Further Reading

The Piagetian distinction between assimilation and accommodation motivates the developmental interpretation used here; Drescher gives its classic computational schema realization [ Piaget , 1985 , Drescher , 1991 ] . For the ALFWorld environment and its alignment between textual and embodied interaction, see Shridhar et al. [ 2021 ] . Standard reinforcement learning is treated systematically by Sutton and Barto [ 2018 ] ; temporally extended actions by Sutton et al. [ 1999 ] ; deep value learning by Mnih et al. [ 2015 ] ; and curiosity-driven exploration by Pathak et al. [ 2017 ] . The distinction between novelty and surprise is especially relevant to the B1 control [ Barto et al. , 2013 ] . The DIAL-GIRL and schema experiments in this chapter are reported as mechanism studies: readers should distinguish their structural claims from state-of-the-art ALFWorld comparisons based on substantially larger training budgets or pretrained language agents.