Source-linked AI summary
Data Patching
ATM Mizanur Rahman, Syed Ishtiaque Ahmed, Sharifa Sultana
TL;DR
Records that are valid in one institution can lose recognition or value when they cross institutional seams. The paper studies these failures through interviews and extends cross-institutional auditing to citizen services. It proposes CCDA to make seam failures visible, contestable, accountable, and repairable.
Problem
Valid records can lose recognition, credibility, or practical value across institutional boundaries, while prior audit approaches centered on individual systems may miss harms at institutional crossings.
Source: §1 · p. 2, §6 · p. 21Method
The researchers conducted interviews with people in Bangladesh and extended Cross-Institutional AI Auditing to datafied citizen services.
Source: §1 · pp. 2–3Results
The interviews identified missing connections, unequal recognition across existing connections, and connections made obsolete by change, alongside citizen workarounds that do not repair the underlying seam.
Source: §1 · p. 2, §6 · p. 21Takeaways & Limitations
The paper proposes CCDA to make institutional seam failures visible, contestable, accountable, and repairable.
Source: §6 · p. 21The study relies primarily on participant accounts, is situated in Bangladesh, and proposes five CCDA principles that have not yet been implemented or evaluated.
Source: §5.5 · pp. 20–21
Abstract
from arXiv · showhide
To provide data-driven quality services to their citizens, every institution determines the acceptability of the citizen data and datafied mechanisms through institutional protocols, including preset standards and policies. However, data meeting standards at one institution might fail to meet a different institution's standards in subsequent phases of the intended citizen services due to mismatched protocols, leading to data devaluation in the citizen service ecosystem. Terming this phenomenon cross-institutional data devaluation, we investigate its causes and workarounds through interviews with 41 Bangladeshi participants. We found that traditional data auditing mechanisms cannot solely address data devaluation; hence, we draw on our findings and theory of cross-institutional AI audits to propose the Citizen-centered Cross-institutional Data Audit (CCDA). We also discuss design and policy implications of CCDA in HCI and datafied citizen services.
1 Introduction
Citizen records can lose recognition or value when they cross institutional boundaries, even when each institution follows its own rules. The paper examines these failures and proposes a citizen-centered audit approach.
1 Introduction
- Cross-institutional data devaluation describes information recognized and useful in one setting losing recognition, credibility, or practical value in another.Source: §1 · p. 1
- 41 interviews in Bangladesh examined how valid records lose recognition or practical value across institutional boundaries and how people respond.Participants described records that could not move, match, or be treated as equivalent, as well as connected institutions that applied unequal rules or changed over time.Source: §1 · p. 2
- Citizens patch failed connections through extra work, but these workarounds can fail when standards, responsibility, or record-update routes are missing.Examples include paying gatekeepers, reproducing evidence, and assembling fragmented records.Source: §1 · p. 2
- The paper identifies limits of existing audit methods for data patching and develops the Citizen-centered Cross-institutional Data Audit (CCDA), drawing on Cross-Institutional AI Auditing.Source: §1 · pp. 2–3
2 Related Work
Prior work explains how documents gain authority, how infrastructure and institutional seams shape movement, and how audits examine bounded systems. The paper identifies gaps around standalone institutions and failures at their crossings.
2.1 Documents, Classification, and Institutional Legitimacy
- Records can remain accurate and officially valid yet lose recognition or practical value in another setting; less is known about crossings between standalone institutions or changing relationships.Institutional categories, procedures, and document forms shape recognition and access to services.Source: §2.1 · p. 3
2.2 Institutional Seams, Infrastructure, and Breakdowns
- When systems do not fully connect, affected people may have to coordinate separate activities, while their workarounds can advance a case without changing the institutional relationship.Governance by patching, as described by Veeraraghavan, depends on an actor with jurisdiction to change rules, procedures, technologies, or arrangements.Source: §2.2 · p. 4
2.3 Auditing and Institutional Accountability
- Individualistic audits rely on an examinable object, observable behavior, a standard, and an accountable actor, but can miss failures emerging between institutions.The paper extends Cross-Institutional AI Auditing to citizen services, including cases without AI.Source: §2.3 · pp. 4–5
3 Methods
The study uses interviews with people in Bangladesh who experienced difficulty moving records across institutions. The team analyzed accounts across service domains and considered how its positionality shaped interpretation.
