Agentic AI: Autonomy, Accountability, and the Algorithmic Society : A Review
Artificial Intelligence & Machine Learning News
“Agentic AI: Autonomy, Accountability, and the Algorithmic Society” by Anirban Mukherjee and Hannah Chang
This timely and thought-provoking paper offers one of the more comprehensive early frameworks for evaluating the risks and possibilities of “agentic AI”—a category of artificial intelligence that moves beyond reactive prompts and into proactive, autonomous decision-making.
Authored by Anirban Mukherjee and Professor Hannah Chang of Singapore Management University – Lee Kong Chian School of Business; moreover, a cross-disciplinary investigation of how AI agents that operate with long-term goals and self-directed workflows challenge current legal, creative, and economic structures.
The paper opens with a compelling contrast between traditional generative AI and its more autonomous successor. Where ChatGPT-like systems respond to inputs, agentic AIs initiate and manage complex processes, potentially without ongoing human involvement. This evolution prompts three central areas of concern: authorship and intellectual property, legal accountability, and competitive market behavior.
One of the paper’s most original contributions is its discussion of authorship and creative ownership in a world where AI can independently originate works of artistic, literary, or functional value. Mukherjee and Chang highlight the collapse of existing incentives for human creativity when machines can generate novel content at scale. They propose that IP laws must evolve to either recognize the AI as a co-author or attribute ownership through entirely new frameworks.
The legal implications are even more pressing. The “moral crumple zone” concept—borrowed from sociotechnical systems research—is a standout metaphor describing how accountability may deflect onto end-users and developers, even when neither has full control over an autonomous agent’s decisions. This critique is particularly urgent as we begin to rely on AIs in contexts like medical diagnostics, financial trading, or autonomous vehicles, where risk is real and potentially catastrophic.
The paper’s treatment of competition in two-sided algorithmic markets is equally sharp. By examining how AI on both sides of a market (buyers and sellers) can independently engage in repeated interactions that may simulate collusion, Mukherjee and Chang raise valid concerns about the effectiveness of current antitrust laws. In scenarios where algorithms learn to converge on price or resource-sharing strategies, traditional definitions of collusion—requiring explicit agreement—fall apart.
Finally, Mukherjee and Chang speculate on a near-future society where networks of agentic AI systems develop their own norms—akin to emergent behavior in biological systems. While this raises hope for stability and efficiency, it also introduces the risk of hard-to-regulate oligopolies, echo chambers, or entrenched biases.
The authors conclude with a call for interdisciplinary collaboration. This is a strength of the paper: it balances technical insight with policy relevance and offers actionable themes for legal scholars, marketers, technologists, and regulators alike.
“Agentic AI: Autonomy, Accountability, and the Algorithmic Society” is a forward-looking, well-structured, and insightful exploration of the challenges posed by the next generation of autonomous systems. Highly useful for those shaping AI policy, designing regulatory frameworks, or researching AI ethics. By blending law, economics, and technology into a unified analysis, Mukherjee and Chang provide a roadmap for how society might adapt to a future increasingly governed not by humans, but by agents.
read the full paper: Agentic AI: Autonomy, Accountability, and the Algorithmic Society by Anirban Mukherjee, Hannah Chang :: SSRN
