ABSTRACT
When conveying information about routes to follow in complex environments, human route-givers adapt to route-receivers’ familiarity with the environments in their choice of landmarks. Meanwhile, as route-givers themselves have experienced the environments within a social role, the landmarks they select may also differ significantly. This research investigated how these two factors influence landmark selection when communicating routes in indoor environments. Two groups of participants were recruited to conduct indoor landmark selection experiments for familiar and unfamiliar route-receivers in a multi-functional university building. The results show an interaction effect between these factors in indoor landmark selection. These findings lay an empirical ground for developing human-centered mobile navigation systems that can adapt to users’ social roles and their familiarity with the environment.
Acknowledgments
The authors would like to thank Prof. Sara Fabrikant, Prof. Ross Purves, Dr. Tyler Thrash, Dr. Sascha Credé, and Dr. Annina Brügger, as well as three anonymous reviewers, for providing their constructive comments on this work.
Notes
1 The level is equivalent to the vista scale as defined in Montello (Citation1993).
2 If time allows, an alternative solution is to employ a think-aloud approach to investigate participants’ reason for each pairwise comparison.
3 Before the actual experiment, 6 additional volunteers (3 being familiar with research on landmarks) were invited to check and improve these questions to make sure that they are generally easy-to-understand by human participants.
4 Please note that these signs are “furniture-level” indoor objects, and they are often attached to their corresponding “room-level” objects (e.g., library).
5 Depending on the conceptualization of the data analysis problem, the generalized estimation equations (GEE) method can be also alternatively employed. To complement the above results, we implemented GEE to provide another perspective of the data analysis. The results are shown in Appendix C. We followed the practice of Szmaragd, Clarke and Steele (Citation2013) to select the appropriate working correlation matrix structure according to the quasi-likelihood information criterion (qIC) by four types of working correlation structures (i.e., independent, autoregressive, unstructured, and exchangeable), which is the key step to apply GEE. Finally, based on the smallest qIC values, we selected the autoregressive correlation structure for the GEE analysis in both the staff-familiar and student-unfamiliar groups, and the exchangeable structure for both the staff-unfamiliar and student-familiar groups, respectively.