Knowledge Spillover, Trust, Effort, and Error Exposure in Peer-Assisted Learning

Peer-assisted learning has the potential to improve learning in academic settings and beyond. However, the cognitive and motivational mechanisms of learning through interaction with other learners are not fully understood. Here we present an empirical study in which we compare a peer-assisted learning condition with two individual learning conditions. The empirical findings suggest that both positive and negative peer effects occurred. On the positive side, learners placed in a peer-assisted learning condition allocated more time to practice and they benefited from selectively interacting with the more knowledgeable peers. On the negative side, error exposure and increased cognitive load may have hindered learning in the peer-assisted learning condition. A computational cognitive model developed in the ACT-R cognitive architecture is presented and used to explain the mechanisms of knowledge spillover, trust, and error exposure. This research has implications for designing collaborative learning protocols to increase human collective intelligence and designing artificial intelligence systems that can support human-machine teaming.