Brian D. Glass

When More Is Less: Feedback Effects in Perceptual Category Learning.

Rule-based and information-integration category learning were compared under minimal and full feedback conditions. Rule-based category structures are those for which the optimal rule is verbalizable. Information-integration category structures are those for which the optimal rule is not verbalizable. With minimal feedback subjects are told whether their response was correct or incorrect, but are not informed of the correct category assignment. With full feedback subjects are informed of the correctness of their response and are also informed of the correct category assignment. An examination of the distinct neural circuits that subserve rule-based and information-integration category learning leads to the counterintuitive prediction that full feedback should facilitate rule-based learning but should also hinder information-integration learning. This prediction was supported in the experiment reported below. The implications of these results for theories of learning are discussed.

Real-Time Strategy Game Training: Emergence of a Cognitive Flexibility Trait

Training in action video games can increase the speed of perceptual processing. However, it is unknown whether video-game training can lead to broad-based changes in higher-level competencies such as cognitive flexibility, a core and neurally distributed component of cognition. To determine whether video gaming can enhance cognitive flexibility and, if so, why these changes occur, the current study compares two versions of a real-time strategy (RTS) game. Using a meta-analytic Bayes factor approach, we found that the gaming condition that emphasized maintenance and rapid switching between multiple information and action sources led to a large increase in cognitive flexibility as measured by a wide array of non-video gaming tasks. Theoretically, the results suggest that the distributed brain networks supporting cognitive flexibility can be tuned by engrossing video game experience that stresses maintenance and rapid manipulation of multiple information sources. Practically, these results suggest avenues for increasing cognitive function.

How Humans Teach Agents

Human beings are a largely untapped source of in-the-loop knowledge and guidance for computational learning agents, including robots. To effectively design agents that leverage available human expertise, we need to understand how people naturally teach. In this paper, we describe two experiments that ask how differing conditions affect a human teacher’s feedback frequency and the computational agent’s learned performance. The first experiment considers the impact of a self-perceived teaching role in contrast to believing one is merely critiquing a recording. The second considers whether a human trainer will give more frequent feedback if the agent acts less greedily (i.e., choosing actions believed to be worse) when the trainer’s recent feedback frequency decreases. From the results of these experiments, we draw three main conclusions that inform the design of agents. More broadly, these two studies stand as early examples of a nascent technique of using agents as highly specifiable social entities in experiments on human behavior.

Category label and response location shifts in category learning

The category shift literature suggests that rule-based classification, an important form of explicit learning, is mediated by two separate learned associations: a stimulus-to-label association that associates stimuli and category labels, and a label-to-response association that associates category labels and responses. Three experiments investigate whether information–integration classification, an important form of implicit learning, is also mediated by two separate learned associations. Participants were trained on a rule-based or an information–integration categorization task and then the association between stimulus and category label, or between category label and response location was altered. For rule-based categories, and in line with previous research, breaking the association between stimulus and category label caused more interference than breaking the association between category label and response location. However, no differences in recovery rate emerged. For information–integration categories, breaking the association between stimulus and category label caused more interference and led to greater recovery than breaking the association between category label and response location. These results provide evidence that information–integration category learning is mediated by separate stimulus-to-label and label-to-response associations. Implications for the neurobiological basis of these two learned associations are discussed.

Normal aging and the dissociable prototype learning systems.

Dissociable prototype learning systems have been demonstrated behaviorally and with neuroimaging in younger adults as well as with patient populations. In A/not-A (AN) prototype learning, participants are shown members of category A during training, and during test are asked to decide whether novel items are in category A or are not in category A. Research suggests that AN learning is mediated by a perceptual learning system. In A/B (AB) prototype learning, participants are shown members of category A and B during training, and during test are asked to decide whether novel items are in category A or category B. In contrast to AN, research suggests that AB learning is mediated by a declarative memory system. The current study examined the effects of normal aging on AN and AB prototype learning. We observed an age-related deficit in AB learning, but an age-related advantage in AN learning. Computational modeling supports one possible interpretation based on narrower selective attentional focus in older adults in the AB task and broader selective attention in the AN task. Neuropsychological testing in older participants suggested that executive functioning and attentional control were associated with better performance in both tasks. However, nonverbal memory was associated with better AN performance, while visual attention was associated with worse AB performance. The results support an interactive memory systems approach and suggest that age-related declines in one memory system can lead to deficits in some tasks, but to enhanced performance in others.

