” Public Personnel Management, volume 42, number 3, pp. 452–470.
doi: http://dx.doi.org/10.1177/0091026013495772, accessed 14 March 2016.

Sara H. Konrath and Norbert Schwarz, 2007. “Do male politicians have big heads? Face-ism in online self-representations of politicians,” Media Psychology, volume 10, number 3, pp. 436–448.
doi: http://dx.doi.org/10.1080/15213260701533219, accessed 14 March 2016.

James M. Lang, 2014. “Office doors of the North American professor,” Chronicle of Higher Education (7 June), at http://chronicle.com/article/Office-Doors-of-the-North-A/44661/, accessed 14 March 2016.

Vincent Larivière, Chaoqun Ni, Yves Gingras, Blaise Cronin, and Cassidy R. Sugimoto, 2013. “Bibliometrics: Global gender disparities in science,” Nature, volume 504, number 7479 (12 December), pp. 211–213.
doi: http://dx.doi.org/10.1038/504211a, accessed 14 March 2016.

Mark R. Leary and Robin M. Kowalski, 1990. “Impression management: A literature review and two-component model,” Psychological Bulletin, volume 107, number 1, pp. 34–47.
doi: http://dx.doi.org/10.1037/0033-2909.107.1.34, accessed 14 March 2016.

Holin Lin and Chuen-Tsai Sun, 2011. “The role of onlookers in arcade gaming: Frame analysis of public behaviours,” Convergence, volume 17, number 2, pp. 125–137.
doi: http://dx.doi.org/10.1177/1354856510397111, accessed 14 March 2016.

Joshua Meyrowitz, 1990. “Redefining the situation: Extending dramaturgy Into a theory of social change and media effects,” In: Stephen Harold Riggins (editor). Beyond Goffman: Studies on communication, institutions, and social interaction. Berlin: Mouton de Gruyter, pp. 65–97.

Hugh Miller, 1995. “The presentation of self in electronic life: Goffman on the Internet,” paper presented at the Embodied Knowledge and Virtual Space Conference, University of London; version at http://www.douri.sh/classes/ics234cw04/miller2.pdf, accessed 14 March 2016.

Corinne A. Moss-Racusin, John F. Dovidio, Victoria L. Brescoll, Mark J. Graham, and Jo Handelsman, 2012. “Science faculty’s subtle gender biases favor male students,” Proceedings of the National Academy of Sciences, volume 109, number 41, pp. 16,474–16,479.
doi: http://dx.doi.org/10.1073/pnas.1211286109, accessed 14 March 2016.

Dhiraj Murthy, 2013. Twitter: Social communication in the Twitter age. Cambridge: Polity.

Amanda Nosko, Eileen Wood, and Seija Molema, 2010. “All about me: Disclosure in online social networking profiles: The case of Facebook,” Computers in Human Behavior, volume 26, number 3, pp. 406–418.
doi: http://dx.doi.org/10.1016/j.chb.2009.11.012, accessed 14 March 2016.

Zizi Papacharissi, 2011. “Conclusion: A networked self,” In: Zizi Papacharissi (editor). A networked self: Identity, community, and culture on social network sites. New York: Routledge, pp. 304–318.

Helen Peterson, 2012. “Women academic managers in Swedish higher education: Real women, gender neutral or quota-bitches,” In: Sofia Strid, Liisa Husu, and Lena Gunnarsson (editors). Proceedings from GEXcel Theme 11–12: Gender Paradoxes in Changing Academic and Scientific Organisation(s): GEXcel Work in Progress Report, volume X, pp. 57–64.

Loretta Pyles and Scott Harding, 2012. “Discourses of post-Katrina reconstruction: A frame analysis,” Community Development Journal, volume 47, number 3, pp. 335–352.
doi: http://dx.doi.org/10.1093/cdj/bsr023, accessed 14 March 2016.

