Intact Reinforcement Learning But Impaired Attentional Control During Multidimensional Probabilistic Learning in Older Adults.

To efficiently learn optimal behavior in complex environments, humans rely on an interplay of learning and attention. Healthy aging has been shown to independently affect both of these functions. Here, we investigate how reinforcement learning and selective attention interact during learning from trial and error across age groups. We acquired behavioral and fMRI data from older and younger adults (male and female) performing two probabilistic learning tasks with varying attention demands. Although learning in the unidimensional task did not differ across age groups, older adults performed worse than younger adults in the multidimensional task, which required high levels of selective attention. Computational modeling showed that choices of older adults are better predicted by reinforcement learning than Bayesian inference, and that older adults rely more on reinforcement learning-based predictions than younger adults. Conversely, a higher proportion of younger adults' choices was predicted by a computationally demanding Bayesian approach. In line with the behavioral findings, we observed no group differences in reinforcement-learning related fMRI activation. Specifically, prediction-error activation in the nucleus accumbens was similar across age groups, and numerically higher in older adults. However, activation in the default mode was less suppressed in older adults in for higher attentional task demands, and the level of suppression correlated with behavioral performance. Our results indicate that healthy aging does not significantly impair simple reinforcement learning. However, in complex environments, older adults rely more heavily on suboptimal reinforcement-learning strategies supported by the ventral striatum, whereas younger adults use attention processes supported by cortical networks.SIGNIFICANCE STATEMENT Changes in the way that healthy human aging affects how we learn to optimally behave are not well understood; it has been suggested that age-related declines in dopaminergic function may impair older adult's ability to learn from reinforcement. In the present fMRI experiment, we show that learning and nucleus accumbens activation in a simple unidimensional reinforcement-learning task was not significantly affected by age. However, in a more complex multidimensional task, older adults showed worse performance and relied more on reinforcement-learning strategies than younger adults, while failing to disengage their default-mode network during learning. These results imply that older adults are only impaired in reinforcement learning if they additionally need to learn which dimensions of the environment are currently important.

Reinforcement learning in multidimensional environments relies on attention mechanisms.

In recent years, ideas from the computational field of reinforcement learning have revolutionized the study of learning in the brain, famously providing new, precise theories of how dopamine affects learning in the basal ganglia. However, reinforcement learning algorithms are notorious for not scaling well to multidimensional environments, as is required for real-world learning. We hypothesized that the brain naturally reduces the dimensionality of real-world problems to only those dimensions that are relevant to predicting reward, and conducted an experiment to assess by what algorithms and with what neural mechanisms this "representation learning" process is realized in humans. Our results suggest that a bilateral attentional control network comprising the intraparietal sulcus, precuneus, and dorsolateral prefrontal cortex is involved in selecting what dimensions are relevant to the task at hand, effectively updating the task representation through trial and error. In this way, cortical attention mechanisms interact with learning in the basal ganglia to solve the "curse of dimensionality" in reinforcement learning.

A universal role of the ventral striatum in reward-based learning: evidence from human studies.

Reinforcement learning enables organisms to adjust their behavior in order to maximize rewards. Electrophysiological recordings of dopaminergic midbrain neurons have shown that they code the difference between actual and predicted rewards, i.e., the reward prediction error, in many species. This error signal is conveyed to both the striatum and cortical areas and is thought to play a central role in learning to optimize behavior. However, in human daily life rewards are diverse and often only indirect feedback is available. Here we explore the range of rewards that are processed by the dopaminergic system in human participants, and examine whether it is also involved in learning in the absence of explicit rewards. While results from electrophysiological recordings in humans are sparse, evidence linking dopaminergic activity to the metabolic signal recorded from the midbrain and striatum with functional magnetic resonance imaging (fMRI) is available. Results from fMRI studies suggest that the human ventral striatum (VS) receives valuation information for a diverse set of rewarding stimuli. These range from simple primary reinforcers such as juice rewards over abstract social rewards to internally generated signals on perceived correctness, suggesting that the VS is involved in learning from trial-and-error irrespective of the specific nature of provided rewards. In addition, we summarize evidence that the VS can also be implicated when learning from observing others, and in tasks that go beyond simple stimulus-action-outcome learning, indicating that the reward system is also recruited in more complex learning tasks.

Striatal activations signal prediction errors on confidence in the absence of external feedback.

Research on the neural bases of learning has mainly focused on reinforcement learning where the central role of the dopaminergic system is well established. However, in everyday life many decisions are not followed by feedback, in which case humans have been shown to code the most probable outcome into memory. We used functional magnetic resonance imaging (fMRI) to examine the neural basis of internally generated signals on correctness and decision confidence in the complete absence of feedback in a categorization task. During test trials after observational training activation in dopaminergic target regions was modulated by the correctness of the answer similarly as during feedback-based training. Moreover, activation in the nucleus accumbens and putamen was correlated with the prediction error on confidence as estimated by a reinforcement learning model. In this model subjective confidence ratings acquired after each trial served as outcome measure. Activation in the striatum therefore follows a similar pattern in response to prediction errors on confidence as it does during reinforcement learning in response to reward prediction errors, but with respect to internally generated signals based on knowledge of the structure of the environment. Furthermore, ventral striatal activation decreased with stimulus novelty, which might support the allocation of attention to unfamiliar stimuli. These results provide a parsimonious account for the neural bases of learning, indicating overlapping neural substrates of reinforcement learning and learning when outcome information has to be internally constructed.

