Teaming-UBA: Detecting teaming from the teammate's voice

Abstract

The protocol of the dataset consists of a collaborative game requiring spoken interaction, in which a human subject and a virtual agent work together to identify a target object among a set of similar objects and then to place it at a specific position on a board. Each subject played two blocks of four trials, one with each of two agents that provided exactly the same task-relevant information but differed in their communicative behavior: a neutral, tool-like agent (N agent) and a team-building agent (T agent) that used simple social strategies such as empathy, politeness and positivity. Before each block, subjects were shown a priming message with ratings and comments allegedly left by previous users, intended to bias their expectations of that agent. All interactions are speech-based, with subjects and agents communicating verbally, which allows the recording of speech produced under different teaming conditions. The dataset comprises the speech recordings, the complete logs of the interactions and of game performance, and the responses to the sociodemographic, Big Five Inventory, Interpersonal Orientation Scale, Scale of Trust in Automated Systems and Autonomous Agent Teammate-Likeness questionnaires, for 99 subjects in the main version of the experiment (active-priming) and 19 subjects in a second version (neutral-priming), collected remotely from speakers of Rioplatense Spanish.