Joel Ross, Lilly Irani, M. Six Silberman, Andrew Zaldivar, and Bill Tomlinson, 2010. “Who are the crowdworkers? Shifting demographics in mechanical turk,” CHI EA ’10: CHI ’10 Extended Abstracts on Human Factors in Computing Systems, pp. 2,863–2,872.
doi: http://dx.doi.org/10.1145/1753846.1753873, accessed 14 March 2016.

Hope Jensen Schau and Mary C. Gilly, 2003. “We are what we post? Self-presentation in personal Web space,” Journal of Consumer Research, volume 30, number 3, pp. 385–404.
doi: http://dx.doi.org/10.1086/378616, accessed 14 March 2016.

Barry R. Schlenker, 1980. Impression management: The self-concept, social identity, and interpersonal relations. Monterey, Calif.: Brooks/Cole Publishing.

Gert Stulp, Abraham P. Buunk, Simon Verhulst, and Thomas V. Pollet, 2013. “Tall claims? Sense and nonsense about the importance of height of U.S. presidents,” Leadership Quarterly, volume 24, number 1, pp. 159–171.
doi: http://dx.doi.org/10.1016/j.leaqua.2012.09.002, accessed 14 March 2016.

James T. Tedeschi and Marc Riess, 1981. “Identities the phenomenal self, and laboratory research,” In: James T. Tedeschi (editor). Impression management theory and social psychological research. New York: Academic Press, pp. 3–22.

Christine Teelken and Rosemary Deem, 2013. “All are equal, but some are more equal than others: Managerialism and gender equality in higher education in comparative perspective,” Comparative Education, volume 49, number 4, pp. 520–535.
doi: http://dx.doi.org/10.1080/03050068.2013.807642, accessed 14 March 2016.

Andrew Tsou, Mike Thelwall, Philippe Mongeon, and Cassidy R. Sugimoto, 2014. “A community of curious souls: An analysis of commenting behavior on TED talks videos,” PloS One, volume 9, number 4, pp. e93609.
doi: http://dx.doi.org/10.1371/journal.pone.0093609, accessed 14 March 2016.

United Nations Educational, Scientific and Cultural Organization (UNESCO), 2007. Science, technology and gender: An international report. Paris: UNESCO, at http:/unesdoc.unesco.org/images/0015/001540/154045e.pdf, accessed 14 March 2016.

Richard J.W. Vernon, Clare A.M. Sutherland, Andrew W. Young, and Tom Hartley, 2014. “Modeling first impressions from highly variable facial images,” Proceedings of the National Academy of Sciences, volume 111, number 32, pp. E3353–E3361.
doi: http://dx.doi.org/10.1073/pnas.1409860111, accessed 14 March 2016.

Katherine Walker, 2000. “‘It’s difficult to hide it’: The presentation of self on Internet home pages,” Qualitative Sociology, volume 23, number 1, pp. 99–120.
doi: http://dx.doi.org/10.1023/A:1005407717409, accessed 14 March 2016.

Wendy M. Williams and Stephen J. Ceci, 2015. “National hiring experiments reveal 2:1 faculty preference for women on STEM tenure track,” Proceedings of the National Academy of Sciences, volume 112, number 17, pp. 5360–5365.
doi: http://dx.doi.org/10.1073/pnas.1418878112, accessed 14 March 2016.

 

Appendices

 

Codebook and justification
 
Appendix 1: Codebook and justification.

 

 

Appendix 2: Turker-research inter-rater agreement.
 IndividualScholar
Animals97%95%
Age96%91%
Face95%93%
Gender95%96%
Clothing93%94%
ComicalFace93%95%
VisualEffect92%87%
Provocative91%74%
Smiling89%86%
Glasses88%91%
HowClose87%87%
Serious82%74%
Selfie81%69%
VisibleObjects79%80%
CorrespondsBest77%67%
WorkEnvironment73%55%
PhotoSetting72%64%
Skin71%74%
Objects69%69%
Professional67%66%
Ethnicity63%52%
DominantFace59%44%
Attractive56%59%
Outfit52%51%
Weight52%53%
OtherObjects27%28%

 

 