Regulatory Match Effects on a Modified Wisconsin Card Sort Task

The Wisconsin Card Sorting Task (WCST; Heaton, 1980) is commonly used to assess concept formation and set shifting. Cognitive research suggests that set shifting performance is enhanced by a match between a persons regulatory focus (promotion focus: attempting to earn an entry into a cash drawing; prevention focus: attempting to avoid losing an entry into the drawing) and the task reward structure (gains: attempting to maximize points gained; losses: attempting to minimize points lost). A regulatory match results when attempting to earn an entry by maximizing points or attempting to avoid losing an entry by minimizing losses. We test the hypothesis that performance on a modified WCST is accentuated in younger, healthy participants when there is a match between the global performance incentive and the local task reward structure. As predicted, participants in a match showed better set shifting but equivalent initial concept formation when compared with participants in a mismatch. Furthermore, relative to a baseline control group, mismatch participants were significantly worse at set shifting than were participants in a regulatory match. These results suggest that set shifting performance might be impacted by incentive and task reward factors in ways that have not been considered previously.

The effects of 24-hour sleep deprivation on the exploration-exploitation trade-off

Sleep deprivation has a complex set of neurological effects that go beyond a mere slowing of mental processes. While cognitive and perceptual impairments in sleep deprived individuals are widespread, some abilities remain intact. In an effort to characterize these effects, some have suggested an impairment of complex decision-making ability despite intact ability to follow simple rules. To examine this trade-off, 24-hour total sleep deprived individuals performed two versions of a resource acquisition foraging task, one in which exploration is optimal (to succeed, abandon low value, high saliency options) and another in which exploitation is optimal (to succeed, refrain from switching between options). Sleep deprived subjects exhibited decreased performance on the exploitation task compared to non-sleep deprived controls, yet both groups exhibited increased performance on the exploratory task. These results speak to previous neuropsychological work on cognitive control.

Attenuating Age-Related Learning Deficits: Emotional Valenced Feedback Interacts with Task Complexity

Previous research reveals that older adults sometimes show enhanced processing of emotionally positive stimuli relative to negative stimuli, but that this positivity bias reverses to become a negativity bias when cognitive control resources are less available. In this study, we test the hypothesis that emotionally positive feedback will attenuate well-established age-related deficits in rule learning whereas emotionally negative feedback will amplify age deficits-but that this pattern will reverse when the task involves a high cognitive load. Experiment 1 used emotional face feedback and revealed an interaction among age, valence of the feedback, and task load. When the task placed minimal load on cognitive control resources, happy-face feedback attenuated age-related deficits in initial rule learning and angry-face feedback led to age-related deficits in initial rule learning and set shifting. However, when the task placed a high load on cognitive control resources, we found that angry-face feedback attenuated age-related deficits in initial rule learning and set shifting whereas happy-face feedback led to age-related deficits in initial rule learning and set shifting. Experiment 2 used less emotional point feedback and revealed age-related deficits in initial rule learning and set shifting under low and high cognitive load for point-gain and point-loss conditions. The research presented here demonstrates that emotional feedback can attenuate age-related learning deficits-but only positive feedback for tasks with a low cognitive load and negative feedback for tasks with high cognitive load.

Regulatory fit effects on stimulus identification

This article examines the effects of a fit between a person’s global regulatory focus and the local task reward structure on perceptual processing and judgment. On each trial, participants were presented with one of two briefly presented stimuli and were asked to identify it. Participants were placed in a promotion focus (a situationally induced sensitivity to gains) or a prevention focus (a situationally induced sensitivity to losses) and were asked to maximize gains or minimize losses. An asymmetric payoff ratio biased the overall reward toward one identification response over the other. Two experiments tested the role of regulatory fit when internal familiarity and perceptual sensitivity were low or high. When familiarity and sensitivity were low, participants in a regulatory fit (promotion focus with gains or a prevention focus with losses) showed greater perceptual sensitivity but no response bias differences, relative to participants in a regulatory mismatch. When familiarity and sensitivity were high, participants in a regulatory fit showed a response bias toward the high-payoff stimulus but no differences in perceptual sensitivity. Speculations are offered on the neurobiological basis of this effect, as well as implications of this work for clinical disorders such as depression.

Approaches to Learning to Control Dynamic Uncertainty

In dynamic environments, when faced with a choice of which learning strategy to adopt, do people choose to mostly explore (maximizing their long term gains) or exploit (maximizing their short term gains)? More to the point, how does this choice of learning strategy influence one’s later ability to control the environment? In the present study, we explore whether people’s self-reported learning strategies and levels of arousal (i.e., surprise, stress) correspond to performance measures of controlling a Highly Uncertain or Moderately Uncertain dynamic environment. Generally, self-reports suggest a preference for exploring the environment to begin with. After which, those in the Highly Uncertain environment generally indicated they exploited more than those in the Moderately Uncertain environment; this difference did not impact on performance on later tests of people’s ability to control the dynamic environment. Levels of arousal were also differentially associated with the uncertainty of the environment. Going beyond behavioral data, our model of dynamic decision-making revealed that, in actual fact, there was no difference in exploitation levels between those in the highly uncertain or moderately uncertain environments, but there were differences based on sensitivity to negative reinforcement. We consider the implications of our findings with respect to learning and strategic approaches to controlling dynamic uncertainty.