Tag: education

Information-Erasing Artefacts

silvima86

ID artefacts are bearers of information. On the subject’s identity, but also on many other aspects concerning them: their entitlements, their household size, the markers to which certain entitlements are connected. As such, ID artefacts play a significantly greater role than simply affording authentication at the point of access of a given service: they tell significant parts of a person’s story, materialising or denying the opportunity to access particular entitlements. The information-bearing role of such artefacts is, to say it with the latest, great book by Mareile Kaufmann, a route to “Making Information Matter”: information on users, but also on what their demographics, citizenship status, family and socioeconomic classification mean to accessing key means to livelihood.

Work on India’s ration cards, in primis the great book “In Pursuit of Proof: A History of Identification Documents in India” by Tarangini Sriraman, has expanded on the materiality of identification documents and its consequences. Ration cards are the central document for accessing subsidised items under India’s Public Distribution System (PDS), the country’s food security system which provides subsidised foodgrains and other commodities to entitled households. Crucial to the ration card is its nature as a household-based document: moved from an individually issued form in July 1952, the card is, Sriraman notes, a reflection of the use of «family as a category of governmentality» as made by the Indian government. It is in virtue of household membership, rather than their own individuality, that people are subjects of rights in the PDS: and at the same time, it is in the materiality of the ration card that key information is contained and enacted. Pre-biometric, older forms of ration cards consisted of a paper booklet with numbered spaces, where each month, at the time of ration delivery, the ration dealer would put a physical sign or stamp indicating that the ration had been collected. This provided central information to (a) the ration dealers themselves, and (b) the users, for whom an empty space on the card constituted the material proof that a ration had not been delivered.

Paper-based receipt, Karnataka, April 2018

The transition to digital ID artefacts, moving core aspects of identification, authentication and authorisation to the digital world, may leverage digitality to strengthen the core aspects of accountability and transparency promised in the transmission of key information. At the same time, the question on which information is preserved – and which is lost – in the transition from a physical to a digital ID artefact remains open. A core example is offered by the work of Grace Carswell and Geert De Neve, who investigated the transition of India’s state of Tamil Nadu from physical ration cards to smart cards, carrying a scannable QR code to be used by people for authentication in ration shops. The smart card, note Carswell and De Neve, promises easy and secure identifiability of individuals at the point of sale: a QR code affords retrieving the person’s name, family size, and all details connected to determination of entitlement for that month. At the same time, the card promises delinking of the act of identification from the person’s fingerprint, which constitutes advantages especially for elderly people whose mobility – and ability to physically visit the ration shop – is limited. All these notes make the smart card an artefact that promises informational justice, defined as the ability of an ID artefact to reflect and enact correct information on subjects.

But the parallel story, concerning information that is lost in the transition to digital ID artefacts, may be as important as that on information that is gained. Work from Carswell and De Neve illustrates also this point. As their research on the Tamil Nadu PDS reveals, the transition to a smart card eliminated the material component of the acknowledgement of ration delivery: before it was a stamp on the ration card booklet, physically provided by the ration dealer, that provided proof of delivery, and its absence could be used as a means to prove that ration collection had not happened yet. In other words: an absent collection stamp was a tool of bargaining power. With the transition to smart cards, however, proof of delivery was delegated to a text message on the registered person’s phone: this posed problems of a gendered nature, due to ration cards often being registered in the name of a male member of the family rather than that of women. It also posed problems of accessibility, due to text messages being sent in English, and many PDS users in the state not speaking this language. But at the heart of the problem, is a transition where the bargaining power of an empty space on a ration card – physically showing that collection has not happened – being lost: users cannot show the “lack of reception” of a text  message, which leaves them with no option to prove that a month’s ration delivery has not happened. This information is lost in the digital artefact, while the physical artefact, the paper booklet, was explicitly designed to carry it.

In my forthcoming book Unfair ID, I propose the notion of information-erasing artefacts to conceptualise the loss of information that transition from physical to digital forms of ID can involve. The transition to smart cards in the PDS is a poignant example: another one is in the work of Margie Cheesman, who studied a blockchain-based digital wallet offered to women refugees in a cash-for-work programme in Jordan. Women refugees, Cheesman notes, used to receive an envelope containing the cash corresponding to worked days and hours: his made it possible for them to be sure of the exact correspondence of salaries to the worked time, as well as ensuring actual usability of the cash. But the transition to the blockchain-powered, biometrically enabled EyePay machines introduced informational hurdles that were not present before: recognition of the user through iris-based matching produced a paper receipt, reporting only some of the key information needed. The receipts, refugees reported to Cheesman, did not report what each payment was for, and neither how many and which working days the payment corresponded to: they only enabled the user to collect money from a cash counter, at the same time not knowing which days’ salary payment was being collected. The issue, Cheesman continues, transposes to the material usability of cash, which stands in stark contract with the “balance” held in a digital wallet: paper receipts, she notes, were held securely to preserve their materiality, with women folding receipts into their bras during the working day to prevent the ink from rubbing off.

The transition to smart ration cards in a large food security programme, as well as that to a blockchain-powered digital wallet in a cash-for-work scheme, induce thinking on a type of information loss that a transition to digitality – with its aims of transparency and accountability – does not necessarily contemplate. Against this backdrop, I find the idea of information-erasing artefacts a powerful one to conceptualise this type of informational loss. It is not, I argue in the book, only unintended loss: it is meaningful erasure, a point that the information-deficient receipts described in Cheesman’s work bring to material visibility. The informational advantages of digitality need attention, as it is proven for instance by electronic weighing machines in India’s PDS: electronic weighing of goods visualises quantities sold, and reduces the scope for manipulation at the last mile of large food security programmes. But to be complete, the story of informational justice in digital ID needs theoretical tools to conceptualise erasure, and that is where the idea of information-erasing artefacts can be of some use. More work is needed on the informational consequences of such a game-changing transition.

