EUCA for End-User-Centered Explainable AI Framework has forms that can help build a first explainable AI tool, just to "try and test". The principle is then to iterate co-design, co-creation, review of the prototype to refine the quality of explanation and make it matching better the user needs/goals. and of course as multiple user profiles have different needs, User interface and user experience as well adn Human-Machine-Interaction are core component of the iterations.
These principles are synergies with what drives iSee community and project.
In iSee we distinguish end-users-likely expecting explanation to justify the AI system prediction/recommendation (what-if, counterfactuals, alternative results with similarity score, key words causing the result to appear). Such can be derived from input/output data set as post-hoc explanation for instance.
AI dvelopers or designers will more be interested in how the AI works (algorithm behaviour) which requires low-level features from data set.
About The Author: ISee Team
More posts by iSee Team