3 Methods
- The study examined cross-institutional record difficulties through 41 semi-structured interviews in Bangladesh.Participants described educational, medical, professional, financial, and government-service records.Source: §3 · p. 5
3.1 Participant Recruitment
- Participants were recruited through community networks, word of mouth, and snowball sampling, and had personal experience moving a record between institutions.Eligibility included records rejected, questioned, downgraded, or requiring additional verification.Source: §3.1 · pp. 5–6
3.2 Interviews
- Interviews were conducted in Bengali, in person or over Zoom, and lasted approximately 30 to 60 minutes.38 of 41 were audio-recorded with consent; detailed notes were taken for the other three.Source: §3.2 · p. 6
3.3 Data Collection and Analysis
- The researchers used iterative open coding and thematic analysis, comparing cases across domains to identify missing, unequal, and obsolete institutional connections.They also examined citizens’ workarounds and cases where those workarounds failed.Source: §3.3 · p. 6
3.4 Positionality and Ethical Considerations
- The Bangladeshi researchers used follow-up questions and team discussion of alternative interpretations to mitigate assumptions from their contextual familiarity.The study received institutional review approval, obtained informed consent, and removed direct identifiers.Source: §3.4 · p. 7
4 Findings
The findings distinguish three forms of cross-institutional data devaluation and examine citizen workarounds, including their limits.
4 Findings
- The findings identify relational, hierarchical, and temporal devaluation as three ways cross-institutional failures emerge.Source: §4 · p. 7
- A fourth theme examines citizen-borne workarounds used to make failed institutional connections work.Source: §4 · p. 7
- A final theme identifies limits of workarounds and cases where people still cannot make the institutional connection work.Source: §4 · p. 7
4.1 Missing Links Between Standalone Systems
Missing links between standalone systems devalue records when institutions cannot transfer them, treat different evidence as equivalent, or match records across systems.
4.1.1 No Transfer Channel Between Systems.
- Without an accepted transfer pathway, recognized records or funds could not move between institutions, leaving them usable only in the originating system.Examples include a school system without a transfer procedure for a recognized madrasa record and a bank that could not transfer funds from another bank.Source: §4.1.1 · p. 8
4.1.2 No Basis for Cross-Institutional Equivalence.
- Seven participants described valid experience or records that could not substitute for the different evidence another institution required.A government plumbing position required a BTEB trade certificate rather than a private-company experience certificate; BRAC Bank did not consider a borrower’s BRAC NGO repayment history.Source: §4.1.2 · p. 8
4.1.3 No Cross-System Record Linkage.
- Without cross-system matching, small differences in names or IDs could prevent institutions from recognizing that records referred to the same person.P17 had to obtain a university letter confirming that “Mohammad” and “Md.” referred to the same student.Source: §4.1.3 · p. 9
4.2 Unequal Recognition Across Existing Couplings
Existing institutional connections did not ensure equal recognition: institutional status, a record holder’s social position, and geographic origin shaped how records were treated.
4.2.1 Rigid Recognition Based on Institutional Status.
- Five participants described rigid judgments that reduced the value of records from institutions considered lower status, despite their formal validity.A UGC-approved university degree and a Class 7 transfer certificate received less favorable treatment because of the issuing schools’ status.Source: §4.2.1 · pp. 9–10
4.2.2 Social Hierarchies of Recognition.
- Four participants described existing institutional connections working less readily when records were presented by people with less social influence or authority.P26 brought valid inheritance, ownership, and identity documents to the land office, but officials would not process them without a male relative.Source: §4.2.2 · p. 10
4.2.3 Geographic Hierarchies of Trust.
- Five participants described more rigid verification for records or applicants associated with locations outside the capital.A death certificate from outside Dhaka required additional verification, while a passport applicant from Cox’s Bazar described a process taking six months to a year rather than a month in Dhaka.Source: §4.2.3 · p. 10
4.3 Institutional Coupling Becomes Obsolete
Institutional connections that once worked became obsolete after political, infrastructural, or rule changes altered recognition or compatibility.
4.3.1 Recognition Changed After Political Transition.
- After a political transition, new officials refused to recognize training certificates issued under the previous administration, although the certificate itself had not changed.The participant was told to take the matter to a higher authority.Source: §4.3.1 · p. 11
4.3.2 Compatibility Broken by Infrastructural Transition.
- Seven participants described previously working connections disrupted when institutions introduced new technologies, systems, procedures, or data standards.After old-age allowance payments shifted to mobile banking, one recipient’s record was missing from the online system and her NID and SIM information did not match; her allowance remained suspended for months.Source: §4.3.2 · p. 11, Figure 2 · p. 11
4.3.3 Compatibility Broken by Institutional Rule Changes.
- After institutions changed their rules, records accepted for years could no longer be used in the same process.BRTA stopped accepting a Civil Surgeon’s medical fitness report, limiting accepted reports to designated centers or certain government hospitals in Dhaka.Source: §4.3.3 · pp. 11–12
4.4 Citizen-Borne Workarounds for Interoperability Gaps
Participants used citizen-borne workarounds to move individual cases through institutional connections, but these efforts did not create lasting connections for others.
