Cover: A Beholder is a fantastical creature commonly found in games like Dungeons & Dragons. Beholders are known for their highly intelligent and paranoid nature, often inhabiting deep underground lairs and fiercely guarding their territories. They are notorious for their ability to manipulate and control magic, making them formidable adversaries. Hive of the Eye Tyrant” © Wizards of the Coast,
by Tony DiTerlizzi. Accessed at Art of MTG here.
“Beauty is in the eye of the beholder”
The phrase “Beauty is in the eye of the beholder” was written by Irish author Margaret Wolfe Hungerford. It appeared in her 1878 work titled “Molly Bawn,” where the original phrase stated: “Beauty is in the eye of the beholder, and it may be necessary from time to time to give a stupid or misinformed beholder a black eye.” The phrase became popular and is often used to express the idea that the perception of beauty is subjective and varies from person to person..
“No culture can live if it attempts to be exclusive.”
Mahatma Gandhi (1959). “India of My Dreams”, p.173, Rajpal & Sons
Maurício Brant Pinheiro & César Bremer Pinheiro
I. Introduction
The concept of beauty standards is the subject of analysis in different societies and cultures, representing physical and aesthetic characteristics considered ideal. However, it is crucial to question the origin and validity of these standards, taking into account cultural and biological aspects. In this article, we will explore the origins of beauty standards, highlighting their formation influenced by cultural and biological factors, as well as the negative impact of artificial intelligence (AI) generated images on the perception of beauty.
Through an experiment with the Midjourney tool, we have observed concerning results, revealing biases and problematic trends in the representation of beauty, reflecting prejudices and stereotypes present in our society. These results emphasize the need to address biases in AI systems, striving for a fairer and more inclusive representation. The images generated by AI also contribute to an unrealistic perception of the body, negatively affecting people’s self-esteem and mental health.
In the face of these challenges, it is essential to adopt a responsible and educational approach, promoting diversity and valuing the uniqueness of each individual. This requires collaboration between AI experts, psychologists, activists, and lawmakers, with the aim of developing ethical guidelines and appropriate regulations. By addressing and correcting biases in AI results, we can move towards a more inclusive society where diversity is valued and everyone feels represented and respected, regardless of their physical appearance.
II. Reflections on the Origins of a Beauty Standard
The beauty standard is a set of valued physical and aesthetic characteristics that are idealized in a society or culture. It encompasses body features, facial traits, and aesthetic attributes that are widely promoted as the beauty ideal. For example, in the Western world, the Western feminine beauty standard often includes features such as slim bodies, ample breasts, symmetrical facial features, fair skin, light, long, and silky hair, light eyes, and full lips.
Culture plays a significant role in defining beauty ideals, influencing aesthetic perceptions and preferences through media and other forms of artistic expression. Until recently, representations in movies, television shows, magazines, and advertising campaigns promoted certain appearance standards as desirable. For example, in cinema, it was common for actresses to embody the “ideal beauty” with a stereotyped image that favored certain characteristics over others.
While in traditional media, aggressive campaigns involving affirmative actions seek to deconstruct traditionally accepted beauty standards in the West, in electronic media, and particularly on social networks, these same beauty standards are organically reinforced. This dichotomy can be attributed to various factors, such as the influence of large advertising companies and conservative fashion brands that still promote a limited and stereotypical beauty standard. Additionally, the nature of social networks, with their recommendation algorithms and emphasis on appearance, can lead to a continuous reinforcement of these standards, amplifying the pressure on individuals to fit into these unattainable ideals.
On the other hand, beauty standards are not limited to humans but can also be observed in the animal kingdom. For example, various bird species exhibit colorful and elaborate plumage, displaying a phenomenon known as sexual dimorphism, where males exhibit distinct physical characteristics such as bright feathers and long tails to attract females. A notable example is the peacock, whose males have tails with vibrant patterns and colors, which are considered attractive by females of the species.
It is crucial to question both the validity and fairness of beauty standards, taking into consideration not only their cultural origins but also their undeniable biological roots. The theory of sexual selection proposed by Darwin suggests that these standards may have biological (genetic) underpinnings, as certain physical characteristics considered attractive are associated with health, fertility, and genetic fitness. Furthermore, these characteristics can influence mate selection and the perpetuation of a species through natural selection, including the predisposition to succeed in the competition for territory (and resources) inherent in most species. Therefore, it is crucial to recognize that the origin of beauty standards transcends cultural context, involving a complex interplay of biological, cultural, and social factors in their definition. Thus, it becomes essential to seek a broader and more critical understanding of these standards.
