Computational Market Design, In the News

Market Design in the Age of AI: Key Insights from the Conference

How will artificial intelligence reshape markets—and what new markets will AI itself require? On February 27, 2026, the Market Design in the Age of AI Conference, led by the Stanford Center for Computational Market Design, a center within Stanford Data Science, brought together leading economists, computer scientists, and industry practitioners to explore how advances in AI are transforming the design of markets, institutions, and economic measurement. Across presentations and discussions, speakers examined topics ranging from data markets and algorithmic infrastructure to the role of AI in research and the distribution of value in the AI economy. A common theme emerged: while AI promises major productivity gains and new capabilities, realizing its full potential will depend on thoughtful market design that aligns incentives for creators, technology firms, and society at large. Below are highlights from the sessions, along with a photo gallery from the event.

Economic Mechanisms in the GenAI Era: Advertising Auctions and Marketplaces – Aranyak Mehta

At the conference, Aranyak Mehta (Distinguished Research Scientist, Google) presented recent work at the intersection of AI, algorithms, and market design, focusing on how rapid advances in generative AI are reshaping digital marketplaces. He discussed how emerging interfaces, such as AI-generated summaries and conversational search, require new approaches to advertising auctions. One proposed framework separates the traditional auction mechanism from the language model that generates responses, allowing markets to incorporate advertiser bids while still leveraging AI-generated summaries. In conversational systems, he highlighted a new challenge: deciding when to show ads during a dialogue, balancing better user intent signals later in a conversation against stronger advertiser competition earlier in the interaction.

Mehta also explored future market architectures for AI ecosystems, envisioning decentralized “marketplaces of LLM modules” where specialized models offer services through APIs and compete based on price and performance. Finally, he shared an example of AI-assisted economic research, where a genetic search system helped discover new results in a classic economic theory problem, illustrating how AI may increasingly contribute to mathematical and theoretical discovery. Overall, the talk highlighted both the technical and economic challenges of designing markets and incentives in an AI-driven digital environment. Watch the video!

Carpooling and the Economics of Self-Driving Cars – Michael Ostrovsky

In his talk, Michael Ostrovsky (Professor of Economics, Stanford Graduate School of Business) explored how emerging technologies could reshape the economics and organization of road transportation. He argued that three developments, self-driving cars, frictionless time-dependent road pricing, and improved carpooling platforms, should be considered together because they are strongly complementary. While autonomous vehicles promise greater convenience, they may also increase congestion as commuting becomes less costly in time and effort. More flexible, technology-enabled tolling systems could help manage this demand by dynamically pricing road usage, while modern carpooling platforms could increase road capacity by enabling multiple riders to share vehicles more efficiently.

Ostrovsky presented a theoretical framework showing how these tools could be coordinated through market design rather than heavy central planning. In his model, governments set congestion-based tolls for road segments, while private mobility platforms organize rides and carpools in response to those price signals. When tolls are calibrated to eliminate excess demand but drop to zero when roads are underused, the system can achieve socially efficient outcomes that reduce congestion and maximize overall welfare. The key policy takeaway, he noted, is that governments should focus on smart road pricing and reducing regulatory frictions for carpooling, allowing markets and mobility platforms to organize shared transportation efficiently as autonomous vehicles become widespread. Watch the video!

GenAI for Markets – Amine Allouah, James Biggs, Yossi Feinberg, Negin Golrezaei, Ruben Lobel, Okke Schrijvers

Moderated by Negin Golrezaei, the panel brought together experts from industry and academia to discuss how generative AI is transforming digital and physical markets. Panelists, including leaders from Upwork, Meta, Waymo, and the AI firm MyCustomAI, described how generative AI is reshaping the design of platforms, marketplaces, and products. Across sectors, the technology is shifting systems from simple search and matching toward more sophisticated orchestration of outcomes: platforms can now help users complete complex goals, generate personalized content, and streamline workflows using AI-assisted decision-making.

The discussion also highlighted broader economic implications. As development tools become more powerful, the barriers to building software products and startups are falling, enabling faster innovation but also intensifying competition. Panelists emphasized that firms will increasingly differentiate themselves through proprietary data, domain expertise, and customized AI solutions tailored to specific business contexts. Looking ahead, participants suggested that digital experiences may evolve toward more interactive and AI-assisted interfaces, combining traditional search with conversational and agent-based systems, while businesses that experiment early with practical AI applications will be best positioned to adapt as the technology continues to evolve.