4.4.1 Paying Gatekeepers to Activate Formal Pathways.
- In two reported cases, formal connections between institutions existed but did not work in practice until participants paid intermediaries: P2 paid around BDT 2,500 before document confirmation, and P10 around BDT 4,000 before the process moved forward.Source: §4.4.1 · p. 12
4.4.2 Reproducing Evidence Inside the Receiving Institution.
- Five participants reproduced evidence inside a receiving institution when it would not accept evidence from elsewhere.Examples include repeating medical tests at Apollo and receiving a Hepatitis B vaccine again at a hospital on the visa agency’s accepted list.Source: §4.4.2 · pp. 12–13
4.4.3 Manual Integration of Fragmented Records.
- Six participants manually combined fragmented records from different institutions into one form of proof so their cases could proceed.P39’s family gathered deeds, mutation documents, khatian records, and tax receipts to support an ownership claim after the records had not been updated across offices.Source: §4.4.3 · p. 13
4.5 Limits of Citizen-Borne Workarounds
Citizen-borne workarounds could not bridge institutional seams when standards conflicted, no actor owned the handover, or an old-to-new system path had closed.
4.5 Limits of Citizen-Borne Workarounds.
- These cases show that citizens’ repeated efforts could not repair institutional connections that required shared standards, seam-level authority, or a functioning update path.The workarounds included carrying records, seeking extra confirmation, changing document formats, and bringing supporting evidence to offices.Source: §4.5 · p. 13, §4.5.1 · p. 14, §4.5.2 · p. 14
4.5.1 When Institutions Could Not Agree on a Shared Verification Standard.
- Participants could not create a shared verification standard when institutions required incompatible forms of confirmation, despite collecting evidence or changing records.A training center’s official letter did not meet a government office’s portal requirement; a bank rejected a company’s printed record without an original physical signature.Source: §4.5.1 · p. 14
4.5.2 When No Actor Was Responsible for the Crossing.
- Neither office could complete a handover when each had authority only within its own system, leaving no actor responsible for the crossing.The municipality accepted the old tax receipts as genuine but lacked the records, while the Union Parishad no longer had authority to update them.Source: §4.5.2 · p. 14
4.5.3 When the Update Path Could Not Be Restored.
- Workarounds could not restore older records when the route into a replacement system had closed, even when officials accepted the supporting evidence as genuine.P40 could not link his 13-digit birth certificate to the current system; P32’s father’s manual record could not enter the online portal after it closed.Source: §4.5.3 · p. 15
5 Discussion
The discussion shows how valid records lose value at institutional seams and how citizens patch failed connections through labor that does not fix the underlying relationship. It argues that these patterns extend across institutional boundaries and over time.
5.1 How Valid Records Lose Value at Institutional Seams
- Records can remain accurate and officially valid yet lose practical value when institutional connections are missing, unequal, or obsolete.Institutions may follow their own protocols while a record still fails at the seam between them.Source: §5.1 · p. 15
5.1.1 Recognition Does Not Always Carry Across Institutions.
- Recognition in one institution does not ensure that another institution can use the record.Examples include work experience recognized by an employer but not treated as equivalent to a required qualification, and madrasa education without a transfer pathway for school admission.Source: §5.1.1 · pp. 15–16
5.1.2 Valid Records Can Still Be Valued Differently.
- Formally valid records can receive less value because of the status, geography, or social position associated with their issuing institution.The discussion describes a UGC-approved degree and a transfer certificate receiving less value because of their issuing institutions.Source: §5.1.2 · p. 16
5.1.3 Institutional Change Can Make Valid Records Unusable.
- Institutional change can make an unchanged record unusable when recognition or verification practices shift.A medical certificate accepted for years became unusable after BRTA changed which institutions it recognized; another certificate’s issuing office stopped verifying it after a political transition.Source: §5.1.3 · p. 16
5.1.4 Proximity Does Not Guarantee Equal Recognition.
- National proximity and shared organizational ownership do not guarantee that records or institutional value move across institutions.A BRAC NGO repayment history was not recognized by BRAC Bank, and Citizens Bank lacked a transfer mechanism to Janata Bank.Source: §5.1.4 · p. 16
5.2.1 Citizen-Borne Coordination Work.
- Citizens perform coordination, pay intermediaries, reproduce evidence, or manually combine records to keep services moving, but these workarounds do not create lasting institutional connections.P2 paid to move an already-submitted passport verification forward without adding evidence; P25 carried reports between hospitals and repeated tests.Source: §5.2 · p. 16, §5.2.1 · pp. 16–17
5.2.2 From Governance by Patching to Data Patching.
- Data patching is citizen-borne work that makes a connection function for one case without repairing the underlying seam.Unlike governance patching, citizens lack jurisdiction to change institutional arrangements; a successful individual outcome may conceal payments, repeated procedures, travel, delay, or personal labor.Source: §5.2.2 · p. 17, §5.2.3 · p. 17
5.2.3 When Successful Patching Hides Institutional Failure.
- Successful patching can hide institutional failure because an eventual service outcome does not reveal the labor or costs needed to achieve it.A successful individual case does not necessarily mean the institutional seam has been fixed.Source: §5.2.3 · p. 17
5.3 Why Individualistic Institutional Audits Can Miss the Seam
Individual audits can miss failures that exist between institutions or across system transitions, motivating CCDA’s focus on auditing institutional seams. The proposed principles center citizens’ experiences, visible decisions, accountable repair, continuity, and ongoing evaluation.
