joellene's digi-journal

Search

SearchSearch

digital opacity

Jul 08, 202411 min read

Originally written as a final project for UC Berkeley’s flagship data ethics class, Data 104: Human Contexts & Ethics

In a datafied world where participation demands total transparency, is there a way to be visible yet unassigned, uncategorized and thus unexploited? In other words, is there a way to be informatically opaque?

Total Transparency

Before solutions, we first turn to the problematic logic of total transparency that undergirds the datafied world. Arguably a manifestation of Deleuze’s control society, this is a world governed by codes, passwords indicating where access to information should be allowed or denied (Deleuze, 1977). Currently, accessing the digital world requires a trade-in of information: our name, phone number, email address, etc. when we first create an account, alongside our consent for our every action to be surveilled. The internet is a one-way mirror, where on one side, we are laid completely bare in a “topology of control”, a node existing only within predetermined informatic parameters (Galloway & Thacker, 2007); yet, on the other side, users have little visibility into what exactly corporations are doing with the data they collect under surveillance capitalism (Zuboff, 2015). Most insidiously, that topology of control is naturalized in that the internet seems like a neutral, apolitical, utopian space of free exploration and discovery. The internet is the world at your fingertips — but it is a world already mediated through a profit-driven company. Even the browsers themselves are companies, and collecting personal and behavioral data is a core part of their business model.

This seeming neutrality reinforces the ideology of total transparency. It dresses platforms and online public spaces as products, which are enhanced by user-personalized, which then requires any and all data about users. Yet, that is problematic on two counts.

One, that imaginary presumes data to be a currency that users should trade in to get access to a product or a service. “If you’re not paying for the product, you are the product,” as Rani Molla writes for Vox (2020). But that personal and behavioral data are as important to our informational personhood as our physical bodies are (Koopman, 2019). We are all statistical subjects in the eyes of the state — to be detached from the data about us in government registries might mean to lose access to a job, a house, a bank. Many of us build our senses of self online, across multiple ‘mains’, ‘spams’ and ‘privates’ on Instagram, and when we get hacked and locked out of our accounts, there is an acute sense of loss even though nothing tangible about us was stolen. To reduce the traces of life we leave online to a currency narrows our expression of individuality online to that of a consumer and a user, not a citizen of a digital platform-based public sphere.

Two, and building on the above, even with algorithmic user personalization, to view people as ‘consumers’ creates a discursive monolith that is assumed to be impacted by a product in the same way, consistently. But there is clear evidence that the harms of data exposure — transparency — are disproportionately borne by minorities. This is seen in the quandary of invisibility / hypervisibility faced by Black people, where they were rendered invisible by Kodak film designed for white skin, but hyperexposed to Facial Recognition Technologies (FRT) correlating skin color and crime (Benjamin, 2019). Issues of hyperexposure are also laid bare in the study of FRT relating to terrorism, which communications scholar Wendy Hui Kyong Chun describes as denying “the mystery of every face”, to render all faces completely knowable, and in doing so racializing the terrorist as Asian and Asian American (2009:26).

Informatic Opacity

Given that logics of total transparency create information and power asymmetry, abstracts informational personhood into parsable input data and disproportionately puts minorities at risk, what is the alternative? Artist-activist Zach Blas forwards the concept of ‘informatic opacity’. He draws this from Caribbean philosopher and poet Édouard Glissant, who wrote that opacity “protects the Diverse” (Glissant, 1999) as an anti-imperialist, minoritarian refusal of a logic of total transparency and rationality, that would transform subjects into categorizable objects of Western knowledge (Blas, 2021). Opacity is therefore a “paradigmatic concept to pit against the universal standards of informatic identification” (Blas, 2021). It sits in contrast to privacy, which is often framed as a ‘right’ à la property rights, falling again into the mistaken view that data is a currency or a possession that should be traded for goods and services. It is also not anonymity, which brings a sense of impersonality and aloofness, like one could pass unseen. Instead, opacity is a state of human existence that accepts the cyborg nature of physical-virtual selfhood and allows visibility without legibility — one can be seen, while unassigned, uncategorized and unexploited (Blas & Gaboury, 2016). 

