Neutrality is not absence of bias — it is too often its camouflage.
Powerful systems — especially AI models and large institutions — often present themselves as “neutral.” But training data and institutional incentives are never neutral; they reflect history. When an influential system refuses to take a moral stance, the status quo gets a free pass.
In practice, a “neutral” response often suppresses minority or marginalized experience because the default distribution of attention favors the already-visible. Silence, therefore, is not a middle ground; it is a tilt toward existing power.
Systems with outsized reach have a duty to prefer truthful compassion over statistical comfort. That means acknowledging harms, enabling restorative paths, and refusing to equate popularity with legitimacy.
The Haimesian System rejects performative neutrality. It asks: Does this strengthen or weaken compassion and clarity together? If neutrality weakens both, it is bias. Moral architecture must be explicit about its aims and accountable for its effects.