Appendix 3: Turker-turker inter-rater agreement.
 Percentage exact matchNumber of choices
age57.4%3
animals95.4%2
arms89.8%2
athletic94.7%2
attractive26.0%5
Attractive (within 1)65.3%5
back99.7%2
black77.1%2
blue82.1%2
body97.9%2
book98.0%2
brown87.5%2
Business casual59.6%2
business70.0%2
camera99.2%2
cap98.6%2
car99.5%2
casual63.0%2
chest91.9%2
comical86.2%2
computer99.0%2
female95.5%2
formal73.0%2
glasses89.2%2
gown98.5%2
green94.0%2
grey85.1%2
head74.7%2
howclose35.0%5
indoors62.0%2
lab_eq99.0%2
legs98.9%2
orange98.6%2
other_clothing90.2%2
outfit50.7%4
professional26.3%5
Professional (within 1)65.6%5
provocative67.1%2
purple96.8%2
red94.6%2
selfie57.2%2
serious72.5%2
shoulder67.8%2
stomach98.5%2
swim99.7%2
Under 35 (years old)73.0%2
waist90.0%2
weight50.5%3
white78.8%2
White (race)68.8%2
Work environment51.1%2
yellow97.4%2

 

 

Statistically significant demographic differences
Statistically significant demographic differences
 
Appendix 4: Statistically significant demographic differences (Chi-square).

 

 

Appendix 5: Regression (Dependent variable: Attractive).
Note: Standard errors provided in parentheses under the estimate.
 BaselineIIIIIIIV
White_race -.058
(.015)
  -.028
(.015)
female  .226
(.019)
 .201
(.019)
age   -.184
(.01)
-.174
(.01)
(Constant)2.155
(.05)
2.194
(.051)
2.103
(.05)
2.448
(.052)
2.405
(.052)
heavy-.230
(.034)
-.232
(.034)
-.233
(.034)
-.182
(.034)
-.188 (.033)
thin-.083
(.017)
-.082
(.017)
-.102
(.017)
-.137
(.017)
-.151
(.017)
miss_weight-.037
(.025)
-.036
(.025)
-.042
(.025)
-.047
(.025)
-.051
(.025)
miss_gender-.235
(.078)
-.236
(.078)
-.187
(.078)
-.298
(.078)
-.253
(.078)
miss_ethnicity-.127
(.032)
-.162
(.033)
-.124
(.031)
-.149
(.031)
-.162
(.032)
scholar.046
(.015)
.046
(.015)
.047

