Emergent Goal‐Anticipatory Gaze in Infants via Event‐Predictive Learning and Inference

• DOI: 10.1111/cogs.13016
• Publication Date: 2021-08-11T12:50:15Z
• Authors: Christian Gumbsch; Maurits Adam; Birgit Elsner; Martin V. Butz
• Keywords: Generative model; Anticipation (artificial intelligence)

Abstract From about 7 months of age onward, infants start to reliably fixate the goal an observed action, such as a grasp, before action is complete. The available research has identified variety factors that influence goal‐anticipatory gaze shifts, including experience with shown events and familiarity agents. However, underlying cognitive processes are still heavily debated. We propose our minds (i) tend structure sensorimotor dynamics into probabilistic, generative event‐predictive, event boundary predictive models, and, meanwhile, (ii) choose actions objective minimize predicted uncertainty. implement this proposition by means event‐predictive learning active inference. implemented mechanism induces inductive, bias, thus developing schematic encodings experienced boundaries. inference principle chooses aiming at minimizing expected future train system on multiple object‐manipulation events. As result, generation shifts emerges while object manipulations: model starts fixating inferred already after having sampled some possible when familiar agent (i.e., hand) involved. Meanwhile, keeps reactively tracking unfamiliar mechanical claw) performing same movement. qualitatively compare these modeling results behavioral data conclude combined may be critical for eliciting behavior in infants.