Virtual Pages - Source Excerpt 85
Summary
This source excerpt preserves a bounded section of 2IA.org/wp-content/themes/twoia-intelligence/inc/virtual-pages.php so readers can inspect the evidence without opening the full source file.
**Source path:** 2IA.org/wp-content/themes/twoia-intelligence/inc/virtual-pages.php
'title' => __( 'AI Risk Scores And Due Process', 'two-identities-anonymous' ),
'overview' => __( 'AI risk scores turn data into rankings, labels, alerts, recommendations, or decisions that can affect real people.', 'two-identities-anonymous' ),
'why' => __( 'It matters because automated suspicion can shape access, investigation, services, employment, moderation, travel, or reputation.', 'two-identities-anonymous' ),
'works' => __( 'Models ingest training data, features, prompts, rules, thresholds, and feedback loops, then produce scores or classifications for human or automated use.', 'two-identities-anonymous' ),
'wrong' => __( 'Failure modes include bias, hallucination, untested proxies, weak validation, rubber-stamp human review, and no appeal route.', 'two-identities-anonymous' ),
'benefits' => __( 'Institutions benefit from speed, scale, budget justification, and a veneer of objectivity; vendors benefit from proprietary opacity.', 'two-identities-anonymous' ),
'harmed' => __( 'People harmed include those denied services, flagged for review, targeted by enforcement, moderated, scored, or burdened without a usable explanation.', 'two-identities-anonymous' ),
'terms' => __( 'Key terms: risk score, feature, proxy, threshold, validation, bias assessment, human review, appeal, explainability, impact assessment.', 'two-identities-anonymous' ),
'records' => __( 'Documents to request: model cards, validation studies, bias assessments, impact reviews, appeal logs, procurement files, and human-review procedures.', 'two-identities-anonymous' ),
'questions' => __( 'Questions for officials: what data trained the system, what error rates are known, and who can override the score?', 'two-identities-anonymous' ),
'actors' => __( 'Actors to map: deploying agency, vendor, model provider, data provider, human reviewers, appeals office, auditors, and affected public.', 'two-identities-anonymous' ),
'references' => __( 'References to verify: NIST AI Risk Management Framework, agency AI inventories, procurement records, and civil-rights guidance.', 'two-identities-anonymous' ),
'related' => __( 'Related 2IA work: /ai-surveillance/, /false-positives/, /methodology/, /public-records-and-foia/.', 'two-identities-anonymous' ),
'steps' => __( 'Request validation records, ask for appeal policies, and document whether human review has real authority.', 'two-identities-anonymous' ),
),
array(
'title' => __( 'Data Brokers And Shadow Public Power', 'two-identities-anonymous' ),
'overview' => __( 'Data brokers collect, infer, append, and sell profiles that can become public-power intelligence through purchase or integration.', 'two-identities-anonymous' ),
'why' => __( 'It matters because public agencies can gain access to intimate commercial profiles without ordinary collection safeguards.', 'two-identities-anonymous' ),
'works' => __( 'Brokers gather public records, app data, purchases, demographics, location signals, device IDs, interests, and modeled inferences.', 'two-identities-anonymous' ),
'wrong' => __( 'Failure modes include invisible collection, sensitive inference, stale data, weak opt-out, inaccurate profiles, and unclear downstream use.', 'two-identities-anonymous' ),
'benefits' => __( 'Brokers monetize asymmetry; agencies and private institutions gain information without building public collection systems themselves.', 'two-identities-anonymous' ),
'harmed' => __( 'People harmed include people with sensitive movement, health, financial, political, immigration, or association patterns exposed through commercial data.', 'two-identities-anonymous' ),
'terms' => __( 'Key terms: data broker, append, inferred segment, device graph, mobile advertising ID, opt-out, downstream recipient, permissible use.', 'two-identities-anonymous' ),
'records' => __( 'Documents to request: vendor lists, invoices, data dictionaries, broker contracts, use policies, opt-out terms, and retention rules.', 'two-identities-anonymous' ),
'questions' => __( 'Questions for officials: what broker data is purchased, what legal authority applies, and what correction route exists?', 'two-identities-anonymous' ),
