Real-time planning of explorative and exploitative actions across development: An exploratory mouse-tracking study

• In Event: Factors Affecting Goal-Directed Behavior

Abstract

Goal-directed action is commonly studied in instrumental settings, where agents pursue expected outcomes with certain or probabilistic value. However, humans also assign value to the information they can gain from the world and interact with their environment motivated by opportunities for learning, driven by the urge to constantly improve their world model – a motivational state often termed curiosity. Neuroscientific findings suggest extrinsic and intrinsic value of actions share brain mechanisms and representations (Blanchard, Hayden, & Bromberg-Martin, 2015). Incorporating these findings with an embodied cognition perspective, all actions can be viewed as decision processes, which show instant preferences for external reward (immediate or delayed) or for learning. It has been suggested that the level of action modulation by curiosity gradually decreases from infancy to adulthood, resulting in less broad exploration (Gopnik, 2017), or otherwise a different balance between specific and diversive curiosity (Bahzydai & Gestermann, 2020). However, curious behavior might also vary as a function of individual cognitive skills (especially executive functions, on top of age differences (Chrysikou, Weber & Thompson-Schill)) and more stable dispositions towards uncertainty and complexity. While the relationship between exploratory behavior and maturation has been studied before (Blanco & Sloutsky, 2021), we aimed to focus on the specific process of planning an exploratory versus an exploitative action, and investigate how these decisions unfold real-time. Furthermore, we aimed to examine how its parameters can be predicted by age and individual differences. In our exploratory study, we examined how 5-7 year-olds, 13-15 year-olds and adults chose between different options to interact with, when these options led either to the attainment of a rewarding goal, to missing information (as a proxy for specific curiosity) or to novel unpredictable stimuli (as a proxy for diversive curiosity). We looked into how the competition between these options was reflected in the real-time action plans; i.e., at the specific hand kinematics while participants made their choices, as we expected an implicit modulation of movement by curiosity in the younger group even when their final choices differed. Participants completed an online decision-making task as we tracked their mouse positions, two stimulus-preference tasks with varying levels of uncertainty and complexity and standard executive functions tasks. Results showed a greater preference for novel stimulation in children compared to the other groups, which replicates previous findings (Figure 1). Furthermore, a mixed-effects analysis on movement parameters revealed differences in the temporal parameters of children’s hand kinematics, suggesting greater conflict between reward and novel experiences in children (Figure 2). Adolescents’ and adults’ movement parameters did not differ as a factor of choice. Complexity preferences, as well as working memory predicted adolescents’ curious choices: greater preference for complexity predicted broader exploration (more diversive curiosity choices), while lower WM scores predicted more choices driven by specific curiosity. In total, our results support that school-aged children’s preferences for new, unpredictable stimuli can be reflected in their movement. Next steps will focus on the unfolding of these decision processes in time, in an attempt to relate them with learning.

Authors

• Penny Bounia-Mastrogianni, Birkbeck University of London
  Presenting Author

• Richard P. Cooper, Birkbeck, University of London

• Denis Mareschal, Birkbeck, Universty of London