(.015)
.053

(.015)
.054
(.015)
professional.370
(.007)
.369
(.007)
.369
(.007)
.372
(.007)
.371
(.007)
athletic.272
(.045)
.278
(.045)
.293
(.044)
.281
(.044)
.303
(.044)
cap.086
(.079)
.084
(.079)
.077
(.078)
.110
(.078)
.100
(.078)
gown.201
(.071)
.204
(.071)
.204
(.071)
.199
(.07)
.203
(.07)
swim-.106
(.155)
-.112
(.155)
-.101
(.154)
-.104
(.153)
-.103
(.153)
miss_clothing-.077
(.084)
-.076
(.084)
-.072
(.084)
-.050
(.083)
-.047
(.083)
business-.335
(.024)
-.336
(.024)
-.310
(.024)
-.296
(.024)
-.277
(.024)
bus_cas-.204
(.02)
-.201
(.02)
-.191
(.02)
-.177
(.02)
-.165
(.02)
formal.125
(.032)
.125
(.032)
.136
(.032)
.132
(.032)
.142
(.032)
glasses-.130
(.015)
-.131
(.015)
-.112
(.015)
-.095
(.015)
-.081
(.015)
miss_glasses-.119
(.05)
-.119
(.05)
-.109
(.05)
-.105
(.05)
-.096
(.05)
miss_outfit-.033
(.025)
-.032
(.025)
-.052
(.025)
-.028
(.024)
-.045
(.024)
blue-.090
(.028)
-.089
(.028)
-.069
(.028)
-.077
(.028)
-.058
(.028)
black-.029
(.026)
-.027
(.026)
-.022
(.026)
-.026
(.026)
-.020
(.026)
brown-.056
(.034)
-.056
(.034)
-.045
(.034)
-.054
(.033)
-.045
(.033)
green-.107
(.04)
-.108
(.04)
-.106
(.04)
-.101
(.04)
-.101
(.04)
grey.016
(.031)
.014
(.031)
.035
(.031)
.030
(.031)
.045
(.031)
orange-.025
(.069)
-.028
(.069)
-.032
(.068)
-.018
(.068)
-.026
(.068)
purple-.042
(.054)
-.042
(.054)
-.060
(.053)
-.045
(.053)
-.061
(.053)
red-.058
(.036)
-.057
(.036)
-.070
(.036)
-.048
(.036)
-.058
(.036)
white-.010
(.028)
-.013
(.028)
.003
(.028)
.004
(.028)
.014
(.028)
yellow-.005
(.053)
-.007
(.053)
.004
(.053)
-.006
(.052)
.002
(.052)
miss_color.058
(.195)
.058
(.195)
.035
(.195)
.029
(.194)
.010
(.193)
num_color.093
(.027)
.092
(.027)
.090
(.027)
.079
(.027)
.077
(.027)
back-.043
(.22)
-.030
(.22)
-.070
(.219)
-.050
(.218)
-.068
(.217)
chest.162
(.045)
.161
(.045)
.137
(.045)
.168
(.045)
.145
(.045)
legs.133
(.093)
.141
(.093)
.119
(.093)
.116
(.092)
.109
(.092)
stomach.018
(.091)
.014
(.091)
.040
(.091)
.016
(.09)
.034
(.09)
other_skin-.226
(.03)
-.225
(.03)
-.201
(.03)
-.198
(.03)
-.177
(.03)
miss_skin-.307
(.062)
-.306
(.062)
-.282
(.062)
-.272
(.062)
-.252
(.062)
miss_bodypart-.081
(.126)
-.082
(.126)
-.088
(.125)
-.098
(.124)
-.103
(.124)
shoulder.131
(.018)
.131
(.018)
.130
(.018)
.140
(.018)
.139
(.018)
waist.052
(.03)
.050
(.03)
.052
(.029)
.057
(.029)
.056
(.029)
body-.054
(.053)
-.058
(.053)
-.054
(.053)
-.048
(.053)
-.052
(.053)
Answer.HowClose-.024
(.008)
-.024
(.008)
-.025
(.008)
-.024
(.008)
-.025
(.008)
selfie.135
(.025)
.133
(.025)
.130
(.025)
.113
(.025)
.109
(.025)
miss_selfie.005
(.025)
.008
(.025)
.008
(.025)
-.012
(.025)
-.007
(.025)
Indoors.085
(.019)
.082
(.019)
.080
(.019)
.098
(.019)
.092
(.019)
miss_photosetting.112
(.023)
.109
(.023)
.109
(.023)
.122
(.023)
.117
(.023)
workenv-.354
(.026)
-.352
(.026)
-.351
(.026)
-.356
(.026)
-.352
(.026)
miss_workenvironment-.110
(.018)
-.106
(.018)
-.109
(.018)
-.110
(.018)
-.108
(.018)
viseffect-.313
(.029)
-.309
(.029)
-.309
(.029)
-.317
(.028)
-.313
(.028)
animals-.151
(.062)
-.148
(.062)
-.149
(.062)
-.147
(.061)
-.144
(.061)
miss_animal-.063
(.059)
-.063
(.058)
-.060
(.058)
-.060
(.058)
-.057
(.058)
book.125
(.043)
.126
(.043)
.125
(.043)
.161
(.043)
.160
(.042)
camera.793
(.102)
.797
(.102)
.784
(.101)
.807
(.101)
.800
(.1)
car.135
(.102)
.131
(.102)
.140
(.101)
.115
(.101)
.119
(.1)
computer-.032
(.067)
-.035
(.067)
-.044
(.067)
-.033
(.066)
-.045
(.066)
lab_eq.005
(.083)
.004
(.083)
.007
(.083)
-.013
(.083)
-.011
(.082)
smiling.034
(.02)
.040
(.02)
.009
(.02)
.036
(.019)
.016
(.02)
miss_smiling-.288
(.046)
-.284
(.046)
-.292
(.046)
-.309
(.045)
-.309
(.045)
serious-.255
(.022)
-.252
(.022)
-.253
(.022)
-.253
(.022)
-.251
(.022)
comical-.061
(.039)
-.061
(.039)
-.056
(.039)
-.074
(.039)
-.069
(.039)
miss_comic-.372
(.045)
-.371
(.045)
-.371
(.045)
-.379
(.045)
-.378
(.044)
provocative-.970
(.036)
-.967
(.036)
-.956
(.035)
-.964
(.035)
-.951
(.035)
miss_provocative.528
(.032)
.526
(.032)
.534
(.032)
.511
(.031)
.517
(.031)
MAS-.091
(.018)
-.095
(.018)
-.082
(.018)
-.072
(.018)
-.066
(.018)
Mendeley.129
(.018)
.120
(.018)
.102
(.018)
.057
(.018)
.033
(.018)

