Shared (Mis)Understandings and the Governance of AI: A Thematic Analysis of the 2023-2024 Oversight of AI Hearings
summary
The gist
Shared (Mis)Understandings and the Governance of AI: A Thematic Analysis of the 2023-2024 Oversight of AI Hearings investigates early legislative deliberations over Artificial Intelligence (AI) in…
In short
This study analyzed Senate AI oversight hearings to see how participants created narratives about Artificial Intelligence. It found that shared, often misleading, understandings of AI—like its potential benefits or inevitability—were used as substitutes for real evidence. These narratives helped establish power arrangements and justify specific types of AI governance.
Key concepts
- Deconstructing the Term AI
- The term 'AI' is often a marketing tool. It makes complex automation sound magical or powerful, which helps companies avoid accountability by portraying machines as thinking entities instead of tools made by people.
- The Three Dominant Understandings
- Participants frequently used three main ideas about AI: 1) Tremendous Potential Benefits (unspecified future good), 2) Inevitability (AI development is unstoppable), and 3) AI Nationalism (American companies are doing the necessary work).
- How Power is Enacted Through Discourse
- These shared understandings act as a stand-in for actual proof. They obscure real harms, such as increased energy use or discrimination, while framing AI benefits as a universal good that protects American innovation and business.
Terminology used across episodes
This episode discusses
- Shared (Mis)Understandings and the Governance of AI: A Thematic Analysis of the 2023-2024 Oversight of AI Hearings · Paper Radio
- Mind the Gap: Foundation Models and the Covert Proliferation of Military Intelligence, Surveillance, and Targeting
- Enhancing LLM Factual Accuracy with RAG to Counter Hallucinations: A Case Study on Domain-Specific Queries in Private Knowledge-Bases
- Estimating the Carbon Footprint of BLOOM, a 176B Parameter Language Model
The paper
Shared (Mis)Understandings and the Governance of AI: A Thematic Analysis of the 2023-2024 Oversight of AI Hearings · Read on arXiv
Rachel Leach
Senate Judiciary Committee
This paper investigates early legislative deliberations over Artificial Intelligence in the United States through a thematic analysis of the 2023 - 2024 Oversight of AI hearings held by the Senate Judiciary Committee's subcommittee on Privacy, Technology, and the Law. I focus on these hearings as a site where participants draw from, and renegotiate, accustomed ways of thinking about technology and society. First, I examine how participants, who overwhelmingly represent the technology industry, work to create narratives for understanding the future impacts of AI, and the relationship between technology and progress. Second, I examine how these narratives are invoked to argue for a particular form of AI governance, while casting alternative approaches as everything from infeasible to anti-American. By tracing industry influence over dominant understandings of the impacts of AI and the proper role of government, I examine the arrangements of power enacted and upheld through these hearings. In all, I ask: what role do shared (mis)understandings of AI play in early deliberations over governing this technology?
Transcript
Introduction to the show: ident: Computer Science. The day in computer science: software, systems and theory on arXiv.
Tom: We're looking at the paper: "Shared (Mis)Understandings and the Governance of AI".
Jane: Here's the gist of the paper. This paper investigates early legislative deliberations over Artificial Intelligence (AI) in the United States through a thematic analysis of the 2023-2024 Oversight of AI hearings held by the Senate Judiciary Committee’s subcommittee on Privacy, Technology, and the Law.
Tom: It examines how participants create narratives for understanding AI impacts to argue for particular forms of AI governance, revealing arrangements of power enacted through these hearings Deconstructing the Term AI The term itself functions to promote a particular understanding of these technologies and their capabilities Bender (2023) explains that “In fact this
AI: is a marketing term" It’s a way to make certain kinds of automation sound sophisticated, powerful, or magical and as such it’s a way to dodge accountability by making the machines sound like autonomous thinking entities rather than tools that are created and used by people and companies The Three Dominant Understandings The thematic analysis identified three understandings of AI frequently invoked by industry representatives and taken up by Senators 1.
Jane: AI has Tremendous Potential Benefits, which focuses on broad and unspecified claims to future possible capabilities of AI across sectors like health and climate change 2.