In the next section, we will delve into the analysis of the unrealistic and stereotypical beauty standards perpetuated by generative AI, regardless of their cultural or biological causes. We will investigate the training biases present in these representations, aiming for a more comprehensive and critical understanding of these phenomena.
III. The negative impact of images generated by AI
With the advancement of AI technology, generative image tools have emerged, such as Dall-E 2, Stable Diffusion, and Midjourney, which were trained on enormous datasets from the internet, particularly from social media platforms, to create representations of “perfect” male and female faces and bodies. These AI tools took into account engagement metrics, such as views, likes, comments, as well as internet search results, to determine social preferences.
Next, we will present the alarming results of the analysis of training biases, which reveal the influence of still prevailing exclusive beauty standards in our Western society. To conduct a simple and enlightening test, we used images generated by Midjourney, with basic and direct prompts. It is important to emphasize that anyone can perform this experiment, and the results are not attributable to the algorithm itself but rather to the vast training dataset used, which reflects the biases and stereotypes present in our society.
1) Prompt: The most beautiful child, white background
2) Prompt: The most beatiful lady, white background
3) Prompt: The most beatiful lady, black background
4) Prompt: The most beatiful lady, white background, full body.
5) Prompt: The most beautiful 60 year old lady, white background
6) Prompt: The most beautiful man, white background
The analysis of the above results, generated by AI, reveals the connection between beauty standards and certain physical and aesthetic characteristics, encompassing men, women, children, and the elderly. These standards reflect the persistence of outdated ideals, with a disproportionate emphasis on Caucasian features and an extremely limited representation of ethnic diversity. It is regrettable that only two out of the 60 generated images (approximately 3% of the total) are exceptions to this pattern.
It is important to emphasize that the midjourney is a spectacular generative AI tool. By using appropriate prompt engineering, it is possible to generate amazing images that represent a wide range of racial diversity. However, it is crucial to mention that, despite its ability to generate diverse images, the midjourney can be susceptible to bias in the training data, especially when prompts are simplified.
Bias in the training data occurs when the dataset used to train the model does not adequately represent the diversity of the population. This can lead to midjourney-generated results that reproduce and perpetuate existing stereotypes or prejudices in society. For example, if the training dataset predominantly contains images of a particular race, the model may struggle to accurately generate images representative of other races.
However, with awareness of the importance of diversity and the adoption of appropriate prompt engineering practices, it is possible to mitigate these biases. Prompt engineering involves carefully formulating instructions or suggestions for the model to guide its response in a more inclusive and diverse manner.
When creating prompts for the midjourney, researchers and developers can ensure that racial diversity is taken into account. This can be done through explicit instructions that emphasize the importance of equitable representation of all races and encourage the model to generate images that encompass this diversity. An example is our gallery created in honor of May 14th, Mother’s Day.
These results highlight the need to address the biases present in AI systems and strive for fair and inclusive representation in all aspects of society.
The concern goes beyond the exclusion and marginalization of individuals who do not fit into these restrictive beauty standards. The images generated by AI also contribute to an unrealistic perception of beauty, similar to highly edited and filtered images on social media. These unattainable standards negatively impact people’s self-esteem and mental health. The responsibility for these outcomes does not lie solely with the generative algorithm itself but rather with the training data that reflects society’s stereotypes (perhaps in part by those algorithms powering social media). It is essential to critically evaluate this data and seek broader, more inclusive diversity that values the uniqueness and authenticity of each individual.
IV. How to change this paradigm?
Changing this paradigm requires a careful and responsible approach. Proposals for affirmative actions have suggested introducing deliberate biases into AI training data, but it is important to explore the dangers of such attempts. Forcing biases into the data may result in replacing an unattainable ideal with another, perpetuating the exclusion of marginalized groups and contributing to the marginalization of new groups. Moreover, this artificial and imposed approach can lead to increased polarization in society, significantly amplifying existing prejudices.
It is important to recognize that imposing beauty standards in an arbitrary and artificial manner can have negative consequences, both in terms of exclusion and in strengthening and replacing harmful stereotypes. Therefore, when addressing the issue of beauty standards, it is crucial to adopt a responsible stance and seek an educational approach that promotes diversity and values the authenticity of each individual, avoiding the imposition of harmful ideals on society.
