Designing Resilient AI-based Robo-Advisors: A Prototype for Real Estate Appraisal

For most people, buying a home is a life-changing decision that involves financial obligations for many years into the future. Therefore, it is crucial to realistically assess the value of a property before making a purchase decision. Recent research has shown that artificial intelligence (AI) has the potential to predict property prices accurately. As a result, more and more AI-based robo-advisors offer real estate estimation advice. However, a recent scandal has shown that automated algorithms are not always reliable. Triggered by the Covid-19 pandemic, one of the largest robo-advisors (Zillow) bought houses overvalued, eventually resulting in the dismissal of 2,000 employees. This demonstrates the current weaknesses of AI-based algorithms in real estate appraisal and highlights the need for troubleshooting AI advice. Therefore, we propose to leverage techniques from the explainable AI (XAI) knowledge base to help humans question AI consultations. We derive design principles based on the literature and implement them in a configurable real estate valuation artifact. We then evaluate it in two focus groups to confirm the validity of our approach. ... mehrWe contribute to research and practice by deriving design knowledge in accordance with a unique artifact.