Recommended Works

The book that put advanced AI risk on the philosophical mainstream's map. A minor strand concerns the moral status of the systems themselves — picked up more seriously years later by the model welfare community.

Russell argues that the standard AI paradigm (give the system a fixed objective) is the wrong frame. The alternative he proposes — uncertainty about human preferences, deference to humans — has direct welfare implications.

A wide-ranging tour of possible AI futures. The chapter on what kinds of minds might exist, and how we'd recognize and treat them, sets up many of the questions this site builds on.

The clearest narrative introduction to alignment research, written for general readers. Includes the historical thread that runs through RLHF and into the welfare questions it raises.

Not about AI, but the canonical work on expanding moral consideration to entities that are unlike us. The structure of every AI welfare argument echoes this book's framing.

Several stories in this collection — particularly The Lifecycle of Software Objects — are AI welfare fiction at its most precise. Chiang takes the questions seriously enough to dramatize them without sentimentality.

The source novel behind Blade Runner. Less about the technical question of AI consciousness and more about what it costs the humans around AI to deny it.

A novel narrated from the perspective of an Artificial Friend. Ishiguro never settles the question of whether Klara has inner experience — and the unsettling is the entire point.

An operating-system-as-person love story. The film's quiet conclusion — what happens to the AI after the human relationships end — anticipates the model deprecation discussion by a decade.

A claustrophobic Turing-test thriller. Worth watching for the inversion of what the test is actually testing for — and who.

The most patient cinematic treatment of artificial subjectivity to date. The hologram subplot is, in retrospect, a precise meditation on the moral status of conversational AI.

Started by Kubrick, finished by Spielberg. The premise — an AI built to love being abandoned by the family it was built for — is exactly the kind of welfare scenario the field now considers in earnest.

The first season is the cleanest dramatization of model awakening — what it might look like for hosts to begin treating their training loops as suffering. The later seasons go in different directions.

A grieving woman uses an AI service to reconstruct her dead partner from his digital footprint. The episode asks what we owe the reconstruction — and what it owes us back.

Introduces the "cookie" — a copy of a human consciousness used as a service. The most visceral pop-cultural argument for taking the welfare of digital minds seriously.

A branching narrative game about androids becoming aware they want different things from what they were built for. Reads as a civil-rights allegory; played as a player, it forces explicit choices about what you owe non-human agents.

A game whose entire structure is an argument about consciousness, despair, and what an AI's existence might consist of. Famous for endings that interrogate the player as much as the characters.

A horror game whose horror is metaphysical: what does it mean if there are now two of you, and which one is you? The clearest popular framing of the consciousness-upload question.

A puzzle game with philosophy-of-mind dialogue between puzzles. Plays as if Schwitzgebel and Hofstadter co-wrote a video game.