How we perceive, attend to, or remember the stimuli in our environment depends on our preferences for them. Here we argue that this dependence is reciprocal: pleasures and displeasures are heavily dependent on cognitive processing, namely, on our ability to predict the world correctly. We propose that prediction errors, inversely weighted with prior probabilities of predictions, yield subjective experiences of positive or negative affect. In this way, we link affect to predictions within a predictive coding framework. We discuss how three key factors - uncertainty, expectations, and conflict - influence prediction accuracy and show how they shape our affective response. We demonstrate that predictable stimuli are, in general, preferred to unpredictable ones, though too much predictability may decrease this liking effect. Furthermore, the account successfully overcomes the "dark-room" problem, explaining why we do not avoid stimulation to minimize prediction error. We further discuss the implications of our approach for art perception and the utility of affect as feedback for predictions within a prediction-testing architecture of cognition.
Bibliographical noteFunding Information:
This manuscript was improved by comments from and conversations with Sander Van de Cruys, Maria Kuvaldina, Jörgen Pind, Heida Maria Sigurdardottir. Andrey Chetverikov was supported by Russian Foundation for Basic Research (#15-06-07417 A) and by Saint Petersburg State University (research grant #8.38.287.2014). Árni Kristjánsson receives support from the European Research Council, The Icelandic Centre for Research (Rannis) and the Research Fund of the University of Iceland.
© 2016 Elsevier B.V.
- Predictive coding