Source-linked AI summary
Free Speech and Artificial Intelligence
Etienne Brown
TL;DR
The paper examines how social media recommendation algorithms and conversational AI raise questions about free speech, including speech visibility and the rights of AI-related actors. It analyzes these technologies philosophically and legally, arguing that recommendation algorithms shape digital speech visibility while distinguishing several possible rights-holders. The paper also identifies limits to current claims, including the difficulty of establishing a causal link between recommendation and polarization.
Problem
The paper addresses how AI technologies, especially recommendation algorithms and conversational AI, affect free-speech rights and speech visibility.
Method
The paper compares two AI technologies and distinguishes rights claims involving AI agents, users, and developers across different conversational and publishing scenarios.
Results
Recommendation algorithms shape speech visibility and can sharply concentrate attention, while the analysis identifies multiple potential rights-holders in conversational AI.
Takeaways & Limitations
Algorithmic recommendation is a free-speech issue because it affects whether digital speech is visible, while chatbot analysis requires separating the rights of agents, users, and developers.
Takeaways & Limitations
The causal link between algorithmic recommendation and polarization is hard to establish.
Abstract
from arXiv · showhide
Philosophers and legal scholars are engaged in debates about the implications of artificial intelligence for freedom of expression. This paper analyzes the free speech issues raised by two distinct AI technologies: social media recommendation algorithms and conversational AI (i.e., chatbots powered by large language models). The first part shows that, through their recommendation algorithms, social media platforms control the dynamics of speech visibility in the digital public sphere, making algorithmic recommendation relevant to the philosophy of free speech. The second part turns to conversational AI. It discusses both the reasons for granting or withholding speech rights to artificial agents and users' right to receive information, which may render specific forms of chatbot regulation illegitimate. Throughout, the chapter also considers whether social media platforms or AI developers hold corporate speech rights. Its general aim is to raise rather than settle questions that arise from the rapid development of AI technologies.
1 Introduction
The chapter examines free-speech questions raised by social media recommendation algorithms and conversational AI. It argues that recommendation algorithms shape speech visibility and distinguishes rights held by AI agents, users, and developers.
- The discussion extends more than a decade of scholarship on whether machine speech is legally protected speech.
- The chapter examines social media recommendation algorithms and LLM-powered chatbots as two influential AI technologies affecting free speech.
- Recommendation algorithms are relevant to free-speech philosophy because they drive speech visibility and invisibility in the digital public sphere.
- The chapter asks whether machine speech is legally protected and identifies AI agents, users, and developers as potentially relevant rights-holders.
- It distinguishes human dialogue with an LLM from publishing AI-generated texts, images, or videos online.
2 Free Speech and Social Media Recommendation Algorithms
Social media recommendation algorithms shape what users encounter and how widely speech is heard, making algorithmic reach a free-speech concern. The chapter argues that severe demotion can function like removal while raising unresolved questions about platforms’ expressive rights and democratic regulation.
- 2.1 Algorithmic Reach and the Value of Speech: Recommendation algorithms determine what users see first and next, exposing them to content without requiring deliberate searches.They also mediate users’ attempts to communicate with others by affecting whether their speech reaches an audience.
- 2.1 Algorithmic Reach and the Value of Speech: Algorithmic demotion can sufficiently limit communicative freedom to require justification, even when content remains searchable rather than removed.The chapter treats making speech very difficult to see as morally insufficiently different from removal in some cases.
- 2.1 Algorithmic Reach and the Value of Speech: Possible responses include limiting platform size or supporting multiple online public forums with different recommendation algorithms, without allowing states to dictate a particular recommender.These proposals aim to give citizens more meaningful opportunities to be heard while avoiding direct state control over algorithmic design.
- 2.2 Algorithmic Recommendation as Expressive Conduct: Whether recommendation algorithms are expressive remains contested because platforms’ choices may reflect corporate speech rights, while curation can operate predictively rather than editorially.The chapter argues that predictive aims can still express values, such as reducing polarization, so the relevant question is what the system predicts and why.
- 2.3 Algorithmic Recommendation and the Moral Foundations of Free Speech: If recommenders hinder truth, autonomy, or democracy, that supplies a reason for regulation, although corporate speech rights may limit how overriding that reason is.The chapter also notes that the causal link between recommendation and affective polarization is difficult to establish.
3 Free Speech and AI chatbots
The chapter examines how AI chatbots raise questions about artificial agents’ speech rights, users’ right to receive information, and the treatment of human-endorsed AI-generated speech. It also considers how recommendation algorithms shape speech visibility and how platform governance may interfere with expression.
- Chatbot regulation: Two regulatory scenarios ask whether governments may require LLM developers to adopt prescribed safety restrictions or viewpoint diversity.The examples include restrictions on dangerous instructions and requirements that political information reflect diverse viewpoints.
- Users’ right to receive information: The chapter asks whether users’ right to receive information may constrain chatbot moderation, amid concerns about both insufficient and excessive content restrictions.The discussion presents worries about dangerous outputs and claims that some corporate moderation policies exceed international human-rights standards.
- AI speech rights: Whether conversational agents possess speech rights depends partly on whether AI systems can count as intentional agents and possess relevant capacities.The discussion considers representational and motivational states, environmental interaction, consciousness, and reflective endorsement.
- AI speech rights: A popular view denies LLMs a non-derivative right to free expression because they lack consciousness, human rational and linguistic capacities, and speech-related interests.An output-only approach is presented as either unconvincing or as derivative attribution of rights, shifting attention to the interests of humans interacting with LLMs.
- Recommendation and AI-generated speech: Recommendation algorithms are relevant to free speech because they drive speech visibility and invisibility, while algorithmic demotion can interfere with expression much like removal.The chapter also examines whether human-posted AI-generated content should count as the user’s own speech and how realistic fabrications challenge platform policies.