Our second line of research focuses on understanding the role of cognitive control in learning. Although human equipped with powerful reinforcement learning mechanisms that can quickly learn new associations via trial and error, these mechanisms alone is not sufficient for human to survive in a changing environment. In real-life learning, the feedback is more probabilistic than deterministic and the reward contingency changes with environments, adaptive behaviors thus require the capacity to quickly overcome prepotent responses and relearn new ones. We conducted a series of studies to examine the neural mechanisms of cognitive control and behavioral flexibility. In one study, we found that the ventral medial prefrontal cortex and ventral striatum were involvement in reward processing, the dorsomedial prefrontal cortex is involved in processing the uncertainty of reward (Xue et al., 2009, Cerebral Cortex). The right ventrolateral prefrontal cortex (VLPFC) plays an important role in response inhibition (Xue et al., 2008, Cerebral Cortex) and reversal learning (Xue et al., 2008, Journal of Neuroscience), serving as the neural basis for behavioral flexibility. In a most recent study, we found that the right DLPFC and VLPFC are commonly involved in reversal learning by both monetary reward and punishment (i.e., mild electric shock) (Wang et al.,2013,Journal of Neuroscience). These results emphasized the role of frontal cognitive control systems in adaptive learning and behavioral flexibility.
Intriguingly, we found the same frontal cognitive control system that is important for adaptive learning could also contribute to maladaptive learning (e.g., the gambler’s fallacy, GF), under certain situation where no consistent stimulus-reward contingency exists (Xue et al., 2012, PNAS). Corroborating this evidence, our large sample (N=438) study further indicates that individuals with higher intelligence and better cognitive control capacity are more likely to engage the GF (Xue et al., 2012, Plos One). We propose that the lateral frontal cortex cognitive control system can be “hijacked” by a false “world model” and contribute to maladaptive learning. By changing the “world model” through agency manipulation (Xue et al., 2013, Plos One), the LPFC can both contribute to adaptive and maladaptive decisions, depending on the “world model” that guide behaviors.
Cognitive control and adaptive learning. (A) Theorectic model: The role of cognitive control system in learning is modulated by the “world model”, when being “hijacked” by a false world model, the frontal cognitive control system can contributed to maladaptiave decision (Xue et al., 2012; Plos One)；(B) functional MRI revealed that the LIFG activity could predict decision making 10s later: A higher LIFG activity was associated with more gambler’s fallacy. (C) Anodal direct current stimulation over the LIFG, as compared to the control site, V1, could enhance LIFG function and also increase the gambler’s fallacy. (Source: Xue et al., 2012, PNAS).