A human or a computer agent: the social and cognitive effects of an e-learning instructor’s identity and voice cues

Abstract

An instructor in an e-learning video can identify as a human or a computer agent. Relatedly, they can project a human-recorded voice or a machine-voice generated from a classical text-to-speech engine. This study examines the effects of an e-learning instructor’s identity and voice cues on an instructor’s social ratings, learners’ cognitive load, and learning performance. A between-subjects laboratory experiment was conducted where university undergraduates (n = 108) interacted with either one of the four e-learning videos featuring different pairings of an instructor’s identity (human versus agent) and voice (human-voice versus machine-voice) cues that delivered a lesson on programming algorithms. The findings affirmed the voice effects in multimedia learning in that the human-voice enhanced social and learning outcomes more than the machine-voice, irrespective of the identity cues. Credibility ratings were diminished when an instructor identified as a computer agent projected a human-voice than a machine-voice—additionally, endowing a human-identified instructor with a machine-voice prompted learners to assign lower intrinsic cognitive load. These observations imply the congruence/incongruence effects of identity-voice cue pairings on social and cognitive load outcomes. Theoretical and practical implications are discussed in this paper.