Preprints

This project was inspired by a virtual talk about consciousness, given by Hakwan Lau at UCL in 2020. Hakwan cited some findings from invasive work on non-human primates as evidence for a specific physiological basis of consciousness, and I asked whether he didn’t feel it was immoral to harm these monkeys, given that he clearly believed they were conscious. He responded that it was not clear to him that consciousness was related to morality at all, which I found very surprising. I thought it would be easy to find papers that show that people generally see consciousness as highly relevant to morality, but I couldn’t find any. So, together with Anna, Arianna and Steve, we empirically tested whether this was the case. In our experiments, the answer was a clear YES. People’s moral judgments in questions of life and death were highly correlated with their ascriptions of consciousness.

Large language models, like humans, struggle to overcome the “curse of knowledge” and ignore potentially biasing information. However, unlike humans, they possess the ability to “self-blind” by calling their own API with an appropriately redacted prompt. In this collaboration with Brian we show that, when given the opportunity, they do just that.