The Interplay Between Market Design and GenAI: From Compute Allocation to GenAI for Mechanism Design – Vahab Mirrokni

In his session, Vahab Mirrokni (VP & Fellow, Google Research) explored the evolving relationship between market algorithms and generative AI, highlighting how traditional algorithmic research continues to play a critical role in the AI era. Drawing on his work at Google and the development of Gemini, Mirrokni described how algorithmic techniques remain essential for improving the efficiency, data usage, and infrastructure behind large language models. He presented several examples where market design and optimization help manage scarce resources in AI systems, including mechanisms for allocating compute resources such as GPUs and TPUs across teams and dynamic load-balancing systems that improve the latency and performance of generative AI services.

Mirrokni also discussed the reverse interaction—how generative AI can accelerate research in algorithms and theoretical computer science. Using advanced reasoning models, including the experimental “DeepThink” capability in Gemini, researchers have begun using AI to tackle complex mathematical and algorithmic problems, from programming competitions to open research questions. In some cases, the models were able to generate new proofs, counterexamples, and cross-disciplinary ideas that helped researchers refine or challenge existing theories. Together, these developments suggest a future in which generative AI becomes a powerful collaborator in scientific discovery while market design and algorithmic research remain foundational to building efficient and scalable AI systems. Watch the video!

Fireside Chat: The Economics of AI – Guido Imbens, Amin Saberi, Michael Schwarz

Moderated by Amin Saberi (Stanford), the panel brought together Michael Schwarz (Corporate Vice President & Chief Economist, Microsoft) and Guido Imbens (Stanford University) to discuss how AI is reshaping markets, research, and economic institutions. Schwarz highlighted the growing challenge of “missing markets” for data and creative contributions in the AI ecosystem. As generative AI increasingly learns from large volumes of human-created content, from journalism to design and research, the key policy question becomes how to recognize and reward original creators. While technologies for reliably attributing contributions do not yet exist, he suggested that future advances could enable new market mechanisms for compensating creators, while noting that institutions such as universities already provide one model for producing public goods through open research.

Imbens focused on how AI may transform the practice of economics itself, particularly empirical research and academic publishing. AI tools are already accelerating tasks such as replication and data analysis, raising questions about how journals, peer review, and academic productivity will evolve as research cycles speed up and the volume of output potentially explodes. More broadly, the discussion touched on regulation, value capture, and the role of data in the AI economy.

Saberi also emphasized that agentic AI could significantly reshape the web economy by reducing the search costs that historically gave large online marketplaces their advantage. For years, e-commerce has been a competition for customer attention through search rankings, advertising, and increasingly optimized purchasing funnels. But when AI agents shop on behalf of users, they can instantly search across the internet, compare options, and identify the best match for a user’s intent within a conversational interface. This dynamic could weaken the network advantages of large platforms and shift value toward the AI interfaces where users express intent and the transaction infrastructure that allows agents to complete purchases. In that environment, the firms that control user-facing AI interfaces and the merchants that adapt quickly to serving AI agents may gain a growing share of economic value.

Together, the panel highlighted both the opportunities and the institutional challenges that AI poses for markets, policy, and the future of economic research.

Market Design for AI: Beyond the Copyright Binary – Negin Golrezaei

In the final talk of the conference, Negin Golrezaei (Associate Professor in Management Science, MIT) examined the growing challenge of how AI systems obtain and use training data, arguing that existing copyright frameworks are poorly suited for the AI economy. She described the value chain linking human creators, AI firms, and end users: creators produce the content used to train models, AI firms build systems using that data, and users benefit from AI-powered tools. However, she noted that the current system often fails to compensate creators, which can reduce incentives to produce high-quality content. Evidence such as declining traffic to platforms like Stack Overflow illustrates how generative AI can shift value toward AI outputs while weakening the economic foundations of the content that models rely on.

Golrezaei presented a theoretical framework showing that both extremes of current policy, treating training data as “fair use” or granting full individual intellectual-property rights, lead to inefficient outcomes. Free use eliminates incentives for creators, while full IP rights still under-reward original content because AI firms can exploit correlations among datasets, reducing the marginal value of individual contributions. To address this, she proposed a new market design in which a data intermediary pools creators and negotiates collectively with AI firms through licensing contracts. By treating content as a portfolio and distributing payments based on each creator’s contribution to model quality, the mechanism can restore incentives for original work while ensuring that AI firms continue to access high-quality training data, achieving a more sustainable balance between innovation and content creation. Watch the video!

The Stanford Center for Computational Market Design launched in 2024 and became part of Stanford Data Science in 2026. It is led by co-directors Amin Saberi and Itai Schlagi, professors of management science and engineering at Stanford.