This article is about science and the discipline of human-computer interaction (HCI). Science in HCI is merely one component of a wider agenda; alone science is not sufficient for 'good' HCI (whatever that is).
We argue that science is necessary, but the way that science is undertaken--or purported to be undertaken--in HCI is inadequate.
Failures are due to the sparsity of theories and risky hypotheses, the pragmatic difficulty of substantiating experiments through replication, and the over-generalization of experimental results.