A recent arXiv preprint proposes co-creative learning, a paradigm in which humans and AI act as biological and artificial agents that mutually integrate their partial perceptual information and knowledge. The goal is not one-sided assistance but the joint construction of shared external representations, according to the abstract.
The method relies on Metropolis-Hastings interaction, a sampling-based approach that likely allows each agent to iteratively propose and accept contributions to the shared output. The abstract does not provide implementation details, but the framing suggests a symmetric, turn-taking exchange rather than a conventional teacher-student setup.
Because the source is only an abstract, specifics about evaluation or applications are not available. The contribution at this stage is conceptual: defining co-creative learning and grounding it in a probabilistic interaction framework. Further reading of the full paper would be needed to assess how the approach works in practice.