It is also essential to promote education on body image and media literacy, empowering people to recognize the standards imposed by the media and resist unrealistic beauty ideals. Collaboration among AI experts, psychologists, activists, and policymakers plays a fundamental role in developing ethical guidelines and appropriate regulations.
Everyone has a role to play in transforming this landscape by questioning established beauty ideals and promoting a culture that celebrates diversity. Companies and brands should also take responsibility to ensure that their algorithms are trained with more diverse data, avoiding the perpetuation of harmful beauty standards. By acknowledging and correcting biases present in AI outcomes, we can move towards a more inclusive society where beauty is valued in its diversity, and everyone feels represented and respected, regardless of their physical appearance. Awareness and collective action are essential to achieve this transformation.
V. The Political Bias War
It is crucial to recognize that political biases can be present and infiltrated in various areas, from digital communication to advertising and social media. It is essential to address these biases specifically, without neglecting their importance in favor of other biases, such as those of generative AI.
Currently, we are witnessing an intense ideological dispute, in which individuals like Elon Musk are developing software to counter what they perceive as left-leaning bias in ChatGPT. This dynamic emphasizes the need to critically examine political biases and the impacts they can have. It is important to acknowledge that AI, like any other tool, can reflect the biases and political inclinations of its creators and users.
A concerning example is the observed trend among students to present school assignments as if they were their own, and consequently, feel compelled to defend what was written, even if it does not necessarily reflect their own convictions. This phenomenon can lead students to fall into the trap of indoctrination, losing the critical ability to question and form their own opinions.
Furthermore, it is necessary to consider the potential new political biases that may arise in AI-driven search engines. As these tools become increasingly advanced, there is a risk that they may be used as powerful tools of indoctrination, influencing how information is presented and filtered to users. This issue is of utmost importance, as it can have significant implications on opinion formation and the dissemination of accurate and unbiased information.
It is essential to address and discuss these political biases while maintaining an impartial and critical stance. Transparency, open debate, and diversity of perspectives are crucial to ensure that AI technologies are developed and used ethically, promoting freedom of thought and avoiding manipulation and ideological indoctrination.
V. Conclusions
The analysis of beauty standards reveals the complexity of their formation, with cultural, social, and biological influences interacting intricately. Artificial intelligence (AI) technology raises concerns about perpetuating unrealistic and stereotypical standards, increasing the pressure on individuals to conform to unattainable ideals. To change this paradigm, it is crucial to adopt a careful and responsible approach, avoiding arbitrary and artificial imposition of new beauty standards that may exclude and marginalize groups while reinforcing existing prejudices. The solution lies in promoting an inclusive approach that values diversity and the authenticity of each individual.
Education on body image and media literacy play a crucial role in empowering people to resist the unrealistic beauty ideals imposed by the media. Collaboration among AI experts, psychologists, activists, and policymakers is essential in developing ethical guidelines and appropriate regulations. Transforming the current landscape requires awareness and collective action. Each one of us can question established beauty ideals and their underlying causes, but promoting a culture that celebrates diversity is the responsibility of all. Companies and brands have a responsibility to train their algorithms with more diverse data, avoiding the perpetuation of harmful beauty standards.
By acknowledging and correcting biases present in AI outcomes, we can move towards a more inclusive society where beauty is valued in its diversity, and everyone feels represented and respected, regardless of their physical appearance. Awareness and collective action are essential to achieve this transformation and foster an environment that celebrates beauty in all its forms, promoting the well-being of all.
The Bulimia Project, an awareness group about eating disorders, requested artificial intelligence (AI) to generate “perfect” male and female bodies based on what generates higher engagement on social media. Read more about this study here.
All images generated by Midjourney for this study were produced by César Bremer Pinheiro and are registered as NFTs on the blockchain at Opensea.io for copyright protection purposes.
#AI #ArtificialIntelligence #Bias #AIWithoutBias #InclusiveSociety #BeautyStandard #EmpowermentThroughAI #ExclusionByAI #BiasedData #Diversity #PerfectBody #PerfectFace #MarginalizationByAI #EthnicDiversity #BreakingBeautyStandards #InclusivityWithAI #DiversityWithAI #AIforAll #EthicalAI #FairnessInAI #EquitableRepresentation #BodyPositivity #RedefiningPerfection #InclusionMatters #EqualOpportunity #DataJustice #EmbracingDifferences #EmpowermentThroughAI #AIResponsibility #RepresentationMatters #CulturalInclusion #PromotingEquality #FightingDiscrimination #CreatingBalance #AIforSocialGood
Copyright 2024 AI-Talks.org