Bias Cuts and Data Dumps

Introduction “Patterns are everywhere”, design researcher Anuradha Reddy told her virtual audience at the 2023 speaker series hosted by Brilliant Labs, a Canadian non-profit focussed on experiential digital learning and coding (Brilliant Labs / Labos Créatifs). Like other technology fora, this public-facing series offered designers an opportunity to highlight the accessibility of code. But unlike many such fora, Reddy’s code was worn on the body. Sitting at the now-standard webinar lectern, Reddy shared a flurry of images and contexts as she introduced a garment she called b00b, a bra that she created in 2021 to probe the encoding of more than aesthetic possibility. Her presentation included knotted motifs of Andean Quipus; symbolic arcs of Chinese Pan Chang knots; geometric transformations of African American cornrow hairstyles (Eglash and Bennett, Brilliant Labs / Labos Créatifs). She followed the patterned imagery with questions of uncertainty that are often central for design researchers like her. Facing what might be a possible swipe, tap, or otherwise engagement, a technologist cannot fully determine what a user does. But they can “nudge”, a term popularised by behavioral economists Richard H. Thaler and Cass R. Sunstein in 2008 and later propagated within technoscientific discourses on risk (see Duffy and Thorson; Rossi et al.; Thaler and Sunstein). Adjacent bodies of scholarship frame the related concept of trust as a form of compliance (Adam et al.; Gass and Seiter). The more trustworthy an interface, the more likely a user is to comply. Rooted in social-psychological precepts, this line of scholarship frames trust less as a condition than a perception. When a user trusts an indicator light, for example, an app is more likely to see increased acceptance and engagement. Reddy approaches trust from and with b00b, an emphatically intimate (soft, pliable, textile) artifact. “How do we use these … perspectives to deal with uncertainty and things we do not know yet in the future?”, Reddy asks her Brilliant Labs audience (Brilliant Labs / Labos Créatifs). To make this argument, I examine Reddy’s b00b in conversation with a legacy feminist textile performance that brings questions of embodiment (and embodied trust) to an ostensibly disembodied technocratic scene. b00b is a decorative bra that emulates two-factor authentication, or what Reddy calls “b00b factor authentication.” The bra uses its two cups to verify a user’s access to a Website describing the project. With this interaction, the bra is self-referential—asking users to unlock a link that brings them back to someone’s chest. In practice, b00b asks users to scan a bra cup that relies on scanning the companion bra cup for a second passcode. Rather than messaging users, an initial passcode that triggers a second passcode sent by text message, the engagement requires bodily proximity. The bra cups take the place of electronic media (such as the text message) so that a close encounter with the bra enlivens digital trust. Under these circumstances, a trusted user becomes a risk-taker—gaining access while transgressing personal boundaries. In the sections that follow, I thread conversations on digital and algorithmic trustworthiness with critiques of trust and compliance that pervade Reddy’s 2021 handmade experiment. To date, technology analysts tend to treat trust as a perception: feelings of confidence in a person or thing (Gilkson and Woolley). As Natasha Schüll notes, a user might trust a slot machine but might miss its implications for further (and potentially excessive) gambling. Additionally, media scholars such as Evgeny Morozov have since mapped this addiction principle within social media development, pointing to a familiar science of incentive structures, gamification dashboards, and behaviour-change techniques, each designed to raise user engagement and keep people in apps longer. Thinking with Reddy’s work, I argue that trust can reveal an embodied desire, something momentarily felt and differentially shared (see also Gregg; Sharma; Irani). Reddy frames the weft of woven material as code, the purl and knit stitches of knitting as binary, and the knots of rope as algorithms. She urges her audience to see fabric as a means of challenging common assumptions about technology. With needles and thread, she proffers algorithmic trust as a relational ethics. In Technology We Trust From a design perspective, trust grows from the strategic balancing of risk and uncertainty (Cheshire). Users who find a digital feature reliable or trustworthy are more likely to grow their engagement and convince others to join in (Hancock et al.). In a recent analysis of the overlapping dynamics of algorithmic trust and bias, communication and information scholars Jeff Hancock, Mor Namaan, and Karen Levy (95) argue that machine learning tools such as the Chrome extension Just Not Sorry often replicate bias within training data. The extension disproportionately alerts femme users