TIPs: Transparency Information Pacts

Phillip Walker, Ugur Kuter, Christopher A Miller, Helen Wauck, Matthew DeHaven, Christopher Geib

Keywords: transparency, trust, human-machine teaming

TL;DR: This paper presents the idea of Transparency Information Pacts, or TIPs, as a way to formally represent transparency information and better allow the information to be requested and conveyed back to the human when they are best able to use it.

Abstract: A key component for building trust in automated systems is improved transparency into the decisions made by the automation and the reasons behind those decisions. Current models of transparency, however, assume that the operator has the time and mental capacity to digest the transparency information the moment it is provided. Often times, this is exactly the moment where the operator is most overloaded, which is why they delegated to automation in the first place. This paper presents the idea of Transparency Information Pacts, or TIPs, as a way to formally represent transparency information and better allow the information to be requested and conveyed back to the human when they are best able to use it. TIPs build off the idea of Lifecycle Transparency presented in (Miller 2021), which encourages the use of other mission phases--pre-mission planning and post-mission debrief--for conveying transparency information where appropriate. Here we present multiple types of TIPs, how they are structured and used, along with illustrative examples.