Marginal charity is the idea that individuals can have the most social gain by unit of private loss by shifting their choices marginally in a prosocial direction. A person's choices are by default close to the private optimum, which typically diverges significantly from the social optimum. Thus, slight deviations away from the latter and toward the former should achieve outsized social gains. The expression "marginal charity" was introduced by Robin Hanson (Hanson 2012), though as the author notes, the idea is a relatively straightforward implication of optimization theory.

Possible examples of marginal charity include divesting from the most harmful companies or industries, reducing consumption of animal products, and being generally nicer to others (Hanson 2018; Trammell 2019).

Kevin Simler and Robin Hanson speculate that marginal charity, despite its efficiency, is not very popular because acts of marginal charity tend to be indistinguishable from ordinary self-interested behavior. As a consequence, such acts are ill-suited to play the role of moral signaling, which requires behavior to be visibly costly to the agent (Hanson 2014; Simler & Hanson 2017: 222-223).

Marginal charity need not involve changes in a person's own behavior—it can apply to cases where others are paid to be more marginally prosocial. For example, longtermists can pay self-interested or short-termist demographers to slightly expand the time horizon of their projections (Trammell 2019).