Comparing the neural basis of monetary reward and cognitive feedback during information-integration category learning.

The dopaminergic system is known to play a central role in reward-based learning (Schultz, 2006), yet it was also observed to be involved when only cognitive feedback is given (Aron et al., 2004). Within the domain of information-integration category learning, in which information from several stimulus dimensions has to be integrated predecisionally (Ashby and Maddox, 2005), the importance of contingent feedback is well established (Maddox et al., 2003). We examined the common neural correlates of reward anticipation and prediction error in this task. Sixteen subjects performed two parallel information-integration tasks within a single event-related functional magnetic resonance imaging session but received a monetary reward only for one of them. Similar functional areas including basal ganglia structures were activated in both task versions. In contrast, a single structure, the nucleus accumbens, showed higher activation during monetary reward anticipation compared with the anticipation of cognitive feedback in information-integration learning. Additionally, this activation was predicted by measures of intrinsic motivation in the cognitive feedback task and by measures of extrinsic motivation in the rewarded task. Our results indicate that, although all other structures implicated in category learning are not significantly affected by altering the type of reward, the nucleus accumbens responds to the positive incentive properties of an expected reward depending on the specific type of the reward.

Assessing the neural basis of uncertainty in perceptual category learning through varying levels of distortion.

The formation of new perceptual categories involves learning to extract that information from a wide range of often noisy sensory inputs, which is critical for selecting between a limited number of responses. To identify brain regions involved in visual classification learning under noisy conditions, we developed a task on the basis of the classical dot pattern prototype distortion task [M. I. Posner, Journal of Experimental Psychology, 68, 113-118, 1964]. Twenty-seven healthy young adults were required to assign distorted patterns of dots into one of two categories, each defined by its prototype. Categorization uncertainty was modulated parametrically by means of Shannon's entropy formula and set to the levels of 3, 7, and 8.5 bits/dot within subsets of the stimuli. Feedback was presented after each trial, and two parallel versions of the task were developed to contrast practiced and unpracticed performance within a single session. Using event-related fMRI, areas showing increasing activation with categorization uncertainty and decreasing activation with training were identified. Both networks largely overlapped and included areas involved in visuospatial processing (inferior temporal and posterior parietal areas), areas involved in cognitive processes requiring a high amount of cognitive control (posterior medial wall), and a cortico-striatal-thalamic loop through the body of the caudate nucleus. Activity in the medial prefrontal wall was increased when subjects received negative as compared with positive feedback, providing further evidence for its important role in mediating the error signal. This study characterizes the cortico-striatal network underlying the classification of distorted visual patterns that is directly related to decision uncertainty.

Dynamic Interaction between Reinforcement Learning and Attention in Multidimensional Environments.

Little is known about the relationship between attention and learning during decision making. Using eye tracking and multivariate pattern analysis of fMRI data, we measured participants' dimensional attention as they performed a trial-and-error learning task in which only one of three stimulus dimensions was relevant for reward at any given time. Analysis of participants' choices revealed that attention biased both value computation during choice and value update during learning. Value signals in the ventromedial prefrontal cortex and prediction errors in the striatum were similarly biased by attention. In turn, participants' focus of attention was dynamically modulated by ongoing learning. Attentional switches across dimensions correlated with activity in a frontoparietal attention network, which showed enhanced connectivity with the ventromedial prefrontal cortex between switches. Our results suggest a bidirectional interaction between attention and learning: attention constrains learning to relevant dimensions of the environment, while we learn what to attend to via trial and error.

The effects of aging on the interaction between reinforcement learning and attention.

Reinforcement learning (RL) in complex environments relies on selective attention to uncover those aspects of the environment that are most predictive of reward. Whereas previous work has focused on age-related changes in RL, it is not known whether older adults learn differently from younger adults when selective attention is required. In 2 experiments, we examined how aging affects the interaction between RL and selective attention. Younger and older adults performed a learning task in which only 1 stimulus dimension was relevant to predicting reward, and within it, 1 "target" feature was the most rewarding. Participants had to discover this target feature through trial and error. In Experiment 1, stimuli varied on 1 or 3 dimensions and participants received hints that revealed the target feature, the relevant dimension, or gave no information. Group-related differences in accuracy and RTs differed systematically as a function of the number of dimensions and the type of hint available. In Experiment 2 we used trial-by-trial computational modeling of the learning process to test for age-related differences in learning strategies. Behavior of both young and older adults was explained well by a reinforcement-learning model that uses selective attention to constrain learning. However, the model suggested that older adults restricted their learning to fewer features, employing more focused attention than younger adults. Furthermore, this difference in strategy predicted age-related deficits in accuracy. We discuss these results suggesting that a narrower filter of attention may reflect an adaptation to the reduced capabilities of the reinforcement learning system. (PsycINFO Database Record