5.3 Why Individualistic Institutional Audits Can Miss the Seam
- Separate audits can find that each institution follows its own rules while missing a failure in the connection between them.The seam itself therefore needs to become an object of audit and accountability.Source: §5.3 · p. 17, §5.3.3 · p. 18
5.3.1 No Shared Standard to Measure Against.
- When institutions have no shared standard, individual audits can assess each institution’s compliance yet leave the missing standard between them unexamined.The examples include incompatible verification letters and portals, and digital versus paper records requiring physical signatures.Source: §5.3.1 · pp. 17–18
5.3.2 No Accountable Actor at the Seam.
- When no actor has authority over a handover, audits of responsible actors within each institution leave an accountability gap at the seam.The municipality and Union Parishad each had officials responsible within their boundaries, but neither could complete the transfer.Source: §5.3.2 · p. 18
5.3.3 No Continuity Over Time.
- Auditing an earlier system and a current system separately can miss the loss of continuity that makes older records unusable.The route for moving records between the two arrangements may disappear even when each system follows its own rules.Source: §5.3.3 · p. 18
5.4 Toward Citizen-centered Cross-institutional Data Audit (CCDA)
- CCDA extends cross-institutional AI auditing to datafied citizen services by making institutional seams, rather than institutions alone, the object of audit.The framework comprises five interrelated principles.Source: §5.4 · pp. 18–19
5.4.1 Principle 1: Conduct Citizen-centered Audits.
- Auditors should trace crossings from citizens’ experiences and compare issuing-institution records, receiving-institution records, and what people experienced.Auditors should actively reach people and protect testimony, including accounts of delays, costs, repeated visits, and informal payments.Source: §5.4.1 · p. 19
5.4.2 Principle 2: Make Recognition Decisions Visible and Contestable.
- Services should document contested recognition decisions, delays, and repeated evidence requests so citizens can see where a record stopped and how to contest the decision.The framework also calls for an accountable participatory authority with jurisdiction to coordinate seam repairs, negotiate shared standards, and assign responsibility.Source: §5.4.2 · pp. 19–20
5.4.4 Principle 4: Build Backward Compatibility into Institutional Transitions.
- Institutional transitions should preserve the usability of earlier records through backward compatibility and working update or revalidation processes.Continuity mechanisms should identify dependent legacy records and keep them actionable through changes in technology, rules, administration, or jurisdiction.Source: §5.4.4 · p. 20
5.5 Limitations and Future Work
The study’s findings are bounded by its participant accounts, Bangladesh setting, and cross-domain scope; the proposed CCDA principles remain unevaluated. Future work should broaden evidence and test the framework.
Limitations
- The study relies primarily on participant accounts and did not interview officials from all involved institutions or independently verify every rule or explanation.The authors note that institutional actors may describe some cases differently; Bangladesh-specific experiences may differ elsewhere, and targeted snowball recruitment does not establish prevalence.Source: §5.5 · p. 20
- Covering six institutional domains enabled cross-domain patterns but limited in-depth analysis of any single system.Some accounts concerned events from years earlier and may reflect incomplete recall.Source: §5.5 · p. 21
Limitations and Future Work
- The authors did not conduct audits or interview auditors, and the five CCDA principles have not yet been implemented or evaluated.Future work should combine participant experiences with officials, records, and observations; compare contexts; follow records longitudinally; and implement and evaluate CCDA.Source: §5.5 · p. 21
6 Conclusion
The conclusion frames cross-institutional data devaluation as a failure of connections between institutions, not simply of individual institutions. It identifies three forms of devaluation, describes citizen workarounds, and proposes CCDA to make seam failures repairable.
6 Conclusion
- The study identifies missing connections, unequal recognition across existing connections, and connections made obsolete by institutional change as three forms of devaluation.Source: §6 · p. 21
- Citizen-borne workarounds may move an individual case forward but do not repair the underlying institutional seam.Source: §6 · p. 21
- The proposed CCDA framework aims to make seam failures visible, contestable, accountable, and repairable.Source: §6 · p. 21
- The paper shifts attention from whether individual institutions work correctly to whether their connections work for people relying on multi-institutional services.Source: §6 · p. 21