Abstract:Science exhibitions, as the major approach for science centers to communicate science, are considered to be scientific educating generally. However, few evaluations have been done to prove its impact on people's learning. Relevant studies could be classified in three categories: using the audience's satisfaction survey to reflect the impact, simply evaluating the impact by the holding power of exhibits and concluding the impact according to audience's cognitive changes before and after exhibitions. None of these approaches value the reasons behind this learning, specifically how exhibits trigger audience to learn at the scene, so that the same old audience-unfriendly issues still remain in the past and current exhibitions hindering the improvement of exhibition quality. Thus, this study proposes a science exhibition evaluation framework towards learning impacts, with the combination of behavior method and comparative study, further exploring its impact on audience's knowledge level through visitors’ behaviors. The framework has also been tested in two exhibitions at the Guangdong Science Center. Results indicate four major conclusions. Certain social stimulus improves its learning impacts. There is an imbalance between entertainment and education. Rethink the combination of multiple exhibit forms. Conventional experimental simulation forms should be upgraded.