Tom: AI is Inevitable, where participants suggest that the development of AI is unavoidable and limits legislation 3.
Paper summary: Tom: So, we've got the overview of "Shared (Mis)Understandings and the Governance of AI: A Thematic Analysis of the two thousand twenty-three-two thousand twenty-four Oversight of AI Hearings <ref:2603.03193#pg1,Shared (Mis)Understandings and the Governance of AI: A Thematic Analysis of>." The paper argues that participants, mostly from industry, are using narratives to frame what AI is capable of and how it should be controlled.
Jane: It claims that these shared understandings function as a substitute for real evidence, which in turn hides existing harms and promotes speculative future capabilities instead. It shows how the discourse used in those hearings actually upholds power arrangements favoring certain ideas about AI governance.
Lu: Specifically, the paper identifies three dominant ways people talk about AI: first, that it has tremendous potential benefits across sectors like health and climate change; second, that its development is inevitable; and third, a focus on American companies doing what's necessary to protect national values.
Meng: Those understandings are really just proxies for evidence because they let people avoid talking about real problems like increased energy use or discrimination that already exist. When you talk about AI nationalism, it gets tied up with American ideas of innovation and business, which shields certain groups from scrutiny.
Lalam: It points out those narratives are full of contradictions. For example, people claim AI will cure cancer and help stop climate change but then suggest the technology is so fragile that we need really limited and permissive rules. This reliance on shared understandings instead of contestable evidence is a key finding > Shared (Mis)Understandings and the Governance of AI: A Thematic Analysis of the two thousand twenty-three-two thousand twenty-four Oversight of AI Hearings <ref:2603.03193#pg1,Shared (Mis)Understandings and the Governance of AI: A Thematic Analysis of>
Conclusion: Tom: So, to wrap up this paper, it shows that these shared understandings function as a way to substitute for actual evidence, which lets people focus on future possibilities instead of current material harms. The narratives used in those hearings are what support the power structures around AI governance.
Jane: It really highlights how early attempts to govern this technology are being shaped by these specific shared ideas rather than a full picture of the situation on the ground. It’s about recognizing that those understandings aren't neutral; they actively shape who gets power in these debates.
Lu: The core implication is that we need to look closely at what those dominant discourses are and how they are being used to justify specific rules for AI development and deployment. It forces a re-evaluation of the evidence being used in these legislative settings > Shared (Mis)Understandings and the Governance of AI: A Thematic Analysis of the two thousand twenty-three-two thousand twenty-four Oversight of AI Hearings <ref:2603.03193#pg1,Shared (Mis)Understandings and the Governance of AI: A Thematic Analysis of>
Meng: For those who are actually trying to deploy models, this means that the way we talk about what AI can do—the hype versus the reality—is heavily influenced by these pre-existing narratives. You have to be aware of that framing when you're building something at scale.
Lalam: It shows how these early deliberations are less about technical solutions and more about establishing a specific language for what AI governance means in the first place > Shared (Mis)Understandings and the Governance of AI: A Thematic Analysis of the two thousand twenty-three-two thousand twenty-four Oversight of AI Hearings <ref:2603.03193#pg1,Shared (Mis)Understandings and the Governance of AI: A Thematic Analysis of>
More episodes
- 2411.16152-Optimizing Winograd Convolution on ARMv8 processors
- 2505.16254-Human-AI Writing as Distributed Cognition: Reassessing Collaborative Writing Theories in the Age of LLMs
- 2506.10913-Choreographic Quick Changes: First-Class Location (Set) Polymorphism
- 2507.14812-A Black-Box Approach for Exogenous Replenishment in Online Resource Allocation
- 2507.15448-Frames over finite fields with respect to the Galois inner product and their connections to self-orthogonal and LCD codes
- 2508.21572-NewsReX: An Open-Source Multi-Framework for Neural News Recommendation
- 2509.17028-Impact of Packetization on Network Calculus Analysis
- 2510.17342-AoA Services in 5G Networks: A Framework for Real-World Implementation and Systematic Testing
- 2511.04088-Efficient and rate-optimal list-decoding in the presence of minimal feedback
- 2511.17264-Notes on Stack Machines and Quantum Stack Machines