Leading artificial intelligence researcher Michael I. Jordan from UC Berkeley and Inria Paris has published a comprehensive academic paper challenging the current individualistic focus of AI development. According to the research paper submitted on July 8, 2025, Jordan argues that the marketing and technology industries should embrace a “collectivist, economic perspective” that treats social welfare as fundamental rather than an afterthought.

The 11-page research paper, published through arXiv and funded by the European Union’s ERC-2022-SYG-OCEAN program, directly challenges Silicon Valley’s dominant approach to AI development. Jordan contends that current AI systems neglect humans’ fundamentally social nature. “Humans are social animals, and much of our intelligence is social and cultural in origin,” the paper states.

The timing of Jordan’s publication coincides with significant industry developments. Marketing professionals increasingly prioritize artificial intelligence according to recent Mediaocean research showing 75% adoption rates for generative AI in 2024, compared to 55% in 2023. However, Jordan’s framework suggests these implementations may miss crucial social dimensions.

Summary

Who: Michael I. Jordan, UC Berkeley and Inria Paris researcher, along with industry professionals in digital marketing and AI development

What: Publication of academic research proposing collectivist economic approach to AI development that prioritizes social welfare over individual cognitive capabilities

When: Research submitted July 8, 2025, with industry discussion following through mid-July 2025

Where: Academic publication through arXiv platform, with implications for global digital marketing industry particularly affecting Silicon Valley AI development and European data privacy regulation

Why: Current AI development neglects humans’ social nature and treats social consequences as afterthoughts, potentially missing opportunities for sustainable creator economies and effective uncertainty management in marketing systems

Economic mechanisms for marketing systems

Jordan’s research proposes specific applications for digital marketing through multi-way market designs. The paper describes a three-way market for recorded music that connects musicians, listeners, and brands through machine learning recommendations and economic incentives. According to the research, “When a brand needs a song, they are supplied with a song from a particular artist (using an ML model), and the artist is paid in that moment.”

This model contrasts sharply with traditional streaming platforms where “money is made by the platform, via subscriptions or by advertising, but there is no direct connection between producer and consumer.” Jordan notes that United Masters, where he serves as board member, has implemented this approach and signed over 1.5 million musicians while partnering with brands including the NBA, Bose, and State Farm.

The research extends these principles to data markets through layered structures. According to Jordan’s framework, platforms providing services to users could simultaneously sell anonymized data to third-party buyers while maintaining formal privacy guarantees through contractual noise levels. “Although the noise level could be subject to government regulation, let’s instead leave the choice in the hands of the platforms,” the paper suggests.

These concepts address ongoing industry challenges. Consumer trust research published in July 2025 revealed that 59% of European consumers oppose AI training data usage, while 62% feel they have “become the product” in current digital ecosystems.

Technical implementation challenges

Jordan identifies three complementary thinking styles required for effective AI system design: computational, economic, and inferential thinking. The research emphasizes that current machine learning approaches focus primarily on computational aspects while neglecting economic incentives and uncertainty management.

“Real-world intelligence is as much a social, communications, economic, and cultural concept as a cognitive concept,” according to the paper. This perspective challenges current large language model development, which Jordan describes as creating an “illusion of personhood” when these systems actually function as “collectivist artifacts” aggregating human contributions.

The paper details specific technical challenges through database design examples. Jordan describes scenarios where banks must balance privacy guarantees, statistical operations, and third-party data sharing. These problems require “inferential thinking” that considers underlying populations and uncertainty quantification beyond simple computational operations.

For marketing applications, Jordan’s framework addresses strategic behavior among data suppliers. The research examines situations where “suppliers of data are agents who have strategic interests in the outcome of data analysis,” creating potential misalignment between agent goals and analyst objectives.

Market dynamics and social welfare

The research paper presents detailed analysis of how recommendation systems could evolve beyond simple customer-product matching. According to Jordan, current recommendation systems are “limited as microeconomic entities—in particular, no money changes hands.” His proposed alternative creates direct economic connections between content creators and audiences through brand partnerships.

