Machine minds on Substack
What Substack writers say about whether AI systems think, feel, or matter morally. A systematic reading of the platform: every post we could find on the topic from January 2024 onward, coded by machine with human spot checks. Posts are linked, never reproduced.
What writers say
Each bar is one of six questions, each combining one to three coded claims. Shares are of posts whose author addresses the question in their own voice; posts that only report others' positions, or never raise it, are excluded. "Yes, qualified" means partly (thinking), after a major advance (possibility), or future AI or precaution (moral status). Hover for counts. Filter by publication type, genre and more →
Who writes about it
Publication types were coded from each publication's name, tagline, author bio and post titles. Nine fine categories are collapsed to five for display; see Methods.
Where to go
- Positions: the six questions with filters, how positions differ by who is writing, and the claims behind them.
- Arguments and sources: which arguments appear, what gets cited, who gets named.
- Features of the systems: which features of the AI systems themselves authors point to as evidence, in which direction, and how that mix has shifted.
- Topic map: the relevant posts laid out by similarity, with labelled clusters.
- Citation network: which publications cite which.
- Over time: how the mix of positions has shifted by quarter.
- Notable posts: posts that beat their own publication's norm, and the most-engaged posts overall.
- Changes of mind: posts whose author reports having once thought differently about machine minds, with the passage that says so.
- Browse posts: every relevant post with its codes, filterable and downloadable.
- Publications: one page per publication, with its posts and how it sits in the discourse.
- Methods: how the corpus was built and coded, and how well the coding holds up.