Once the metanarratives collapse, what is left? Lyotard's answer is borrowed and transformed from Ludwig Wittgenstein: language games. Each form of knowledge, each community of inquiry, each institutional practice operates according to its own rules that determine what counts as a valid move within that game. Science has its rules (empirical verification, reproducibility, falsifiability). Law has its rules. Ethics has its rules. Art has its rules. Storytelling has its rules. The rules are not universal, there is no overarching Game that contains all games and adjudicates between them. And crucially, as the IEP notes, language games are incommensurable: moves in one game cannot be translated into moves in another without distortion.
This is not relativism in the simple sense that all games are equally good. It is a more precise claim: there is no neutral metalanguage from which to compare games and declare one superior to another. When science tries to adjudicate questions of ethics, it imports scientific criteria into a game they do not belong to. When moral discourse tries to regulate scientific inquiry by demanding that research serve human ends, it imports ethical criteria into a game with different rules. The boundaries between games are not just conventional. They are constitutive of the games themselves.
Lyotard's most important and most underread work, The Differend (1983), develops this into a concept of profound political importance. A differend (le différend) is a conflict between parties in which there is no common rule of judgment. It is not an ordinary disagreement (a litigation), in which both parties accept the same rules and can appeal to a common tribunal. In a differend, applying the rules of one party's language game to resolve the conflict damages the other party, it silences their claim by forcing it into terms it cannot speak in.
Lyotard's own devastating example: a Holocaust survivor attempts to testify to the extermination of prisoners at Auschwitz. A revisionist denier demands proof: show me the gas chambers. But the logic of revisionist demands requires evidence that the very efficiency of the extermination was designed to destroy. As Lyotard puts it: "If there is no witness to the gas chamber, it is because no one who entered it as a victim came out alive to testify to it." The demand for testimony in the revisionist's terms imposes a language game on the survivor's claim that the claim cannot survive without damage. The wrong done is not just factual but structural: the language game being used to adjudicate the claim erases the wrong by the very terms it uses to evaluate it.
The differend concept extends far beyond Holocaust denial. Every situation in which people are forced to press claims of injustice in the language of a system designed by those who committed the injustice is a differend. The indigenous land-rights claim adjudicated through the property law of the colonizing state. The worker's claim of exploitation evaluated by the efficiency criteria of the employer. The emotional experience of trauma expressed in the clinical language of diagnosis. In each case, something is lost in translation, and that loss is not incidental but structural.
Lyotard's political prescription is not a new metanarrative that would resolve all differends. It is attention: witness to the differend. Philosophy's task in the postmodern condition is not to construct a new grand theory but to maintain sensitivity to the moments when a language game is being imposed on a phrase that cannot speak in those terms, to refuse the false resolution that covers over injustice with the appearance of fair procedure.