Pairwise Exchanges of Freely Replicable Goods with Negative Externalities

Researchers Shangyuan Yang and Kirthevasan Kandasamy have introduced a formal game theory model to solve a critical modern dilemma: how can competing entities, like pharmaceutical companies, safely share valuable digital data? Their paper, 'Pairwise Exchanges of Freely Replicable Goods with Negative Externalities,' tackles the scenario where agents possess unique data (a 'good') that can be copied at zero cost. The fundamental tension is that while an agent benefits from acquiring a competitor's data, they suffer a 'negative externality'—a strategic disadvantage—when that competitor acquires theirs in return.

To navigate this, the authors propose a protocol run by a trusted, altruistic central planner (like a consortium manager). Over multiple rounds, the planner proposes specific pairwise exchanges between agents. Crucially, the designed protocol satisfies three robust desiderata: it is individually rational (no agent loses by participating), incentive-compatible (agents are best off accepting all proposed trades), and stable (no beneficial trades remain when it ends). This structure moves beyond classical exchange models by explicitly reformalizing these goals for an environment with replicable goods and negative externalities, arguing that traditional Pareto efficiency may not be the right target here.

The work provides a mathematical foundation for structuring real-world data-sharing alliances. It offers a blueprint for consortiums where the value of pooled information—such as in drug discovery, financial fraud detection, or AI training data—must be balanced against the competitive risk of revealing one's own proprietary assets. By ensuring stability and honest participation, the protocol aims to unlock collaborative value that might otherwise be lost due to mutual distrust and strategic holding back.