Hierarchical motor adaptations negotiate failures during force field learning

Belief inference for hierarchical hidden states in spatial navigation

Uncertainty abounds in the real world, and in environments with multiple layers of unobservable hidden states, decision-making requires resolving uncertainties based on mutual inference. Focusing on a spatial navigation problem, we develop a Tiger maze task that involved simultaneously inferring the local hidden state and the global hidden state from probabilistically uncertain observation. We adopt a Bayesian computational approach by proposing a hierarchical inference model. Applying this to human task behaviour, alongside functional magnetic resonance brain imaging, allows us to separate the neural correlates associated with reinforcement and reassessment of belief in hidden states. The imaging results also suggest that different layers of uncertainty differentially involve the basal ganglia and dorsomedial prefrontal cortex, and that the regions responsible are organised along the rostral axis of these areas according to the type of inference and the level of abstraction of the hidden state, i.e. higher-order state inference involves more anterior parts.

Human reaching control in dynamic environments

Closed-loop models of movement control have attracted growing interest in how the nervous system transforms sensory information into motor commands, and several brain structures have been identified as neural substrates for these computational operations. Recently, several studies have focused on how these models need to be updated when environmental parameters change. Current evidence suggests that when the task changes, rapid control updates enable flexible modifications of current actions and online decisions. At the same time, when movement dynamics change, humans use different strategies based on a combination of adaptation and modulation of controller sensitivity to exogenous perturbations (robust control). This review proposes a unified framework to capture these results based on online estimation of model parameters with dynamic updates in control. The reviewed studies also identify the time scales of associated behavioral mechanisms to guide future research on the neural basis of movement control.

Additional cognitive load decreases performance but not adaptation to a visuomotor transformation

Dual-task paradigms are procedures for investigating interference with two tasks performed simultaneously. Studies that previously addressed dual-task paradigms within a visuomotor reaching task yielded mixed results. While some of the studies found evidence of cognitive interference, called dual-task costs, other studies did not. We assume that dual-task costs only manifest themselves within the explicit component of adaptation, as it involves cognitive resources for processing. We suspect the divergent findings to be due to the lack of differentiation between the explicit and implicit component. In this study, we aimed to investigate how a cognitive secondary task affects visuomotor adaptation overall and its different components, both during and after adaptation. In a series of posttests, we examined the explicit and implicit components separately. Eighty participants performed a center-outward reaching movement with a 30° cursor perturbation. Participants were either assigned to a single task group (ST) or a dual-task group (DT) with an additional auditory 1-back task. To further enhance our predicted effect of dual-task interference on the explicit component, we added a visual feedback delay condition to both groups (ST/DT_DEL). In the other condition, participants received visual feedback immediately after movement termination (ST/DT_NoDEL). While there were clear dual-task costs during the practice phase, there were no dual-task effects on any of the posttest measures. On one hand, our findings suggest that dual-task costs in visuomotor adaptation tasks can occur with sufficient cognitive demand, and on the other hand, that cognitive constraints may affect motor performance but not necessarily motor adaptation.

Adaptive Feedback Control in Human Reaching Adaptation to Force Fields

Sensorimotor adaptation is a central function of the nervous system, as it allows humans and other animals to flexibly anticipate their interaction with the environment. In the context of human reaching adaptation to force fields, studies have traditionally separated feedforward (FF) and feedback (FB) processes involved in the improvement of behavior. Here, we review computational models of FF adaptation to force fields and discuss them in light of recent evidence highlighting a clear involvement of feedback control. Instead of a model in which FF and FB mechanisms adapt in parallel, we discuss how online adaptation in the feedback control system can explain both trial-by-trial adaptation and improvements in online motor corrections. Importantly, this computational model combines sensorimotor control and short-term adaptation in a single framework, offering novel perspectives for our understanding of human reaching adaptation and control.

From internal models toward metacognitive AI

In several papers published in Biological Cybernetics in the 1980s and 1990s, Kawato and colleagues proposed computational models explaining how internal models are acquired in the cerebellum. These models were later supported by neurophysiological experiments using monkeys and neuroimaging experiments involving humans. These early studies influenced neuroscience from basic, sensory-motor control to higher cognitive functions. One of the most perplexing enigmas related to internal models is to understand the neural mechanisms that enable animals to learn large-dimensional problems with so few trials. Consciousness and metacognition-the ability to monitor one's own thoughts, may be part of the solution to this enigma. Based on literature reviews of the past 20 years, here we propose a computational neuroscience model of metacognition. The model comprises a modular hierarchical reinforcement-learning architecture of parallel and layered, generative-inverse model pairs. In the prefrontal cortex, a distributed executive network called the "cognitive reality monitoring network" (CRMN) orchestrates conscious involvement of generative-inverse model pairs in perception and action. Based on mismatches between computations by generative and inverse models, as well as reward prediction errors, CRMN computes a "responsibility signal" that gates selection and learning of pairs in perception, action, and reinforcement learning. A high responsibility signal is given to the pairs that best capture the external world, that are competent in movements (small mismatch), and that are capable of reinforcement learning (small reward-prediction error). CRMN selects pairs with higher responsibility signals as objects of metacognition, and consciousness is determined by the entropy of responsibility signals across all pairs. This model could lead to new-generation AI, which exhibits metacognition, consciousness, dimension reduction, selection of modules and corresponding representations, and learning from small samples. It may also lead to the development of a new scientific paradigm that enables the causal study of consciousness by combining CRMN and decoded neurofeedback.

Hierarchical motor adaptations negotiate failures during force field learning

Humans have the amazing ability to learn the dynamics of the body and environment to develop motor skills. Traditional motor studies using arm reaching paradigms have viewed this ability as the process of ‘internal model adaptation’. However, the behaviors have not been fully explored in the case when reaches fail to attain the intended target. Here we examined human reaching under two force fields types; one that induces failures (i.e., target errors), and the other that does not. Our results show the presence of a distinct failure-driven adaptation process that enables quick task success after failures, and before completion of internal model adaptation, but that can result in persistent changes to the undisturbed trajectory. These behaviors can be explained by considering a hierarchical interaction between internal model adaptation and the failure-driven adaptation of reach direction. Our findings suggest that movement failure is negotiated using hierarchical motor adaptations by humans.

How do we improve actions after a movement failure? Although negotiating movement failures is obviously crucial, previous motor-control studies have predominantly examined human movement adaptations in the absence of failures, and it remains unclear how failures affect subsequent movement adaptations. Here we examined this issue by developing a novel force field adaptation task where the hand movement during an arm reaching is perturbed by novel forces that induce a large target error, that is a failure. Our experimental observation and computational modeling show that, in addition to the popular ‘internal model learning’ process of motor adaptations, humans also utilize a ‘failure-negotiating’ process, that enables them to quickly improve movements in the presence of failure, even at the expense of increased arm trajectory deflections, which are subsequently reduced gradually with training after the achievement of the task success. Our results suggest that a hierarchical interaction between these two processes is a key for humans to negotiate movement failures.