Unfair ID and Algorithmic Welfare

silvima86

My book is finally submitted! I am so happy that I barely find the words. It was written in tea stalls, in bars, on buses, on planes, even in a police station. It was mostly written on copybooks, in the materiality of pen and paper where, far away from computing, I find my true writing self. It was submitted on a beautiful Norwegian afternoon, but in its pages are stories of digital ID injustice and resistance from my early days of human rights work in Palestine, all through my work on India’s Public Distribution System, up to narratives on digital ID research from Kenya to Jordan, from Colombia’s algorithmic welfare to Uganda’s introduction of digital ID in social protection. The chapters of my book Unfair ID are a hymn to hope, the hope that, by better understanding injustice in digital ID systems, we can be equipped to build systems that restore the fairness lost by undue exclusions, surveillance and denial of basic entitlements.

The book does not have an ending, at least not one in the traditional sense of the term. That frustrated me a bit when I realised it, but after a bit I thought – maybe unconsciously, it was always planned to have an open ending. My last chapter, titled “Imagining Fair ID”, seeks to build on awareness of injustice to imagine fairer forms of digital ID. It is the longest chapter in the whole book. For a text that promises to build on unfairness to imagine forms of ID that liberate instead of surveilling, that include instead of excluding, this is a greatly encouraging result. The chapter puts the digital ID discourse in relation with that on digital rights, and I am delighted that the splendid report by Giulio Coppi – Mapping humanitarian tech: Exposing protection gaps in digital transformation programmes – was released in time for the book to engage it. With over 400 documents reviewed, 40 interviews and many months of work, the report offers a groundbreaking illumination of the role of private companies in digital humanitarianism, offering up-to-date statistics and highlighting the explicit inscription of digital ID work in the domain of digital rights.

Ration Shop, Karnataka, April 2018

One of the topics reviewed in my final chapters is that of algorithmic welfare systems. A brilliant article published by Tapasya, Kumar Sambhav and Divij Joshi in January 2024 brought the topic back to my attention: I say “back”, because Bidisha Chaudhuri’s work on programmed welfare already illuminated the algorithmic aspects that the Aadhaar biometric database allowed infusing in India’s Public Distribution System (PDS). Tapasya et al., however, bring the argument forward by studying Samagra Vedika, an algorithmic system that, implemented in India’s state of Telangana, crosses different databases to determine the list of people eligible for foodgrains and other goods under the PDS. With inbuilt assumptions of algorithmic fairness, in this case meant as fairness towards users of a programme on which millions of households depend for sustenance, the algorithmic system flags what it determines as non-eligible beneficiaries, who due to asset-owning, incomes or other factors are barred from receiving essentials provisions under the PDS.

But the algorithm seems to replicate the flaws that research on the Aadhaar-based PDS, and on the shift to a targeted PDS before it, already denounced. The story of elderly widow Bismillah Bee, told in the opening of the article, is illustrative of this matter: her deceased husband, a rickshaw driver, was erroneously tagged as a car owner, resulting in the family losing PDS entitlement. Research on the case revealed how the man had been mistaken for a car owner with a name similar to his: authorities chose, however, to follow the algorithm’s flawed decision, even when faced with the wrongful input mechanism. The case, note the authors, is epitomic of many more: from 2014 to 2019, the state of Telangana cancelled over 1.86 million existing ration cards and rejected 142,086 new applications without notice. While the technology was claimed to be of high precision, a Supreme Court-imposed re-verification of cancelled cards in April 2022 suggested that at least 7.5 percent of the cards were wrongfully rejected, meaning an equal number of households becoming unable to access basic food provisions.

The argument of continuity, according to which the algorithm barely replicates the exclusions of needy users generated in the PDS for many years, has some foundation. The shift to a targeted system in 1997 meant that a capped number of entitled households had to be established for each Indian state, determining, in turn, the exclusion of many genuinely entitled households from a system of social protection. A logic prioritising the fight to wrongful inclusions has later pervaded the introduction of the Aadhaar-based PDS: algorithmic systems like Samagra Vedika, it can be argued, are designed to replicate a process that excludes those deemed as non-entitled, without necessarily offering opportunities of redressal. Bee’s case is, again, epitomic of a logic that bars the “flagged” person from any provision, without demonstrating the same anxiety to rectify wrongful exclusions. While the “programmatic welfare” theorised by Chaudhuri is indeed qualitatively different, it can be said to reinstate the same exclusionary logic of previous systems.

And at the same time, there is more to the narrative than replication. While I was finishing my book, the book “Algorithms of Resistance: The Everyday Fight against Platform Power”, by the brilliant Tiziano Bonini and Emiliano Trere’, was published: the book unpacks exactly how algorithmic violence can be turned upside down, by communities building resistance through the same algorithmic tools that led to injustice. That is true for algorithmic welfare as well: the digitally-infused nature of the system means that actors willing to instill fair values, such as the data-for-dignity proposition theorised by Joan Lopez for Colombia’s prosperity scoring system Sisbén, have the scope and material ability to do so. This turns the discourse on algorithmic welfare from one of injustice into one of hope: it is the hope that, by recognising algorithmic leveraging in social welfare systems, exclusions can be combated rather than enabled, and fairness can be designed directly in the programmes rather than bypassed. It is the same hope that, referred to digital identity, has informed my book Unfair ID.

This book will be published in a time of tension between injustice and resistance, and stories of algorithmic welfare have so much to add to it. I cannot imagine a better site to imagine new routes to rebuilding fairness!