Towards an empirical investigation of inner alignment — AI Alignment Forum