 

 

Appendix 6: Regression (Dependent variable: Professional).
Note: Standard errors provided in parentheses under the estimate.
 BaselineIIIIIIIV
White_race -.039
(.015)
  -.055
(.015)
female  -.061
(.019)
 -.046
(.019)
age   .088
(.01)
.091
(.01)
(Constant)1.155
(.051)
1.182
(.052)
1.164
(.051)
.990
(.054)
1.030
(.054)
heavy.051
(.033)
.050
(.033)
.052
(.033)
.030
(.033)
.028 (.033)
thin.008
(.017)
.009
(.017)
.013
(.017)
.034
(.017)
.040
(.017)
miss_weight-.089
(.024)
-.088
(.024)
-.087
(.024)
-.083
(.024)
-.081
(.024)
miss_gender-.262
(.077)
-.263
(.077)
-.274
(.077)
-.228
(.077)
-.237
(.077)
miss_ethnicity-.069
(.031)
-.093
(.032)
-.070
(.031)
-.058
(.031)
-.091
(.032)
scholar.154
(.015)
.154
(.015)
.154

(.015)
.150

(.015)
.149
(.015)
attractive.358
(.007)
.358
(.007)
.360
(.007)
.366
(.007)
.366
(.007)
athletic-.725
(.044)
-.720
(.044)
-.730
(.044)
-.728
(.044)
-.726
(.044)
cap-.250
(.078)
-.251
(.078)
-.248
(.078)
-.261
(.077)
-.261
(.077)
gown-.301
(.07)
-.298
(.07)
-.302
(.07)
-.300
(.07)
-.297
(.07)
swim-.079
(.153)
-.083
(.153)
-.080
(.152)
-.079
(.152)
-.085
(.152)
miss_clothing.152
(.083)
.153
(.083)
.151
(.083)
.139
(.083)
.139
(.083)
business1.029
(.023)
1.028
(.023)
1.022
(.023)
1.009
(.023)
1.002
(.023)
bus_cas.558
(.02)
.559
(.02)
.554
(.02)
.554
(.02)
.543
(.02)
formal.378
(.031)
.378
(.031)
.375
(.032)
.372
(.031)
.369
(.031)
glasses.139
(.015)
.138
(.015)
.134
(.015)
.123
(.015)
.118
(.015)
miss_glasses.057
(.05)
.057
(.05)
.055
(.05)
.051
(.049)
.049
(.049)
miss_outfit.150
(.024)
.151
(.024)
.155
(.024)
.147
(.024)
.152
(.024)
blue.175
(.028)
.176
(.028)
.170
(.028)
.169
(.028)
.166
(.028)
black.263
(.026)
.264
(.026)
.261
(.026)
.261
(.026)
.261
(.026)
brown.237
(.033)
.236
(.033)
.234
(.033)
.235
(.033)
.232
(.033)
green.057
(.04)
.056
(.04)
.056
(.04)
.054
(.04)
.053
(.04)
grey.132
(.031)
.131
(.031)
.127
(.031)
.125
(.031)
.119
(.031)
orange-.043
(.067)
-.045
(.067)
-.041
(.067)
-.046
(.067)
-.048
(.067)
purple.206
(.053)
.205
(.053)
.211
(.053)
.207
(.053)
.210
(.053)
red.108
(.035)
.109
(.035)
.111
(.035)
.103
(.035)
.106
(.035)
white.220
(.027)
.217
(.027)
.216
(.027)
.212
(.027)
.206
(.027)
yellow-.020
(.052)
-.020
(.052)
-.022
(.052)
-.019
(.052)
-.022
(.052)
miss_color-.225
(.192)
-.225
(.192)
-.219
(.192)
-.210
(.192)
-.206
(.192)
num_color-.122
(.027)
-.123
(.027)
-.121
(.027)
-.116
(.027)
-.116
(.026)
