Filters of Identity : AR Beauty and the Algorithmic Politics of the Digital Body

Miriam Doh, Corinna Canali, Nuria Oliver · 2025

Read Filters of Identity : AR Beauty and the Algorithmic Politics of the Digital Body by Miriam Doh, Corinna Canali, Nuria Oliver (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-19611.txt.

Introduction

Miriam Doh, Corinna Canali, and Nuria Oliver write a 2025 position paper on beauty filters. Phone filters that reshape faces are often sold as fun, neutral tools. The authors argue the filters act as a form of governance over how bodies should look.

Naming choices, biased training, and platform rules push narrow standards tied to race, gender, and ability, the paper says. The push happens quietly, since the filters hide their own influence on taste. Drawing on debates in human-computer interaction about body politics, the authors call for open disclosure of how filters work. They also urge a fresh look at the values built into digital body tools and their design.

Catalog record

CREATORMiriam Doh, Corinna Canali, Nuria Oliver
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE16 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-19611

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Filters of Identity : AR Beauty and the Algorithmic Politics of the Digital Body Miriam Doh, Corinna Canali, Nuria Oliver Abstract This position paper situates AR beauty filters within the broader debate on Body Politics in HCI. We argue that these filters are not neutral tools but technologies of governance that reinforce racialized, gendered, and ableist beauty standards. Through naming conventions, algorithmic bias, and platform governance, they impose aesthetic norms while concealing their influence. To address these challenges, we advocate for transparency-driven interventions and a critical rethinking of algorithmic aesthetics and digital embodiment. Keywords Algorithmic Beauty \cdot Body Politics \cdot AR Filters \cdot Postfeminism \cdot Self-Surveillance Note This work was presented at the “Body Politics: Unpacking Tensions and Future Perspectives For Body-Centric Design Research in HCI" workshop at the ACM (Association for Computing Machinery) CHI conference on Human Factors in Computing Systems 2025. 1 Introduction AR beauty filters are marketed as tools for self-expression, yet they subtly enforce racialized, gendered, and ableist beauty standards through opaque algorithmic modifications. Rather than simply expanding self-representation, these technologies, along with content moderation algorithms on social platforms, contribute to the construction of aesthetic hierarchies. While beauty filters impose specific visual norms, content moderation systems reinforce these standards by determining which content is amplified or suppressed, together shaping a stratified landscape of digital desirability. Situating beauty filters within Body Politics in HCI, this position paper argues that these tools do more than mediate appearance—they function as technologies of governance, shaping visibility, desirability, and exclusion in digital spaces. Building on our previous research, we explore how filter naming conventions, algorithmic biases, and platform moderation policies produce and sustain hegemonic beauty norms, often contradicting platform diversity narratives. Addressing the themes of this workshop, we engage with the three key questions proposed: (Q1) What theories have influenced body-centric research in HCI, and how have bodies been conceptualized? (Q2) What new openings or ethical pitfalls emerge when attending to marginalized or more-than-human bodies through different research methods? (Q3) How can we account for the multiplicity of bodies while preserving the significance of individual experiences in body-centric research? By framing beauty filters within the politics of algorithmic aesthetics, this paper aims to spark a critical discussion on transparency, customization, and platform accountability in digital beauty governance. 1.1 (Q1) Beauty Filters and the Politics of Digital Embodiment Body-centric HCI studies conceptualize the gendered body as a site of negotiation between digital productivity and sociocultural norms, aligning with feminist critiques of capitalism, technology, and the body ( e.g. , Federici 2004 ; Haraway 2013 ). Within this framework, AR beauty filters emerge as more than mere appearance enhancements Isakowitsch 2022 ; Cug et al. 2022 ; Fribourg et al. 2021 . They work as powerful technologies of gender LAURETIS 1987 , encoding and enforcing prescriptive beauty ideals Leeat et al. 2019 through algorithmic design, benefiting platforms while shifting responsibility onto users. Objectification Theory Fredrickson and Roberts 1997 provides a key lens for understanding these dynamics, showing how individuals—particularly women—internalize an external gaze, leading to self-surveillance and body dissatisfaction. AR beauty filters amplify this process