Jordan’s three-way market design addresses what he identifies as fundamental flaws in existing platforms. “There’s a strong incentive for the platform to use generative AI tools to replace the musicians,” the paper notes regarding traditional streaming services. His alternative model creates sustainable revenue streams for creators while maintaining brand connections to audiences.

The framework extends to data privacy considerations through game-theoretic analysis. According to the research, platforms face conflicts between providing privacy guarantees to attract users and maintaining data quality for third-party buyers. Jordan proposes modeling these scenarios as “generalized Stackelberg games” to find optimal equilibria.

Statistical contract theory represents another key component of Jordan’s approach. The research describes how regulatory agencies could design menu-based contracts for AI system approval, similar to FDA drug testing protocols. “The agency would like to control the rate of false positives (an unsafe vehicle goes to market) and the false negative rate obtained (a safe vehicle fails to go to market),” according to the paper.

Industry implications for digital advertising

Jordan’s research arrives as marketing fundamentals continue facing implementation challenges despite AI advancement. Industry consultant Richard Angel recently identified 25% misallocation rates in performance media spending, suggesting systemic issues beyond technological capabilities.

The collectivist framework addresses these challenges through integrated economic and technical design. Jordan argues that “social environments create various kinds of uncertainty, including information asymmetry” while simultaneously enabling “cooperation and the sharing of information, mitigating uncertainty for everyone.”

For programmatic advertising, these principles suggest fundamental restructuring of platform relationships. Rather than treating social consequences as afterthoughts, Jordan’s approach would embed social welfare considerations into algorithmic design from inception. “The path forward is not merely more data and compute,” the research states, “but a thorough blending of economic and social concepts with computational and inferential concepts.”

The paper acknowledges existing multi-agent research in computer science, human-computer interaction, and algorithmic game theory. However, Jordan emphasizes that his focus differs by seeking “quantitative design principles for emerging real-world ML-based systems in which many of the participants are human and many are non-human.”

Academic and educational considerations

Jordan’s research identifies significant gaps in current academic preparation for AI development. The paper notes that while pairwise combinations exist—machine learning blends computation and inference, econometrics combines economics and inference, algorithmic game theory merges computation and economics—the tripartite combination remains underdeveloped.

“Machine learning has made relatively little usage of economics, especially incentive-theoretic ideas and concepts of information asymmetry,” according to the research. This educational gap may explain why industry implementations often overlook social welfare considerations.

The paper proposes viewing academic disciplines through a hub model that connects to social sciences, public policy, cognitive science, biology, medicine, and humanities. Jordan argues this connectivity requires common language beyond current “computational thinking” or “AI as currently practiced.”

For marketing education specifically, Jordan’s framework suggests curriculum development incorporating mechanism design, information theory, and uncertainty quantification alongside traditional campaign management and analytics training.

Future technology development directions

Jordan concludes by comparing AI development to historical engineering disciplines. Chemical engineering and electrical engineering achieved maturity by developing “modular, transparent design concepts” based on solid theoretical foundations like Schrödinger’s equation and Maxwell’s equations.

“For AI, we certainly have exceedingly complex phenomena—cognitive, social, commercial, and scientific—but we do not have the equivalent of Maxwell’s equations as a guide,” the research states. Jordan advocates for incorporating “rationality, experimentation, dialog, openness, cooperation, skepticism, creative freedom, empathy, and humility” as foundational principles.

The research emphasizes that developing collectivist perspectives on information technology “can be just as exciting intellectually as AGI, and at least as promising for the future of the species.” This positioning directly challenges current industry focus on artificial general intelligence development.

For marketing technology development, Jordan’s framework suggests prioritizing system-level designs where “social welfare is a first-class citizen” rather than pursuing individual cognitive capabilities. This approach could address ongoing challenges with AI implementation, consumer trust, and sustainable creator economies.

The paper received attention from industry professionals through LinkedIn discussions, with Meta’s Yann LeCun commenting and INRIA researcher Francis Bach promoting the work as “a must-read rethink” that advocates “blending ML, economics, and uncertainty management to prioritize social welfare over mere prediction.”

Timeline