We Need an Ontology of Visual Search Tasks

Visual search tasks are widely used in the laboratory, yet also generalize to important real-world tasks. While some real world tasks (e.g., medical imaging, security screening) are the topic of much research, connecting the cognitive expectations of in-lab tasks to real world scenarios is hampered by the lack of consistent categories and terminology to describe important parameters of visual search. Thus, we here propose an ontology to categorize visual search based on conceptual differences across tasks, hopefully easing generalization of predictions between laboratory tasks and to real world behavior. This ontology initially stemmed from a literature review of research on attentional templates, which revealed that researchers categorize visual search tasks and the accompanying attentional templates using idiosyncratic categories that are often only implicitly defined. In addition to the aforementioned difficulties that stem from this problem, the lack of agreed-upon categories and terms leads to difficulty in comparing between existing studies or identifying gaps in existing knowledge because the mappings from task-specific parameters or conditions to abstract concepts can be unclear or inconsistent. We suggest that these problems might be ameliorated by establishing a systematic classification system to help better organize the existing visual search literature and new studies. These distinctions are based on: (1) the number of target-defining dimensions, (2) number of targets sought per dimension, and (3) disjunctive vs. conjunctive search. These distinctions will allow for more nuanced comparisons between visual search tasks because task-specific parameters will be mapped back to standardized conceptual categories. The proposed ontology will also help to identify potential areas in the visual search literature that are lacking and may require further investigation. Moreover, the systematicity of the ontology makes it readily expandable: for example, further divisions could be identified based upon spatial vs. temporal search.