On Engineering AI Agents for Privacy

Abstract

Artificial Intelligence agents are becoming increasingly present in our day to day lives. Their interactions with users are creating expectations of privacy that so far have been addressed solely by the implementation of mechanisms in the behaviour of the agents. We argue that the engineering process has an influence on this behaviour, so privacy considerations need to be made in both places. We show the dependency between the engineering process and the behaviour of an agent, explore how the framework of Contextual Integrity can inform the engineering process, and how its holistic application can be used to effectively incorporate privacy into an AI agent.

Publication
Symposium on Applications of Contextual Integrity