by embedding dominant beauty standards into their algorithms, subtly dictating how users modify their digital self-representation under the guise of choice and empowerment. This disciplining effect extends to filter naming conventions, which, rather than describing their actual effects, actively shape expectations of digital beauty. Our analysis Doh et al. 2024 of TikTok filters shows that names such as ‘‘Princess Makeup," ‘‘Pure Eyes," and ‘‘Prettiest" reinforce narrow, feminized beauty ideals, linking attractiveness to idealized youth, innocence, and purity. The linguistic framing of filters constructs a hierarchy of desirability, associating smooth skin, symmetry, and delicate features with idealized beauty while obscuring the extent of algorithmic facial modifications. 1 1 1 For instance, the “Prettiest” filter, with 62.2M posts, was used by 98.66% female users, reinforcing how language encodes aesthetic norms and dictates digital desirability in gendered ways. Yet the influence of AR beauty filters extends beyond individual aesthetics to platform-wide governance, where content moderation algorithms and policies dictate whose bodies are visible and desirable. Our research into TikTok’s AR filter ecosystem reveals that the platform not only develops and promotes beauty filters shaped by intersectional biases but also actively curates (in)visibility by incentivizing filter use Doh et al. 2025a . Prior investigations Biddle et al. 2020 ; Hern 2020 have exposed internal moderation documents detailing exclusionary content suppression policies, banning users deemed “too ugly, poor, or disabled" from appearing on the For You feed. The suppression criteria included “abnormal body shape," “ugly facial features," “too many wrinkles,’ or “shabby environments." These practices illustrate how the politics of digital embodiment in HCI extend beyond individual filters to platform-wide visibility infrastructures. AR beauty filters function within a broader system of algorithmic governance, shaping who gets seen, how they are represented, and which bodies are deemed desirable in digital spaces. 1.1.1 Postfeminist and Neoliberal Frameworks of Beauty Optimization Within their inherent gender bias, AR beauty filters operate in a postfeminist media culture Gill 2007 ; McRobbie 2004 , where self-surveillance is reframed as empowerment, and normative self-improvement becomes an obligation rather than a choice. Through AI-driven enhancements and opaque platform governance, beauty filters seamlessly integrate into this logic, offering users the illusion of control over their digital self-presentation while reinforcing deeply embedded gendered norms of attractiveness. Under the neoliberal, postfeminist mandate of self-optimization Gill 2007 ; McRobbie 2004 ; Elias et al. 2017 ; Hearn and Banet-Weiser 2020 body-computer interactions become a site of real-time governmentality 2 2 2 Governmentality, as formulated by Michel Foucault, refers to the governance of people through positive means rather than coercion, shaping their conduct by fostering willing participation. Neoliberal governance exemplifies this by both isolating individuals as self-responsible entrepreneurial units and binding them to collective economic and political structures Brown 2015 , of “conduct of conduct" Foucault 2007 through rhetoric of autonomy, self-actualization, and self-realization Madsen 2014 , while the body turns into a profitable site of aesthetic Elias et al. 2017 and glamour labor Wissinger 2015 ; Hearn and Banet-Weiser 2020 –– i.e. , gendered practices that emphasize female attractiveness “at whatever cost to the self" Elias et al. 2017 . Within platform economies, the idealized self is produced and circulated as a tool for market-driven interactions. AR beauty filters thus function as a mode of gendered governmentality. By conditioning individuals to internalize platform-driven beauty standards, these technologies extend neoliberal postfeminist surveillance, where the feminized body is simultaneously hyper-visible and invisibilized, optimized for digital desirability yet interchangeable within algorithmic economies of attention. Figure 1: Analysis of TikTok’s Bold Glamour filter transformations from Doh et al. 2025a . (Left) Example outputs of the filter applied to male and female faces (“*" indicates filtered images). (Center) Disclaimer Block visualizations: P_{D} highlights pixel-level changes, S_{D} represents semantic mask transformations, and T_{D} quantifies intensity variations. A zoom-in on the ‘Mouth’ region reveals how the filter enlarges lips. (Right) Two cases of misclassification: the first row shows a male face receiving female-targeted transformations, while the second row shows a female face with a misclassified male-targeted modification. Faces from Ma et al. 2015 1.2 (Q2) Algorithmic Bias and Aesthetic Exclusion in AR Filters Despite their potential for inclusivity, current implementations of AR

Read the complete text