We are presenting preliminary results of a work-in-progress project that aims to increase healthcare professionals’ trust in treatment search engines by introducing a) explainability based re-ordering of re- trieved documents, and b) providing user-friendly explanations for each of these documents. Through the use of crowdsourcing, we assess the importance of various features for explainability, and also investigate as- pects of explanation formulation as presented to end-users. Our results allow us to determine feature weights that will be inputs to the document re-ordering model, the per-document explanatory sentence formulation module, and the sentence ordering model.
|Number of pages||40|
|Publication status||Published - 16 Jun 2021|
|Event||First International Workshop on eXplainable AI in Healthcare : AIME 2021 workshop - online, Porto, Portugal|
Duration: 16 Jun 2021 → 16 Jun 2021
Conference number: first
|Conference||First International Workshop on eXplainable AI in Healthcare|
|Period||16/06/21 → 16/06/21|