Attribution Markets Assign Credit to Tasks and Actions

Personal and organizational planning systems often struggle to accurately connect planned tasks with the actual actions performed, leading to misrepresentations of effort and progress. Existing methods typically use an all-or-nothing link, which fails to account for related but unlinked work or falsely reports stalled goals. This paper proposes "Attribution Markets," a novel framework formulated as a quasi-linear Fisher market. In this model, planned tasks act as budget-constrained buyers, and performed actions are divisible goods. A combined text, structural, and temporal signal determines each task's valuation of an action. The market incorporates instruments like a seller reserve price and a buyer cash option, ensuring properties such as conservation of effort, hard budget caps, and a provable junk filter. The research also extends the market with a concave completion utility and addresses noise sensitivity with an entropy-regularized generalization, offering a more nuanced and accurate way to attribute credit for work done.