One of the main objectives of the Artificial Intelligence is to reproduce computational mechanisms that can replace or assist humans in several day-to-day tasks. In particular, the researches within Knowledge Engineering and related areas, like Ontology Engineering and Conceptual Knowledge, have sought tools to replace or assist human workers on tasks that require a high level of expertise. Generally, this effort has generated results in the way that knowledge systems are currently successfully employed in several areas of the industry.

However, advances are still necessary in several areas where the actual state of the art of Knowledge Engineering doesn’t provide solutions. In particular, domains where the tasks have a strong visual component, like Medicine and Geology, offers difficulties and resistance to automation or to development of auxiliary tools to the experts. This happens fundamentally by the difficult process of externalization and representation of the knowledge involved in the solution of this kind of task, called here of visual knowledge. More specifically, we define visual knowledge as the set of mental models that support the process of reasoning over the information associated to the spatial arrangement and other visual aspects of the domain’s entities. The objective of the BDI Group is to generate methods and tools for the creation of mechanisms that assist the solution of tasks involving visual knowledge. For such, several aspects of representation of knowledge and reasoning shall be explored. In the field of representation, the challenges are to provide constructs of representation that allow capturing the iconic nature of the visual knowledge. Another challenge is to arrange those constructs in a framework for integrated representation of visual knowledge and symbolic/propositional knowledge, what involves challenges in the development of effective representation languages, besides challenges in the ontological validity of the created models.

In the aspect related to reasoning, the challenge is in utilize the visual knowledge models in the solution of tasks. This involves researching ways of utilize the internal structure of the construct of visual representation so that it is possible to infer significant relations within the application domain. The relation between visual/propositional knowledge and visual reasoning raises others questions, like the problem of symbol grounding (or symbol reference), that deals with the relation between the symbol and it’s real manifestation; a model of visual knowledge should ideally foresee ways of keeping this relation, to be possible build mechanisms for interpretation of raw sensory data (e.g. images and videos).