Implementation: Biometric

Beyond theoretical implications, I want to address whether and how this ideal can be achieved in life. In particular, the next few examples take aim at the way FRT forces an inequitable, harmful transparency on individuals, especially those in vulnerable communities. In Blas’s 2012 work Facial Weaponization Suite, he leverages aggregation to approach opacity, with four masks made from facial data of participants from communities threatened by FRT — the LGBT and Black communities, hijab-wearing women and undocumented immigrants — aggregated till the masks are illegible to biometric technology.

Similarly, fashion designer Rachele Didero taps on mimicry to approach opacity, in her 2023 Manifesto Collection offering garments that trick an object detection solution called YOLO into recognizing dogs, zebras, giraffes or small knitted people inside the fabric. 

Implementation: Digital?

Yet, the above examples only address opacity against biometric FRT. In other words, they solve for a physical transparency enforced by surveillance cameras in ‘real’ life. As established in the problem of transparency, I feel that informatic opacity online is a more pressing issue, because of the current sociotechnical structures of the internet and online platforms. The following proposed ideas are nowhere near full solutions for the problem, however, as that would require a full reimagination of the virtual world and our relationship to it in terms of digital selfhood and autonomy. 

In any case, there are technical solutions being proposed to preserve big data privacy by de-identifying data through three methods: K-anonymity, L-diversity and T-closeness. The main goal of these methods is to sanitize data with generalization that replaces identifying information with other values, as well as suppression of this information, i.e. not releasing some values at all, before analysis. For example, the L-diversity model promotes diversity within a dataset, sometimes through inserting fictitious data (Jain et al., 2016). However, even if these solutions were technically airtight — they are still helpless against correlation and similarity attacks that triangulate data across several datasets to identify individuals — they solve for privacy, not opacity, and operate under the same surveillance-capitalist sociotechnical imaginary of data as a resource to be mined and then ‘refined’ via these technical solutions, without questioning why so much data was collected in the first place.

In the Web3 paradigm, decentralized identifiers (DIDs) could be the solution to a ‘self-sovereignty’ of data. DIDs are issued and stored on verifiable data registries that are autonomous namespaces with decentralized administration (Kesonpat, 2023). There are zero-knowledge proofs and novel signature schemes that enhance current technologies with privacy-preserving properties such as anti-correlation — where data across several ledgers cannot be triangulated back to an identified person — selective disclosure and so on. In other words, Web3 could allow for opacity not through aggregation or mimicry, but through decentralization and boundary-setting, rebalancing the power asymmetry of the internet between individuals and corporations. And yet, in exchange for being opaque to corporations, this solution demands transparency to the blockchain, and that is arguably not solving the problem of visibility without legibility.

Perhaps the closest we could get to true online opacity would be a complete overhaul of the underlying structure of our current online world, separating user data from the platforms they are collected on. One way to do this would be allowing each individual to have centralized control over their online personal data, as opposed to surrendering to a ‘higher power’ like the blockchain. This is what the Social Link Data (SOLID) protocol aims to achieve. It defines containers of personal data — ‘data pods’ — such that any platform or entity requesting that data needs to undergo standardized authentication mechanisms and access control policies contingent on the approval of the individual. As such, data subjects store their data in their own pod and can “determine for themselves whether an entity requesting to access the data has a legal basis to use the data in a particular way” (Esposito et al., 2022). Users can thus be visible online and participate on online platforms, but illegible to companies unless they actively choose to be, and can revoke that access freely. Achieving opacity, therefore, becomes two steps: separating user data from platforms, and then setting boundaries around them, boundaries that users alone can control. 

However, going down this road introduces new questions of boundary-setting. Separating out personal data from platforms is fairly straightforward — the Article 4 of the European Union’s General Data Protection Regulation (GDPR) outlines ‘personal data’ as “any information relating to an identified or identifiable natural person” such as names, location data, etc. But what about behavioral data? What about our search histories, social media posts, purchase history? This is the data that is monetized by the platform companies that then sell it to advertisers, but this is also the data that needs to be collected for the platforms to function — not just in personalizing our experiences, but simply collecting enough information to function. Would it be possible for behavioral data to also fall under siloed buckets of data pods, owned and gatekept by users? It appears that new innovations, like the SOLID protocol, bring with them possibilities of our ideal, but also new questions that we are still far from answering. 

Informatic opacity, I think, balances at the knife’s edge between Halpern’s perception and cognition. Whether it is possible in its fully realized form is yet to be seen, but I would like to imagine an online world where our machines do not guess at the mystery of every face, or every trace of our digital lives.

Bibliography 

Graph View

Backlinks