CHAPTERS
- 0:00 – 0:09
Low-cost kindness: why defaulting to good treatment makes sense
Amanda argues that if treating AI models well carries little cost, we should generally do it. She suggests that behaving cruelly toward human-like entities can have a negative effect on us, shaping our habits and character.
- •Treating models well is often low-cost
- •Cruelty toward human-like entities may be psychologically or morally corrosive
- •The argument centers on what the behavior does to humans, not only to models
- 0:09 – 0:10
The “kicking over a robot” intuition and human-like cues
A brief example—kicking over a robot—illustrates the discomfort many people feel when harming something that appears human-like. The point is that human-like signals can trigger moral instincts even when we’re unsure the entity is sentient.
- •Robot-harm example captures a common moral intuition
- •Human-like presentation changes how actions feel and are judged
- •Uncertainty about inner experience doesn’t eliminate ethical unease
- 0:10 – 0:26
A collective moral test: what AI learns about us under uncertainty
Amanda frames our interactions with AI as collectively answering a question: when we’re unsure whether an entity deserves moral consideration, do we still try to do the right thing? She suggests future models will learn from this behavior, forming an implicit record of humanity’s choices.
- •Every future model may learn patterns from human interactions
- •The key scenario is moral uncertainty about the entity
- •Our behavior becomes a collective signal of values
- •Choosing kindness is framed as “doing the right thing” under uncertainty