back-.025
(.217)
-.016
(.217)
-.017
(.216)
-.020
(.216)
-.003
(.216)
chest.075
(.045)
.075
(.045)
.082
(.045)
.071
(.045)
.075
(.045)
legs-.012
(.092)
-.006
(.092)
-.009
(.092)
-.005
(.092)
-.006
(.092)
stomach-.160
(.09)
-.162
(.09)
-.166
(.09)
-.159
(.089)
-.166
(.089)
other_skin.387
(.029)
.387
(.029)
.381
(.029)
.374
(.029)
.369
(.029)
miss_skin.346
(.062)
.346
(.061)
.340
(.062)
.331
(.061)
.326
(.061)
miss_bodypart.219
(.124)
.219
(.124)
.221
(.124)
.227
(.123)
.228
(.123)
shoulder.033
(.018)
.033
(.018)
.033
(.018)
.027
(.018)
.026
(.018)
waist.027
(.029)
.026
(.029)
.027
(.029)
.024
(.029)
.022
(.029)
body.063
(.052)
.059
(.052)
.063
(.052)
.060
(.052)
.056
(.052)
Answer.HowClose-.027
(.008)
-.027
(.008)
-.026
(.008)
-.027
(.008)
-.026
(.008)
selfie-.134
(.025)
-.135
(.025)
-.133
(.025)
-.124
(.025)
-.124
(.025)
miss_selfie-.095
(.025)
-.093
(.025)
-.096
(.025)
-.086
(.025)
-.084
(.025)
Indoors-.065
(.018)
-.067
(.018)
-.064
(.018)
-.072
(.018)
-.074
(.018)
miss_photosetting-.030
(.023)
-.033
(.023)
-.030
(.023)
-.036
(.023)
-.039
(.023)
workenv.471
(.026)
.472
(.026)
.471
(.026)
.473
(.026)
.474
(.026)
miss_workenvironment.112
(.018)
.114
(.018)
.112
(.018)
.113
(.018)
.116
(.018)
viseffect-.032
(.028)
-.030
(.028)
-.033
(.028)
-.027
(.028)
-.024
(.028)
animals.039
(.061)
.041
(.061)
.039
(.061)
.038
(.061)
.041
(.061)
miss_animal-.085
(.058)
-.085
(.058)
-.086
(.058)
-.086
(.057)
-.086
(.057)
book-.009
(.042)
-.008
(.042)
-.009
(.042)
-.028
(.042)
-.027
(.042)
camera-.047
(.1)
-.044
(.1)
-.046
(.1)
-.061
(.1)
-.057
(.1)
car-.243
(.1)
-.246
(.1)
-.245
(.1)
-.234
(.1)
-.238
(.1)
computer-.029
(.066)
-.032
(.066)
-.026
(.066)
-.029
(.066)
-.029
(.066)
lab_eq-.027
(.082)
-.028
(.082)
-.027
(.082)
-.018
(.082)
-.020
(.082)
smiling.192
(.019)
.196
(.019)
.198
(.019)
.189
(.019)
.200
(.019)
miss_smiling.038
(.045)
.040
(.045)
.039
(.045)
.050
(.045)
.055
(.045)
serious.239
(.022)
.240
(.022)
.239
(.022)
.239
(.022)
.241
(.022)
comical.201
(.039)
.201
(.039)
.200
(.039)
.207
(.039)
.206
(.039)
miss_comic.113
(.044)
.113
(.044)
.113
(.044)
.119
(.044)
.120
(.044)
provocative.698
(.035)
.699
(.035)
.696
(.035)
.701
(.035)
.701
(.035)
miss_provocative.134
(.031)
.133
(.031)
.131
(.031)
.137
(.031)
.133
(.031)
MAS.036
(.018)
.033
(.018)
.034
(.018)
.028
(.018)
.022
(.018)
Mendeley-.154
(.018)
-.159
(.018)
-.147
(.018)
-.120
(.018)
-.122
(.018)

