Would you let AI cheat for you?

Our new paper in @nature.com, 5 years in the making, is out today.

nature.com/articles/s4158… As we delegate more hiring, firing, pricing and investing decisions to machine agents, particularly LLMs, we need to understand what ethical risks it may entail.

Our new research, based on 13 studies involving over 8,000 participants and commonly used LLMs, reveals two risks of how machine delegation can drive dishonesty and highlights strategies for risk mitigation.

New paper:

Using Facebook's advertising infrastructure (i.e. Surveillance Capitalism) to measure culture at the scale of 2 Billion humans

Brief thread 🧵👇

royalsocietypublishing.org/doi/full/10.10… Culture has played a pivotal role in human evolution, as @JoHenrich summarized in his amazing book The Secret of Our Success. See his Google talk for a summary.

Moral Machine Experiment (MME) results challenged!

New study by @kurtjgray & @yo_bigman published today in @nature along with our reply (@EdmondAwad @SohanDsouza R. Kim @azimshariff @JFBonnefon @JF_Schulz @JoHenrich).

(thread)

Both papers here: rdcu.be/b2v6R The Moral Machine Experiment always gave people two options: should a self driving car go straight, killing one group of people, or should it swerve, killing another group? Focus was on cases of unavoidable harm.

Ethicists, Moralists, Psychologists, and Trolleyologists! I am pleased to announce our latest paper: "Universals and variations in moral decisions made in 42 countries by 70,000 participants" just out in @PNASNews

pnas.org/content/early/…

(thread) In our "Moral Machine" platform, we added "classic" mode to collect data on the three most-studied variants of the Trolley Problem: classic "pull the switch", "push the fat man over the footbridge", and the "loop" case designed to eliminate the means vs. side effect confound