Are experiments on human-robot interaction replicable? Some critical thoughts.
Sprache des Vortragstitels:
12. Tagung der Fachgruppen Arbeits-, Organisations- und Wirtschaftspsychologie sowie Ingenieurspsychologie der DGPs
Sprache des Tagungstitel:
2011 hit psychology as a science hard. Cases of fraud became public, Bem (2011) published controversial studies on parapsychological phenomena and Simmons et al (2011) showed that even absurd hypotheses can be made statistically significant due to researcher?s degree of freedom. Events like these eventually led to a replicability crisis. Since then, many studies have failed to replicate, inevitably leading to an erosion of confidence in research outcomes. Reformers have pointed out many reasons for the crisis including methodological problems (e.g. low statistical power), publication practices, lack of theory, lack of validations, and even epistemological shortcomings. These problems have been found in different scientific disciplines. But how is human-robot interaction (HRI) research doing in this respect? Is HRI facing the very same problems of confidence as other disciplines?
This position talk aims to give a first impression of the current state concerning replicability of HRI research. In doing so, results from a recent systematic literature review on personal space in HRI (Leichtmann & Nitsch, 2020) and single replication studies in HRI are summarized and discussed in light of psychology?s confidence crisis. The results indicate that HRI is facing the very same problems of replicability. The talk ends with some possible solutions, that had been widely recommended by reformers previously, to tackle these issues in future studies.
Taken together, HRI research could benefit from the bitter lessons learned in psychology. Intensifying collaboration between psychology and the more technical disciplines could be key for better replicability of future HRI studies.