A paper posted on arXiv, titled "Justice After Identity: Large Language Models and the View from Everywhere," takes up a long-standing problem: the search for a common view of justice and fairness. The abstract notes that human judgments inevitably diverge because they are shaped by self-interest, social position, personal benefit, and cultural inheritance. This makes any single, universal standpoint difficult to achieve.
The paper's title points toward a possible alternative—not a "view from nowhere," but a "view from everywhere." Rather than trying to strip away perspective, large language models might be able to hold and reason across many situated viewpoints at once. The abstract frames this as a response to the challenge that identity and position pose to collective decision-making.
The significance, if the approach holds, is that LLMs could help surface or mediate between conflicting notions of fairness. But the abstract also makes clear that the difficulty is real: the search for a common view has challenged human collective activity all along. Whether LLMs can actually move beyond that challenge remains an open question, and the paper appears to be an early attempt to map the terrain.