Abstract: With the continuous advancement of the wave of artificial intelligence, algorithmic empowerment has gradually become a key engine for the social assistance system to move towards a high-quality development stage. The research is devoted to grasping the underlying operation logic of algorithm empowering social assistance decision-making, deeply excavating the alienation risk that may be induced, and then exploring targeted optimization strategies. It is found that the algorithm relies on the internal logic of the chain of ' perceived demand-allocation of resources-collaborative governance ', realizes the explicitization of implicit demand, the precision of resource allocation and the coordination of governance actions, and promotes the paradigm transformation of social assistance from ' passive relief after the event ' to ' active monitoring before the event '. However, when digital intelligence technology is deeply embedded in social assistance decision-making, there are also complex risks and hidden dangers : first, the ' algorithm black box ' at the technical level obscures procedural justice, making the decision-making process difficult to be reasonably explained and easy to solidify prejudice ; the second is that the fuzzy boundary of power and responsibility at the institutional level leads to a vacuum in the regulatory field, resulting in the dilution of responsibilities; the third is that the excessive expansion of instrumental rationality at the value level dispels the dominant position of the recipients, and even leads to a new type of digital exclusion. In response to these pain points, this study attempts to build a ' technology-system-value ' trinity governance logic : reshape technology trust by constructing an interpretable and auditable algorithm mechanism, anchor human-machine coordination rights and responsibilities based on the whole chain accountability system, and promote the return of technical rationality to humanistic value, so as to build a new pattern of modern social assistance with both wisdom and humanity.