Time Machine Experiments: Using Historically-Bounded AI for Inquiry into the Human Mind

Open original

Researchers built AI models trained only on pre-1930 text to let people interact with "historical minds," then tested whether this changed how participants viewed the past. Talking to a 1930s-bounded AI reduced the "moral decline illusion"—the common belief that people were more ethical back then.

  • A randomized experiment with 240 participants showed that interacting with historically-bounded LLMs reduced the illusion of moral decline compared to using contemporary AI
  • The study demonstrates a new experimental method where temporal knowledge boundaries become controllable variables, turning philosophical thought experiments into testable science
  • This approach reveals how our understanding of the past gets distorted by everything that happened after it—we can't see history clearly without the filter of hindsight

Researchers built a language model trained exclusively on text from before 1930 and had 240 people interact with it in a controlled experiment. The core question was whether talking to an AI with a genuinely bounded historical perspective—one that literally knows nothing about the last century of events—could shift how people think about the past. Most of us judge history through a filter of everything that's happened since: wars, social movements, technological change, moral progress we assume we've made. This experiment strips that filter away by creating an actual historical mind to talk to. The results showed a measurable effect. Participants who interacted with the pre-1930 model experienced less of what psychologists call the "illusion of moral decline"—that tendency to believe people were better, more virtuous, or more principled in the old days compared to now. Talking to someone from 1930 who had no knowledge of subsequent events apparently made the past feel less like a lost golden age. The control group used a contemporary model trained on current text, so the difference came directly from that temporal boundary. The paper frames this as opening up a new experimental method. Instead of just thinking through hypotheticals, researchers can now turn thought experiments into actual testable interactions by treating time-bounded knowledge as an experimental variable. They call it "science fiction science"—taking speculative ideas and making them runnable. This could extend beyond moral perception studies. You could use historically-bounded models to explore how people reason differently when they lack hindsight, or how specific pieces of missing information reshape judgment and decision-making.

Ask about this article