Tap ’N Seek: Causal Learning from Interactive Graphics

Anonymized data for manuscript submission "Tap 'N Seek". Includes the cleaned interaction data, behavioral coding results, and analysis code and results for each participant. Abstract: Interventions, in both adults and children, are advantageous for causal learning. However, not all interventions are possible in the physical world. Digital environments have no such deficiency, but it is unclear if causal learning in digital environments benefit from interventions like their physical counterparts. This begs whether interactive graphics offer a comparative advantage over video graphics for causal learning in digital environments. In this study, 48 children aged 5-8 years of age were either freely able to generate interventions on causal video games or were passively shown videos of interventions produced by a past intervening participant. The intervening children scored better than their counterparts passively observing information-matched videos. Furthermore, the quality of interventions was correlated with learning outcomes only when children actively selected interventions themselves. Leveraging the flexibility of graphics, we also assessed the impact of different causal graphs and scenario priors on causal learning. We found that causal relations likely more familiar to children benefitted less from the ability to intervene than less familiar causal scenarios. However, we found no effect of causal graph structure, nor age, on learning ability. Unlimited by real world resource, safety, and feasibility constraints, these findings confirm and clarify results in physical settings, set the stage for more interventions accessible for causal learning, and make a case for the provision of interactive graphics for causal learning in children.