Why The Focus on Expected Utility Maximisers? — AI Alignment Forum