Investigate Using Publicly Available Information (OSINT) Without Harming People

## Write Down Why the Investigation Is Justified and Who It Could Harm Define the question, intended benefit, affected groups, and the decision or public outcome the investigation is meant to inform. Distinguish genuine public or organizational value from curiosity, attention, or the technical ability to obtain information. Document why the work is justified and which rights or interests may conflict. Public availability does not remove privacy, safety, legal, or ethical obligations. Revisit the justification when the scope, likely impact, or evidence changes. Before proceeding, confirm: - Who may benefit, who may be harmed, and how. - Whether the same objective can be met with less intrusion. - Which laws, professional duties, and organizational policies apply. - What event or finding would require the investigation to stop or narrow. ## Plan Which Sources and Methods Answer Each Research Question Create a living plan that connects each research question to sources, collection methods, verification steps, analysis methods, and intended outputs. Define what is in scope, what is excluded, and how those boundaries may affect the conclusions. Record assumptions, knowledge gaps, sampling choices, tool limitations, and foreseeable bias. Set decision points for changing the method rather than allowing convenience or an early theory to determine the evidence gathered. The plan should identify: - Objectives, deliverables, owners, and review points. - Data fields and sources that are necessary for each objective. - Preservation, verification, analysis, and documentation methods. - Quality thresholds, escalation paths, and stopping conditions. ## Weigh Risks to People and Put Together a Team with the Right Skills Assess risks to data subjects, sources, researchers, partners, and audiences before collection begins and whenever conditions change. Consider privacy, physical security, online safety, legal exposure, psychological strain, and the risk of amplifying harmful material. Match the team to the work. Include the technical, linguistic, cultural, legal, and subject-matter knowledge needed to interpret evidence responsibly. Seek perspectives that can reveal blind spots, and do not assign exposure to disturbing content without safeguards and support. For each material risk, record: - Likelihood, severity, affected parties, and uncertainty. - Prevention, mitigation, monitoring, and response measures. - A responsible owner and a review date. - Conditions requiring specialist advice, suspension, or withdrawal. ## Collect Only the Data the Goal Needs and Hide Personal Details Early Collect no more information than the defined objective requires, especially when personal or sensitive data is involved. Prefer less intrusive sources and methods, and reassess necessity before expanding the dataset or linking records across contexts. Follow applicable law and approved access conditions. Protective research identities may reduce risk where permitted, but collection must not rely on misleading people into disclosing information or granting access. Do not bypass controls merely because a technique is available. During collection: - Record source location, capture time, method, and relevant context. - Separate original material from cleaned, transformed, or inferred data. - Correct corrupt or incomplete records without obscuring the change. - Pseudonymize or anonymize personal data as early as practical. ## Keep Evidence Safe and Record Where Each Piece Came From Maintain a traceable path from each claim back to the material used to support it. Preserve originals where lawful and necessary, record transformations, and keep enough context to prevent an isolated item from being misinterpreted. Protect confidentiality, integrity, and availability throughout the evidence lifecycle. Limit access to people with a legitimate need, use secure storage and transfer methods, and define retention and disposal rules that account for both reproducibility and exposure risk. Maintain: - A research log of actions, decisions, dilemmas, and changes in scope. - Provenance records for collected, derived, and corroborating material. - Versioned methods, scripts, filters, and tool settings where relevant. - Backups and archives with tested recovery and controlled access. ## Test Conclusions Against All the Evidence and State How Sure You Are Test explanations against the complete relevant dataset rather than selecting only supportive items. Search for contradictory evidence, plausible alternatives, missing populations, and artifacts introduced by collection or processing. Corroborate consequential findings with independent sources whenever feasible. Evaluate source reliability separately from whether a source supports the working theory. State confidence according to the strength, independence, consistency, and completeness of the evidence. For every major finding: - Identify supporting and undermining evidence. - Explain material gaps, uncertainty, and tool or dataset limitations. - Separate observation, inference, attribution, and prediction. - Seek additional evidence when confidence is below the required threshold. ## Have Others Challenge the Findings Before They Are Published Use a genuine review process that can change or reject a conclusion. Reviewers should inspect evidence, methods, calculations, and reasoning, then look for unsupported leaps, alternative explanations, cultural assumptions, and avoidable harm. Choose reviewers with relevant but varied expertise, and disclose conflicts that could affect independence. Resolve material objections or document why they remain unresolved. Approval should never be reduced to a ceremonial sign-off. The review record should show: - What was checked and by whom. - Which weaknesses or disagreements were found. - How the analysis or publication changed in response. - Which unresolved limitations must accompany the output. ## Publish Findings Clearly While Keeping Private Details and Sources Safe Write for the intended audience in precise, accessible, and non-sensational language. Match the strength of every conclusion to the confidence of the evidence. Provide enough methodology and supporting material for readers to understand how the conclusion was reached. Before release, balance transparency against privacy, security, source protection, and the risk of amplification. Include only the personal details and harmful material necessary to support the legitimate purpose. When evidence cannot safely be shown, explain the basis of the conclusion and the reason for withholding it. The output should: - Distinguish facts, analysis, assumptions, and allegations. - Present relevant counterevidence, data gaps, and cutoff dates. - Credit sources and contributors when doing so is accurate and safe. - Explain material edits to images, data, or other representations. ## Fix Errors After Publication and Turn Lessons into Better Methods Responsibility continues after release. Monitor foreseeable effects, provide an accessible channel for good-faith challenges, and correct material errors promptly and transparently. Use version history so later readers can distinguish the original context from subsequent changes. Review whether the work achieved its intended benefit and caused unexpected harm. Evaluate the method, risk controls, team wellbeing, and publication choices; then convert lessons into specific improvements for future investigations. Where lawful and safe: - Preserve an accurate archive under a defined retention policy. - Share reusable methods, tools, and lessons without exposing protected people. - Report security weaknesses or threats to parties able to address them. - Contribute knowledge and training while respecting credit and consent.