XAI for learning: Narrowing down the digital divide between “new” and “old” experts
Auste Simkute, Aditi Surana, Ewa Luger, Michael Evans, Rhianne Jones
Abstract
Regular eXplainable AI (XAI) approaches are often ineffective in supporting decision-makers across domains. In some instances, it can even lead to automation bias or algorithmic aversion or would simply be ignored as a redundant feature. Based on cognitive psychology literature we outline a strategy for how XAI interface design could be tailored to have a long-lasting educational value. We suggest the features that could support domain-related and technical skills development this way narrowing the digital divide between “new” and “old” experts. Lastly, we suggest an intermitted explainability approach that could help to find a balance between seamless and cognitively engaging explanations.