CNPEN - Ethical issues of conversational agents
Conversational agents (also called "chatbots") are digital systems capable of interacting with their users in natural language, both written and spoken. Today, they can be found in many everyday objects (smartphones, connected speakers, cars, websites): they provide health advice, facilitate recruitment, manage after-sales services, provide remote training, help vulnerable people, etc.
In France and at European level, conversational agents, because they rely on personal data, are beginning to attract the attention of the authorities, regulators and the public. Beyond these regulatory issues, many ethical questions arise. What behaviours and beliefs do we have in relation to conversational agents? Can a chatbot lie to its user? Will the mistakes of conversational agents be more acceptable than those of a human being? Should a chatbot systematically imitate a human being?
Among the various ethical issues identified by the committee, the main tensions concern the blurring of the boundaries between a machine and a human being, the imitation of language and emotions by chatbots, and the ability of conversational agents to manipulate human interlocutors.