Context matters: Observer perception of dog-human social interactions becomes more variable in a novel human context

Subjective ratings are frequently used to interpret the social behavior of dogs, especially in the context of human–dog interactions due to their convenience and practicality. To assess whether ratings are sufficiently reliable, researchers are tasked with documenting the level of agreement between multiple observers. Several of these comparisons are conducted within a single context (e.g., interactions solely with the experimenter), hence raising the question whether interpretations would also converge across different social contexts. In this study, three observers with varying levels of experience, one expert in canine behavior and welfare and two dog owners, rate both the cooperative tendencies and petting responses of 20 dogs across two social contexts: interactions with (a) their owners (i.e., a familiar social context) and (b) the experimenter (a novel social context). We found that observer scores aligned only in the novel social context, suggesting a divergence among observers while interpreting the interaction of dogs with their owners. We believe the unique nature of the long-term bonding between dogs and their owners, coupled with observers’ limited access to the development of these bonds, contributed to the lack of inter-rater reliability. Our findings highlight the importance of conducting behavioral tests across multiple contexts.