Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).
📌 EXECUTIVE & METHODOLOGICAL FRAMEWORK: How to Read This Document This white paper goes beyond a standard advocacy text; it serves as a foundational methodological pillar within a longitudinal participatory observation corpus. Readers, institutional reviewers, and policymakers should evaluate this document through the following core principles: Civic Observation as an Evidence Layer: Citizen auditing bridges the gap between everyday encounters with automated systems and formal institutional review, providing structured records without replacing regulatory bodies. Methodological Discipline over Speculation: Aligned with rigorous recording standards (Observe → Record → Verify → Notify → Archive), this framework emphasizes separating raw primary evidence from analytical interpretation. Traceability and Institutional Memory: By standardizing how digital and algorithmic events are documented, this approach builds an immutable, cross-referenced evidentiary architecture that supports independent verification. Introduction Artificial intelligence governance cannot rely exclusively on legislation, institutional oversight, technical compliance, and internal auditing. As AI-enabled systems increasingly influence communication, access to information, employment, public services, education, digital platforms, and everyday decision-making, citizens are also becoming direct observers of their effects. This raises a practical question: How can citizens and independent observers document AI-related events in a consistent, verifiable, and responsible way? Citizen Auditing proposes a simple answer: create a repeatable recording standard that separates what was observed from what was subsequently interpreted. The objective is not to turn citizens into lawyers, regulators, or professional investigators. The objective is to enable ordinary people to create reliable records that can later be independently examined by institutions, researchers, journalists, civil society organisations, regulators, or other competent bodies. The basic principle is: Observe → Record → Verify → Notify → Archive Documentation is the first act of civic accountability. 1. Why Citizen Auditing Matters for AI Governance AI governance is often discussed at the level of policies, regulations, technical standards, and organisational compliance. These mechanisms are essential. However, there is another layer of governance that begins at the point where people actually encounter technology. A citizen may observe: an unexpected automated decision; a change in the behaviour of an AI-enabled service; misleading or inconsistent AI-generated information; a possible transparency problem; an unexplained moderation or recommendation outcome; a recurring pattern affecting users; a publicly visible statement that appears inconsistent with observed system behaviour; or a significant change that may warrant further examination. A single observation does not necessarily establish wrongdoing. It does, however, create a potential record. The quality of that initial record can determine whether the event can later be independently verified. Citizen Auditing therefore does not attempt to replace formal oversight. It creates an additional evidence layer between everyday experience and institutional review. 2. The Five-Step Citizen Auditing Standard OBSERVE Record what actually happened. The observer should distinguish direct observation from assumptions about why something happened. Where possible, record date and approximate time, platform or system involved, relevant page, interface or process, what was displayed or experienced, the circumstances in which it occurred, and any immediately available supporting material. The first principle is simple: Record the event before explaining the event. RECORD Preserve the available evidence. Depending on the circumstances, this may include screenshots, documents, publicly available URLs, system messages, notifications, correspondence, transaction or submission records, timestamps, relevant versions of documents, or other lawful and appropriately obtained records. A record should retain enough context to allow another person to understand what was actually observed. Where possible, the original source should be preserved rather than relying exclusively on a later summary. VERIFY Separate observation from interpretation. Verification can involve checking whether the source is authentic, whether the information can be independently reproduced, whether another source confirms the event, whether the relevant document or webpage has changed, whether the same phenomenon occurs under comparable circumstances, and whether an independent institution or source has reported a related development. The purpose is not to manufacture certainty, but to make the level of certainty visible. NOTIFY If the observation appears significant, it can be communicated to an appropriate institution or responsible organisation. Depending on the circumstances, this could include the relevant service provider, a public authority, a regulatory body, an ombuds institution, a research organisation, a civil society organisation, or another competent recipient. A notification should describe the evidence without presenting an unverified interpretation as established fact. This distinction is fundamental to responsible citizen participation. ARCHIVE Preserve the complete record. An archive should ideally contain the original observation, supporting evidence, dates and timestamps, copies of notifications, delivery or submission confirmations, subsequent responses, independent corroboration, and later developments. This creates a chronological record rather than an isolated screenshot. Over time, individual records can become a structured body of evidence that is available for independent review. 3. A Simple Evidence-Level Matrix To avoid confusing evidence with interpretation, Citizen Auditing uses a four-level structure: A — Primary Record: A direct or official record. Examples include an official document, an original decision, a direct system record, an original communication, or other primary material. B — Notice Record: A record showing that an issue was communicated to an institution or responsible organisation. Examples include a formal submission, notification, complaint, delivery confirmation, acknowledgement, or institutional response. C — Independent Corroboration: Evidence originating independently from the original observation. Examples include another independent record, a separate institutional development, independent reporting, reproducible observations, or other external confirmation. D — Analysis: Interpretation, hypothesis, or analytical inference. This level can be valuable, but it should remain clearly distinguished from primary evidence. A hypothesis should never be presented as an A-level record. This separation is particularly important in AI governance, where complex system behaviour can easily lead to premature conclusions about causation or intent. 4. Citizen Auditing and Responsible AI Citizen Auditing can complement existing AI governance mechanisms in several areas: Transparency: Citizens can document what information is presented to them and whether explanations or disclosures are actually accessible in practice. Accountability: Repeated and properly documented observations can provide useful material for further institutional examination. AI Literacy: The method encourages people to distinguish evidence, assumptions, and interpretation when interacting with AI systems. Independent Oversight: Well-preserved records can potentially assist researchers, civil society organisations, and competent authorities in identifying patterns that deserve further investigation. Public Participation: AI governance becomes stronger when citizens are not only recipients of AI-enabled services but also capable observers of their real-world effects. 5. From Individual Observations to Collective Knowledge The real value of Citizen Auditing does not necessarily lie in a single observation. It lies in the possibility of creating a structured collection of observations. If multiple independent records use a comparable recording standard, researchers and institutions may be better positioned to identify: recurring patterns; systemic issues; changes over time; discrepancies between stated policies and observed outcomes; or areas requiring further technical or regulatory investigation. This does not mean that a collection of citizen observations automatically constitutes proof of a systemic problem. Rather, it creates a more structured starting point for independent verification. 6. A Complement — Not a Replacement Citizen Auditing should not replace regulatory supervision, professional auditing, scientific research, judicial processes, institutional complaint mechanisms, data protection procedures, cybersecurity practices, or other established accountability mechanisms. It is intended as a complementary civic documentation layer. Its strength is simplicity. A person does not need to possess specialised legal or technical expertise to make a careful record of something they have observed. The more important requirement is methodological discipline: Observe first. Record accurately. Verify independently. Notify responsibly. Archive completely. References and Related EU Frameworks Regulation (EU) 2024/1689 — Artificial Intelligence Act European Parliament and Council of the European Union. Official Journal of the European Union, 12 July 2024. https://eur-lex.europa.eu/eli/reg/2024/1689/oj European Commission — Governance and Enforcement of the AI Act European AI Office and national market surveillance authorities are responsible for implementing, supervising, and enforcing the AI Act. https://digital-strategy.ec.europa.eu/en/policies/ai-act-governance-and European Commission — Guidelines on Tr
No comments yet — start the discussion below.