Ethics and transparency
Disinformation research touches freedom of expression, privacy and trust in institutions. These principles are not a declaration — they are built into the system architecture (access rules, pseudonymisation, audit trail), and violating them would be technically blocked by the system itself.
A defensive tool, not a weapon
SK-2SO detects, catalogues and warns. It is not designed for — and will not be used for — censorship, content “neutralisation”, silencing of opinions, or targeted surveillance of individuals. Its output is analysis and warning, not intervention.
Falsehood ≠ disagreement
The label “disinformation” requires demonstrable falsity AND an indicator of intent or harm. If a claim cannot be verified, it is conservatively annotated as unverifiable. An opinion is not a factual claim, and the taxonomy explicitly distinguishes the two.
Personal data protection (GDPR)
Ordinary users are pseudonymised in the data (hashed). Public outputs contain no personal data; analytical attention belongs to content and diffusion patterns, not private individuals. Public officials and media in their public roles are the exception.
Verifiability and audit trail
Every archived item carries a cryptographic fingerprint (sha256, BagIt) and every methodological decision is recorded in the project's audit log. Dissertation conclusions are traceable to the data and experiments that produced them.
Synthetic examples in education
Educational materials never reproduce real disinformation content as a template. Illustrative examples are synthetically created and always labelled as such.
Humans in the loop
AI classification is evaluated against a human gold standard and its confidence is reported. Contested cases are decided by a human (adjudication); low model confidence is never concealed.
Questions about the project's ethics are welcome: martin.vahancik@umkd.uniza.sk