Machine Learning (ML) has achieved remarkable milestones in recent years, but its future hinges on a robust, ethically grounded, and well-incentivized data infrastructure. The next era of AI innovation requires specialized data that is scarce, proprietary, or sensitive (e.g., medical records, autonomous vehicle data, pharmaceutical test results). Collecting such data is cost and labor-intensive and subject to strict privacy and legal constraints. Simultaneously, the era of “free” or lightly regulated data (e.g., scraped from the public internet) is drawing to a close as copyright and user consent concerns rise — evidenced by legal challenges to large generative models and platform restrictions on web scraping. Taken together, these developments underscore the need for a new AI paradigm — one in which data rights are respected, data owners receive fair compensation, and valuable datasets are shared and reused to unlock novel applications. This raises the central question we want to address in our workshop

"How can we design incentive mechanisms to promote data sharing while safeguarding privacy, fairness, and intellectual property?"

To make progress towards the above question, our workshop would focus on themes such as:

• Data valuation and attribution: Methods to determine data's value, detect bad/fake data, credit contributors, and establish fair compensation.

• Mechanism design for data and model markets: theoretic approaches to foster honest participation, data quality, and fair profit-sharing.

• Data privacy and autonomy: designing federated/distributed learning mechanisms and effectively utilizing personalization and differential privacy tools that respect data ownership and autonomy. 

• Collaboration among competitors: managing conflicts of interest between participants when they are competing for finite resources (e.g., firms compete for market share, autonomous vehicles compete for road usage)

• Fairness and efficiency: study the potential definitions of fairness and the tension between designing fair systems and socially optimal efficiency.  

• Legal perspectives: Address data/model ownership, data market regulations, and alternatives such as data co-ops.

• Practical algorithms: Develop implementations that can sustain multiple rounds of collaboration.

• Empirical studies: Simulate multi-agent environments to demonstrate the need for or value of incentives.

CONFIRMED SPEAKERS LIST