The most common misreading of Lyotard is to conclude that the end of metanarratives means that anything goes, that all claims are equally valid, that standards of truth and justice are merely local and conventional. This would be simple relativism, and Lyotard explicitly rejects it. But his alternative is subtle, and it is worth working through carefully.
Lyotard's positive proposal in The Postmodern Condition is paralogy (from the Greek para, beside, beyond, and logos). In contrast to homology (the goal of consensus, of bringing all participants in a discourse to agreement through the power of the better argument. Is, roughly, Habermas's position), paralogy is the production of new moves, new rules, new language games that were not possible within the existing framework. Paralogy does not aim at consensus. It aims at what Lyotard calls dissensus: the moment when someone makes a move that changes the game itself, when a new idiom is created that could not have been anticipated from within the old game.
The history of genuine scientific innovation is full of paralogical moves: Copernicus did not argue within the Ptolemaic game more successfully than Ptolemy. He changed the game. Einstein did not refine Newtonian mechanics. He showed that the framework of absolute space and time was not a given but a particular model with a domain of validity, and then provided a different framework. The paralogical move is the one that reveals the contingency of the current rules and opens the possibility of playing differently.
Lyotard is explicitly critical of Habermas at this point. Habermas's communicative ethics argues that the goal of discourse is consensus reached through the force of the better argument in an ideal speech situation. Lyotard's objection: consensus is not the goal of knowledge but its tombstone. When everyone agrees, the game has stopped producing new knowledge. The productive moments are the moments of instability, of disagreement, of the odd move that does not fit the current rules. Postmodern science, he cites Gödel's incompleteness theorems, quantum indeterminacy, catastrophe theory, is precisely the science of the limits of its own rules, the science that produces its own instability as a methodological result.
A clarification: Lyotard is not saying that any destabilizing move is good. He is not celebrating trolling, noise, or pure disruption. His criterion is the differend: the paralogical move is politically and epistemically valuable when it gives voice to something that could not speak within the existing rules, when it creates a new idiom for a claim that was previously a differend. The enemy is the false consensus that covers over genuine difference and genuine wrong under the appearance of neutral procedure.