The BESTBOT was developed at the School of Business FHNW from March to August 2018. Predecessor projects were GOODBOT (2013) and LIEBOT (2016). Prof. Dr. Oliver Bendel has been doing research in the young discipline of machine ethics for several years. In cooperation with robotics and artificial intelligence (AI), it designs and produces moral machines. At the beginning of 2018 Bendel presented his paper “From GOODBOT to BESTBOT” at Stanford University, which laid the foundation for the BESTBOT project. David Studer programmed the chatbot in Java. Prof. Dr. Bradley Richards assisted him in technical matters. Like LIEBOT, BESTBOT is a networked system that exploits search engines and dictionaries. It analyzes the user’s text input with text-based emotion recognition software. At the same time, face recognition is used, again with emotion recognition. For example, if the user states that he is doing well but reveals something else on his face, the chatbot addresses this contradiction. It recognizes both small and big worries. Like the GOODBOT, the BESTBOT can escalate over several levels and provide a suitable emergency number. Like its predecessor, it makes it clear that it is only a machine. It is also special that it cites the source of factual allegations. The BESTBOT will be presented at conferences in 2019.
Fig.: The chatbot uses face recognition