Ventral pallidum response to initial confirmation of bad advise predicts subsequent impediments in reinforcement learning

Bias to search for information that confirms one’s prior expectations and to dismiss hypothesis-incongruent information (until sufficient amount of contradictory information accumulates and forces one to abandon one’s initial hypothesis) is one of the most fundamental aspects of human information processing. As a specific example of this, it has been shown that after receiving initial expert advise, persons tend to be less likely to seek out alternative actions and, in case of bad advice, end up with less optimal outcomes. While this bias, known as the law of primacy in persuasion (i.e., out of two opposing arguments, the first one has a stronger effect), has been behaviorally demonstrated already in the 1920s, relatively little is known about the neural basis of learning based on bias due to initial advise by others and the possible role of initial confirmatory evidence in strengthening the bias.

In their recent study, Staudinger et al. (2013) had 35 right-handed volunteers carry out a probabilistic reinforcement-learning task on Japanese Hiragana symbols during functional magnetic resonance imaging, wherein the subjects received, prior to task onset, a misleading tip that two of the symbols would have the highest probability of being correct. The first trials were presented to the subjects non-randomly, providing two out of three advise-confirmatory trials for one of the advised symbols, and two out of three advise-contradictory trials for the other advised symbol. After these initial trials the task sequence became random.  After the scanning, the subjects completed a lottery task where the probability of winning 8€ was specified by the displayed symbol that the subjects were learning during their task in the fMRI scanner.

The results show that one-time misleading advise robustly biases subsequent learning and decision making; the initial advise-confirmatory trials strengthened the biasing effects of the misleading advise on learning. Furthermore, interestingly, stronger ventral pallidum responses to initial positive reinforcement of misleading advise predicted worse learning over the remainder of the experiment. Given that the ventral pallidum is part of the brain’s reinforcement-learning circuitry, these results are highly important in demonstrating how receiving a priori advise shapes the way that the reward circuitry processes subsequent advise-associated information, and thus biases the learning process.

Reference: Staudinger MR, Büchel C. How initial confirmatory experience potentiates the detrimental influence of bad advice. Neuroimage (2013) e-publication ahead of print. http://dx.doi.org/10.1016/j.neuroimage.2013.02.074