Information and communications technologies (ICTs) in healthcare are often introduced with expectations of higher-quality, more efficient, and safer care. Many fail to meet these expectations. We argue here that the well-documented failures of ICTs in healthcare are partly attributable to the philosophical foundations of much health informatics research. Positivistic assumptions underpinning the design, implementation and evaluation of ICTs (in particular the notion that technology X has an impact which can be measured and reproduced in new settings), and the deterministic experimental and quasi-experimental study designs which follow from these assumptions, have inherent limitations when ICTs are part of complex social practices involving multiple human actors. We suggest that while experimental and quasi-experimental studies have an important place in health informatics research overall, ethnography is the preferred methodological approach for studying ICTs introduced into complex social systems. But for ethnographic approaches to be accepted and used to their full potential, many in the health informatics community will need to revisit their philosophical assumptions about what counts as research rigor.