when they use qualifying words like “sorry”, and “I think”. In ​​other contexts, Hancock and colleagues suggest, an AI-aided tool may help mitigate interpersonal biases since if it “imparts signals of trustworthiness between peer-based social exchange partners, these countervailing cues may neutralise stereotypes that would otherwise impede the transaction” (ibid). Here, the signal of trustworthiness holds the promise of accountability. But because the signals focus on cognition (manipulating an individual’s perceptions), what they refer to and how they may alleviate harms caused by entrenched cultural bias remains less clear. Grounded in social-psychological tenets, technology analysts codify trust as the relationship between two primary concepts: risk and uncertainty. As information scholar Coye Chesire (50) explains, “trust is not simply the absence of risk and uncertainty. More accurately, trust is a complex human response to situations that are rife with risk and uncertainty”. Through a range of controlled methods including observations, self-reports, survey questions, and the experimental conditions of a lab study, researchers measure the trustworthiness of user interface features as assessments of risk and uncertainty that explain differing motivations for use and disengagement. For example, design researcher Nick Merrill’s and Cheshire’s study of heart rate monitors finds that listening to an acquaintance's normal heart rate can lead to negative trust-related assessments in challenging contexts such as waiting to meet the acquaintance about a legal dispute. Parallel work by Hancock and colleagues uses self-reports and large-scale experiments on platforms like Facebook to map the significance of AI-enabled curation features like news feeds (Hancock et al.). As a psychological state, trustworthiness tends to indicate a behavioral metric that can be numerically encoded and individually addressed. By measuring trust-infused dimensions of user activity, analysts seek to systematically identify new ways of scaffolding trust-building behaviour by manipulating perception (Hancock, Namaan, and Levy), ultimately convincing a user to comply. A core goal is to maximise participation. The US government applied these principles to mass data collection and dissemination efforts during national census such as the COVID response (Halpern). But a secondary effect grows from the political-economic dimensions of user experience. Through compliance, users become easier to place, measure, count, and amend—a process Michelle Murphy names the economisation of life. When people’s certainty in interpersonal relationships grows, “the source of uncertainty then shifts to the assurance system, thereby making trustworthiness and reliability of the institution or organisation the salient relationship” (Cheshire 54). For instance, we may trust people in our text messages because we meet them face to face and put their numbers in our phones. But once we trust them, this assurance moves to our social media service or cellular phone provider. The service that manages our contacts also preserves the integrity of our contacts, such as when a messaging platform like WhatsApp automatically updates a cell phone number without our knowledge or explicit consent. Conversely, feelings of assurance in a digital interface feature may dwindle with decreased feelings of assurance by a platform. Until November 2022, users may have trusted someone with a blue checkmark on Twitter more than someone without one, even if they did not trust them at an interpersonal level. But with a chaotic acquisition that, according to a Washington Post report (Weatherbed), led to shifting check mark meanings and colours, this assurance grew more complicated. Murphy (24) might call these quantitative practices enriched with affect the “phantasmagrams” of rationalised assurance. Like a check mark that may or may not index a particular measure of confidence, excitement or worry, these shifting dynamics reveal the “trust and belief that animates numbers” (52). A less considered outcome of this framing is how individuated expressions of distrust (situations that foster psychological and physiological concern, skepticism, or fear for a single person) overshadow its complement: non-unconditional expressions of care. How might a user interface foster networks of connection for self and community? As Anna Lauren Hoffmann suggests, efforts to thwart algorithmic discrimination undergird this conundrum—“mirroring some of antidiscrimination discourse’s most problematic tendencies” (901). The particular value placed on trust often proceeds quick-fix techniques such as multi-factor authentication and cryptography that reduce trust to a neutral transaction (see Ashoori, et al.). In this discussion, design researchers have only begun to conceive trust (and distrust) as a deeply embodied process. Looks, Cuts, and Scans Reddy’s b00b invites audiences to explore em

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.