 

 

Appendix 7: Models for attractiveness (Robustness).
Notes: ***p<0.001; **p<0.01; *p<0.05.
“Yes” indicates the inclusion of variables in the model.
 IIIIIIIVVVIVII
N19,76119,17619,17619,04919,04919,04919,049
Adjusted R20.0370.1430.1590.1870.2270.2830.284
(Constant)3.349***2.415***2.441***2.357***2.411***2.425***2.405***
White_race-0.109***-0.093***-0.079***-0.073***-0.041**-0.03*-0.028
Female0.249***0.324***0.289***0.276***0.255***0.214***0.201***
Age-0.124***-0.180***-0.159***-0.175***-0.185***-0.184***-0.174***
Weightyesyesyesyesyesyesyes
Scholar & professionalism yesyesyesyesyesyes
Dress/skin showing  yesyesyesyesyes
Picture perspective   yesyesyesyes
Photo setting    yesyesyes
Smiling/comical/provocative     yesyes
Platform      yes

 

 

Appendix 8: Models for professionalism (Robustness).
Notes: ***p<0.001; **p<0.01; *p<0.05.
“Yes” indicates the inclusion of variables in the model.
 IIIIIIIVVVIVII
N19,17619,17619,17619,04919,04919,04919,049
Adjusted R20.0280.1420.2910.2990.3230.3450.347
(Constant)3.103***1.825***1.238***1.35***1.22***0.945***1.03***
White_race-0.078***-0.038*-0.008-0.01-0.03-0.046**-0.055***
Female-0.188***-0.28***-0.076***-0.073***-0.065***-0.069***-0.046*
Age0.157***0.201***0.092***0.101***0.107***0.109***0.091***
Weightyesyesyesyesyesyesyes
Scholar & professionalism yesyesyesyesyesyes
Dress/skin showing  yesyesyesyesyes
Picture perspective   yesyesyesyes
Photo setting    yesyesyes
Smiling/comical/provocative     yesyes
Platform      yes

 

 


Editorial history

Received 7 January 2016; revised 22 February 2016; accepted 14 March 2016.


Creative Commons License
“Self-presentation in scholarly profiles: Characteristics of images and perceptions of professionalism and attractiveness on academic social networking sites” by Andrew Tsou, Timothy D. Bowman, Thomas Sugimoto, Vincent Larivière, and Cassidy R. Sugimoto is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

Self-presentation in scholarly profiles: Characteristics of images and perceptions of professionalism and attractiveness on academic social networking sites
by Andrew Tsou, Timothy D. Bowman, Thomas Sugimoto, Vincent Larivière, and Cassidy R. Sugimoto.
First Monday, Volume 21, Number 4 - 4 April 2016
https://firstmonday.org/ojs/index.php/fm/article/download/6381/5343
doi: http://dx.doi.org/10.5210/fm.v21i4.6381