'actors' => __( 'Actors to map: brokers, resellers, analytics platforms, public agencies, procurement offices, law-enforcement units, and consumer-protection regulators.', 'two-identities-anonymous' ),
'references' => __( 'References to verify: FTC data broker reports, public procurement records, privacy policies, and consumer-protection enforcement actions.', 'two-identities-anonymous' ),
'related' => __( 'Related 2IA work: /surveillance-systems/data-brokers-are-shadow-infrastructure/, /metadata-is-identity/, /public-records-and-foia/.', 'two-identities-anonymous' ),
'steps' => __( 'Request invoices and contracts, compare vendor claims with policy, and ask how broker data is validated or deleted.', 'two-identities-anonymous' ),
),
array(
'title' => __( 'Keyword Monitoring And False Positives', 'two-identities-anonymous' ),
'overview' => __( 'Keyword monitoring watches words, phrases, labels, or signals, then routes matches into queues, alerts, reports, or moderation decisions.', 'two-identities-anonymous' ),
'why' => __( 'It matters because language is contextual. A word can be joke, quote, research, protest, art, threat, grief, slang, or translation noise.', 'two-identities-anonymous' ),
'works' => __( 'Systems define terms, match text or metadata, score context, apply thresholds, and send hits to reviewers or automated rules.', 'two-identities-anonymous' ),
'wrong' => __( 'False positives emerge from ambiguity, sarcasm, reclaimed language, multilingual context, OCR errors, stale lists, and reviewers who never see the surrounding record.', 'two-identities-anonymous' ),
'benefits' => __( 'Institutions benefit from cheap triage and broad monitoring claims; vendors benefit from dashboards that make simple matching look like intelligence.', 'two-identities-anonymous' ),
'harmed' => __( 'People harmed include students, protesters, workers, journalists, artists, marginalized communities, and anyone whose context is flattened by a list.', 'two-identities-anonymous' ),
'terms' => __( 'Key terms: keyword list, watch term, threshold, context window, reviewer queue, precision, recall, false positive, audit sample.', 'two-identities-anonymous' ),
'records' => __( 'Documents to request: policy, training, term governance, review procedures, audit samples, appeal logs, accuracy reports, retention schedules, and vendor manuals.', 'two-identities-anonymous' ),
'questions' => __( 'Questions for officials: who chooses terms, how are they reviewed, what context is preserved, how many hits are wrong, and how are people notified or corrected?', 'two-identities-anonymous' ),
'actors' => __( 'Actors to map: monitoring vendor, program office, reviewers, policy approvers, legal counsel, school or workplace administrators, and appeal contacts.', 'two-identities-anonymous' ),
'references' => __( 'References to verify: platform transparency reports, school or workplace monitoring policies, civil-liberties reports, and public-records responses.', 'two-identities-anonymous' ),
'related' => __( 'Related 2IA work: /keyword-monitoring/, /false-positives/, /ai-surveillance/, /methodology/.', 'two-identities-anonymous' ),
'steps' => __( 'Request term-governance records, ask for error rates, inspect appeal routes, and document context loss without republishing private details.', 'two-identities-anonymous' ),
),
array(
'title' => __( 'Protest Privacy And Civil Liberties', 'two-identities-anonymous' ),
'overview' => __( 'Protest privacy covers how lawful assembly, association, speech, location, images, devices, and records can be collected or chilled.', 'two-identities-anonymous' ),
'why' => __( 'It matters because the right to assemble weakens when attendance becomes an intelligence product or a future risk marker.', 'two-identities-anonymous' ),
'works' => __( 'Records can come from cameras, drones, social media, location data, permit files, mutual-aid lists, arrest databases, vendor tools, and interagency sharing.', 'two-identities-anonymous' ),
'wrong' => __( 'Failure modes include guilt by association, overbroad retention, misidentification, chilling effects, private-person exposure, and protest records repurposed for unrelated decisions.', 'two-identities-anonymous' ),
'benefits' => __( 'Agencies and vendors benefit when protest activity is treated as a searchable event stream instead of protected civic participation.', 'two-identities-anonymous' ),