By Abeer Dyoub,
Dipartimento di Ingegneria e Scienze dell’Informazione e Matematica, Università degli Studi dell’Aquila
This work presents a framework for monitoring chat text in online customer service, to check ethical violations from the employees w.r.t codes of ethics and conduct of their company, and then report these violations to managers in real time. We focus on learning ethical principles from chat text and then use the learned principles for ethical evaluation of other future cases.
This work was performed with a future perspective of ethical chatbots in general and in customer service in particular. Autonomous intelligent agents are playing an increasingly important roles in our lives. They store information about us, and are becoming able to perform tasks on our behalf. Chatbots are an example of such agents that need to engage in a complex conversations with humans. Thus, we need to ensure that they behave ethically. The approach proposed in this thesis is a hybrid logic-based approach, that makes use of Answer Set Programming as a primary knowledge representation and reasoning language, and uses Inductive Logic Programming for learning ethical evaluation rules. Non-monotonic reasoning is an appealing paradigm for ethical reasoning. Generalizing Principles from cases is the solution for ill-defined domains like the ethical domain. Moreover, our approach as a pure logic-based approach provides support for two very important aspects of Machine ethics viz. explainability and accountability.
We claim that no similar work have been conducted in the field of machine ethics so far.