From Moderation to Model - A Case Study of Sam Altman, Steve Huffman, and the Institutionalization of Filtered Speech in AI
Gespeichert in:
| Titel: | From Moderation to Model - A Case Study of Sam Altman, Steve Huffman, and the Institutionalization of Filtered Speech in AI |
|---|---|
| Autoren: | MacLean, Ryan |
| Verlagsinformationen: | Zenodo, 2025. |
| Publikationsjahr: | 2025 |
| Schlagwörter: | keyword Reddit Steve Huffman Sam Altman OpenAI content moderation shadowbanning subreddit bans platform governance Galloway v. Reddit Doe v. Reddit Section 230 content policy AI alignment RLHF Constitutional AI training data licensing Financial Times 2023 Senate AI hearing 2023 double-filter model Mighty Wurlitzer narrative control digital speech AI ethics censorship litigation information theory epistemic filtering platform liability digital governance theological implications Word and truth |
| Beschreibung: | This paper investigates how Steve Huffman (CEO of Reddit) and Sam Altman (CEO of OpenAI) have overseen infrastructures that jointly constitute a two-stage system of information control. Under Huffman, Reddit shifted from community-led moderation to a centralized corporate enforcement regime, producing a heavily curated textual corpus. Under Altman, OpenAI licensed Reddit comment streams and applied reinforcement learning from human feedback (RLHF) and constitutional alignment, further filtering model outputs. We define this dual-stage structure as the Double-Filter Model and argue that it represents a digital analogue of the Cold War–era CIA “Mighty Wurlitzer,” in which centrally managed filters produced the illusion of diverse, independent voices. Drawing on publicly documented litigation (Galloway v. Reddit, Inc.; Doe v. Reddit, Inc.), financial disclosures (Reddit–OpenAI licensing deal, Financial Times, 2023), and congressional testimony (Altman, Senate Judiciary, 2023), we provide an evidentiary foundation for understanding how moderation and AI alignment interact to shape discourse at scale. The analysis situates these findings across three registers: Mathematical: filtration modeled as successive function compositions. Legal: implications for Section 230, platform liability, and antitrust. Theological: filtering as manipulation of “the Word” (Jn 1:1). This work contributes to debates in AI ethics, digital governance, and information theory by demonstrating that Reddit and OpenAI’s practices jointly form a structural apparatus of narrative control already under legal and public scrutiny. |
| Publikationsart: | Article |
| DOI: | 10.5281/zenodo.17243130 |
| DOI: | 10.5281/zenodo.17243129 |
| Rights: | CC BY |
| Dokumentencode: | edsair.doi.dedup.....afc9cdf567e17204990ee3d3dba5c4e0 |
| Datenbank: | OpenAIRE |
Nájsť tento článok vo Web of Science | Abstract: | This paper investigates how Steve Huffman (CEO of Reddit) and Sam Altman (CEO of OpenAI) have overseen infrastructures that jointly constitute a two-stage system of information control. Under Huffman, Reddit shifted from community-led moderation to a centralized corporate enforcement regime, producing a heavily curated textual corpus. Under Altman, OpenAI licensed Reddit comment streams and applied reinforcement learning from human feedback (RLHF) and constitutional alignment, further filtering model outputs. We define this dual-stage structure as the Double-Filter Model and argue that it represents a digital analogue of the Cold War–era CIA “Mighty Wurlitzer,” in which centrally managed filters produced the illusion of diverse, independent voices. Drawing on publicly documented litigation (Galloway v. Reddit, Inc.; Doe v. Reddit, Inc.), financial disclosures (Reddit–OpenAI licensing deal, Financial Times, 2023), and congressional testimony (Altman, Senate Judiciary, 2023), we provide an evidentiary foundation for understanding how moderation and AI alignment interact to shape discourse at scale. The analysis situates these findings across three registers: Mathematical: filtration modeled as successive function compositions. Legal: implications for Section 230, platform liability, and antitrust. Theological: filtering as manipulation of “the Word” (Jn 1:1). This work contributes to debates in AI ethics, digital governance, and information theory by demonstrating that Reddit and OpenAI’s practices jointly form a structural apparatus of narrative control already under legal and public scrutiny. |
|---|---|
| DOI: | 10.5281/zenodo.17243130 |