Surprise and novelty in the brain

Predictive coding in the human olfactory system

The human olfactory system is unusual. It deviates from the classical structure and function of other sensory cortices, and many of its basic computations remain mysterious. These idiosyncrasies have challenged the development of a clear and comprehensive theoretical framework in olfactory neuroscience. To address this challenge, we develop a theory of olfactory predictive coding that aims to unify diverse olfactory phenomena. Under this scheme, the olfactory system is not merely passively processing sensory information. Instead, it is actively issuing predictions about sensory inputs before they even arrive. We map this conceptual framework onto the micro- and macroscale neurobiology of the human olfactory system and review a variety of neurobiological, computational, and behavioral evidence in support of this scheme.

Exploration in 4-year-old children is guided by learning progress and novelty

Humans are driven by an intrinsic motivation to learn, but the developmental origins of curiosity-driven exploration remain unclear. We investigated the computational principles guiding 4-year-old children's exploration during a touchscreen game (N = 102, F = 49, M = 53, primarily white and middle-class, data collected in the Netherlands from 2021-2023). Children guessed the location of characters that were hiding following predictable (yet noisy) patterns. Children could freely switch characters, which allowed us to quantify when they decided to explore something different and what they chose to explore. Bayesian modeling of their responses revealed that children selected activities that were more novel and offered greater learning progress (LP). Moreover, children's interest in making LP correlated with better learning performance. These findings highlight the importance of novelty and LP in guiding children's exploration.

Synapses learn to utilize stochastic pre-synaptic release for the prediction of postsynaptic dynamics

Synapses in the brain are highly noisy, which leads to a large trial-by-trial variability. Given how costly synapses are in terms of energy consumption these high levels of noise are surprising. Here we propose that synapses use noise to represent uncertainties about the somatic activity of the postsynaptic neuron. To show this, we developed a mathematical framework, in which the synapse as a whole interacts with the soma of the postsynaptic neuron in a similar way to an agent that is situated and behaves in an uncertain, dynamic environment. This framework suggests that synapses use an implicit internal model of the somatic membrane dynamics that is being updated by a synaptic learning rule, which resembles experimentally well-established LTP/LTD mechanisms. In addition, this approach entails that a synapse utilizes its inherently noisy synaptic release to also encode its uncertainty about the state of the somatic potential. Although each synapse strives for predicting the somatic dynamics of its postsynaptic neuron, we show that the emergent dynamics of many synapses in a neuronal network resolve different learning problems such as pattern classification or closed-loop control in a dynamic environment. Hereby, synapses coordinate themselves to represent and utilize uncertainties on the network level in behaviorally ambiguous situations.

Surprising sounds influence risky decision making

Adaptive behavior depends on appropriate responses to environmental uncertainty. Incidental sensory events might simply be distracting and increase errors, but alternatively can lead to stereotyped responses despite their irrelevance. To evaluate these possibilities, we test whether task-irrelevant sensory prediction errors influence risky decision making in humans across seven experiments (total n = 1600). Rare auditory sequences preceding option presentation systematically increase risk taking and decrease choice perseveration (i.e., increased tendency to switch away from previously chosen options). The risk-taking and perseveration effects are dissociable by manipulating auditory statistics: when rare sequences end on standard tones, including when rare sequences consist only of standard tones, participants are less likely to perseverate after rare sequences but not more likely to take risks. Computational modeling reveals that these effects cannot be explained by increased decision noise but can be explained by value-independent risky bias and perseveration parameters, decision biases previously linked to dopamine. Control experiments demonstrate that both surprise effects can be eliminated when tone sequences are presented in a balanced or fully predictable manner, and that surprise effects cannot be explained by erroneous beliefs. These findings suggest that incidental sounds may influence many of the decisions we make in daily life.

A shared novelty-seeking basis for creativity and curiosity: Response to the commentators

In our target article, we proposed that curiosity and creativity are both manifestations of the same novelty-seeking process. We received 29 commentaries from diverse disciplines that add insights to our initial proposal. These commentaries ultimately expanded and supplemented our model. Here we draw attention to five central practical and theoretical issues that were raised by the commentators: (1) The complex construct of novelty and associated concepts; (2) the underlying subsystems and possible mechanisms; (3) the different pathways and subtypes of curiosity and creativity; (4) creativity and curiosity "in the wild"; (5) the possible link(s) between creativity and curiosity.

Computational models of intrinsic motivation for curiosity and creativity

We link Ivancovsky et al.'s novelty-seeking model (NSM) to computational models of intrinsically motivated behavior and learning. We argue that dissociating different forms of curiosity, creativity, and memory based on the involvement of distinct intrinsic motivations (e.g., surprise and novelty) is essential to empirically test the conceptual claims of the NSM.

Curiosity-driven exploration: foundations in neuroscience and computational modeling

Curiosity refers to the intrinsic desire of humans and animals to explore the unknown, even when there is no apparent reason to do so. Thus far, no single, widely accepted definition or framework for curiosity has emerged, but there is growing consensus that curious behavior is not goal-directed but related to seeking or reacting to information. In this review, we take a phenomenological approach and group behavioral and neurophysiological studies which meet these criteria into three categories according to the type of information seeking observed. We then review recent computational models of curiosity from the field of machine learning and discuss how they enable integrating different types of information seeking into one theoretical framework. Combinations of behavioral and neurophysiological studies along with computational modeling will be instrumental in demystifying the notion of curiosity.