Investigating causal understanding in LLMs — AI Alignment Forum