Causal Modeling

Like most complex systems learning environments, Connected Worlds portrays dynamic emergent behavior which results in nonlinear state transitions. These state transitions demand that the learners evolve their solutions while path dependencies prevent backtracking. Further, the problems presented within complex systems are open-ended, meaning they are characterized by loosely defined reward functions that allow a range of: (1) solutions states or configurations, (2) strategies for reaching those solutions, and (3) ways to implement those strategies, many of which are often unknown a priori. All these factors make understanding and supporting learners’ explorations challenging. This work leveraged the structural and computational advantages of causal models in representing these complex behaviors and is useful in documenting the large solution space inherent in open-ended learning environments. 

The constructed causal model for Connected Worlds museum exhibit involved devising a novel framework called Formative Exploration Support Template (FEST) that guides the representation of the learners' actions, evolving system states and learning goals that could inform salient scenarios as nodes in the causal model. The causal model serves as a foundation for the computational pipeline that generates formative feedback for the visitors interacting with the exhibit. This is first to our knowledge to generate formative feedback that could aid exploration in open-ended learning environments.