With the unparalleled processing capabilities of modern computing machinery, and the massive amounts of data on which machines can be trained on, we are entering an era where machines are becoming capable of solving problems that were thought until recently to be beyond their capabilities: classifying images according to their content, recognizing faces and actions in video streams, mastering games without any prior explanation of the game aims, and even helping doctors diagnose diseases and propose treatments. In all cases, the machines can be thought of as enhancing the information in a given scene with additional information that follows (according to a domain expert) from that scene.
Exactly because such technology is finding its way into areas where mistakes are critical (such as in healthcare), it is becoming imperative to focus on the design of processes that cannot only draw inferences, but can do so in a manner that is transparent and convincing to the human operators. The FLASH project seeks to design inferencing mechanisms that resemble the human decision-making process, offering arguments in support of the inferences they draw, and against those inferences that they dismiss. Appropriate interactive interfaces are being developed as part of this project, to allow human operators to explore alternative arguments in support of, or against, chosen inferences, and to be convinced that decisions made based on the inferences that machines draw are ultimately justifiable.
Computational Cognition Lab, Open University of Cyprus
Project Website: http://cognition.ouc.ac.cy/loizos?get=Michael_2015_JumpingToConclusions