SFB 1187 ›Medien der Kooperation‹ an der Universität Siegen
Ringvorlesung: „Learning (in) Digital Media“ – Katharina Rohlfing (Universität Paderborn): „Scaffolding and monitoring: Aspects of learning in the social design of explainable AI systems“
Tuesday, 19. October 2021, 18:00 - 20:00 Uhr

Die Ringvorlesung findet als hybride Veranstaltung mit Teilnahme vor Ort und online-Zugang statt. Der Zoom-Link für die Vorlesung wird rechtzeitig über die Mailingliste des SFB zur Verfügung gestellt. Gäste können sich per Mail bei Damaris Lehmann anmelden. Sende eine E-Mail

 

 

Katharina Rohlfing (Universität Paderborn): „Scaffolding and monitoring: Aspects of learning in the social design of explainable AI systems“

What is driving a recent surge of interest in explainable AI (XAI) are technological advancements in machine learning affecting humans’ lives on the one hand but also  regulatory initiatives fostering transparency in algorithmic  decision-making on the other hand. Explainability is discussed as a solution to socio-technical challenges such as intelligent software providing incomprehensible decisions or big data enabling fast learning but becoming too complex to fully comprehend and judge its achievements. The vision is that with explainable AI, more insights  into the function, decisions, and usefulness of algorithms are expected.
  
  Yet, if an explanation is successful, it results in an understanding. Current XAI research is centering around solutions of how to explain and what could be a successful explanation leading to understanding. In the presentation, I will elaborate on a new approach that draws on two concepts known from learning and education: scaffolding and monitoring. Both are important resources for the social design of explainable AI systems: Whereas in current XAI research,  the addressee/explainee is mostly seen as a passive receiver, these concepts lead to a particular form of interaction to which both partners actively contribute. The assumption is that it will result in the explanation being tailored to a particular form  of understanding thus gaining on relevance.

 

Katharina J. Rohlfing received her Master’s in Linguistics, Philosophy, and Media Studies from Paderborn University, Germany, in 1997. As a member of the interdisciplinary Graduate Program “Task-Oriented Communication,”  she received her PhD in  Linguistics from Bielefeld University in 2002. In 2006, she became a  Dilthey Fellow (Volkswagen Foundation) and Head of the Emergentist Semantics Group at Bielefeld University’s Cluster of Excellence Cognitive Interaction Technology (CITEC). Currently, she is professor of psycholinguistics at Paderborn University and speaker of the Transregional Research Center „Constructing explainability“, funded by the DFG. Her work is on multimodal dialogical coordination and learning with a strong interest in cognitive  modeling, developmental robotics, and HRI.