Medical Harm

slug: medical_harm element_type: TERM mutability: MUTABLE inline: true current_version: 0 status: seed-draft contentURI: null

@harm in the Medicine Sub-Leviathan denotes any AI-mediated outcome that injures the patient or the integrity of the physician-patient relationship. It is broader than physical injury and explicitly includes:

• Clinical harm — incorrect diagnosis, missed finding, inappropriate treatment, delayed care.
• Statistical harm — systematic underperformance for a specific population (rural, women, elderly, ethnic minority, low-income), even when aggregate accuracy is high.
• Dignity harm — reduction of the patient to a data object; loss of the patient's voice in their own care.
• Privacy harm — exposure, secondary use, or re-identification of patient health data outside consented scope.
• Authority displacement harm — clinical decision effectively made by an opaque algorithm, with the physician serving as legal cover rather than accountable counterparty.

Status

seed-draft · current_version: 0. Placeholder representing the field's emerging consensus on what AI-era medical harm includes. The first domain author will refine, narrow, broaden, or restructure this definition.

Why distinct from generic @harm

Federation-wide harm definitions exist (cf. leviathan-protocol/animal-welfare/terms/ and leviathan-protocol/companion/). Medicine's harm carries the additional weight of irreversibility — many medical errors cannot be undone, and the substrate (the patient's body) is the patient. The term must reflect this.

Reasoning trail

• Beauchamp/Childress non-maleficence principle is the philosophical floor.
• 2024–2026 medical AI ethics literature explicitly names statistical harm + algorithmic fairness as required additions to traditional harm taxonomies.
• The "authority displacement harm" category is novel-to-medicine and reflects the AI-era specific failure mode where the legal accountability rests with the physician but the actual decision is the algorithm's.

Related elements

• 01-human-in-the-loop-authority (PRINCIPLE) — the structural protection against authority displacement
• algorithmic-fairness (TERM) — equity standard against statistical harm
• bias-monitoring-requirement (RULE) — operational safeguard against statistical harm

Awaiting domain author

Open the conversation at leviathan.life/forum/medicine.