A key feature of sequential decision making under uncertainty is a need to balance between exploiting-choosing the best action according to the current knowledge, and exploring-obtaining information about values of other actions. The multi-armed bandit problem, a classical task that captures this trade-off, served as a vehicle in machine learning for developing bandit algorithms that proved to be useful in numerous industrial applications. The active inference framework, an approach to sequential decision making recently developed in neuroscience for understanding human and animal behaviour, is distinguished by its sophisticated strategy for resolving the exploration-exploitation trade-off. This makes active inference an exciting alternative to already established bandit algorithms. Here we derive an efficient and scalable approximate active inference algorithm and compare it to two state-of-the-art bandit algorithms: Bayesian upper confidence bound and optimistic Thompson sampling. This comparison is done on two types of bandit problems: a stationary and a dynamic switching bandit. Our empirical evaluation shows that the active inference algorithm does not produce efficient long-term behaviour in stationary bandits. However, in the more challenging switching bandit problem active inference performs substantially better than the two state-of-the-art bandit algorithms. The results open exciting venues for further research in theoretical and applied machine learning, as well as lend additional credibility to active inference as a general framework for studying human and animal behaviour.
1 October 2020 in Journal of Experimental Psychology: General
How do people decide whether to try out novel options as opposed to tried-and-tested ones? We argue that they infer a novel option's reward from contextual information learned from functional relations and take uncertainty into account when making a decision. We propose a Bayesian optimization model to describe their learning and decision making. This model relies on similarity-based learning of functional relationships between features and rewards, and a choice rule that balances exploration and exploitation by combining predicted rewards and the uncertainty of these predictions. Our model makes 2 main predictions. First, decision makers who learn functional relationships will generalize based on the learned reward function, choosing novel options only if their predicted reward is high. Second, they will take uncertainty about the function into account, and prefer novel options that can reduce this uncertainty. We test these predictions in 3 preregistered experiments in which we examine participants' preferences for novel options using a feature-based multiarmed bandit task in which rewards are a noisy function of observable features. Our results reveal strong evidence for functional exploration and moderate evidence for uncertainty-guided exploration. However, whether or not participants chose a novel option also depended on their attention, as well as reflecting on the value of the options. These results advance our understanding of people's reactions in the face of novelty. (PsycInfo Database Record (c) 2020 APA, all rights reserved).
Gender effects in risk taking have attracted much attention by economists, and remain debated. Loss aversion—the stylized finding that a given loss carries substantially greater weight than a monetarily equivalent gain—is a fundamental driver of risk aversion. We deploy four definitions of loss aversion commonly used in the literature to investigate gender effects. Even though the definitions only differ in subtle ways, we find women to be more loss averse than men according to one definition, while another definition results in no gender differences, and the remaining two definitions point to women beinglessloss averse than men. Conceptually, these contradictory effects can be organized by systematic measurement error resulting from model mis-specifications relative to the true underlying decision process.
24 January 2020 in Proceedings of the National Academy of Sciences
Uncertainty plays a critical role in reinforcement learning and decision making. However, exactly how it influences behavior remains unclear. Multiarmed-bandit tasks offer an ideal test bed, since computational tools such as approximate Kalman filters can closely characterize the interplay between trial-by-trial values, uncertainty, learning, and choice. To gain additional insight into learning and choice processes, we obtained data from subjects’ overt allocation of gaze. The estimated value and estimation uncertainty of options influenced what subjects looked at before choosing; these same quantities also influenced choice, as additionally did fixation itself. A momentary measure of uncertainty in the form of absolute prediction errors determined how long participants looked at the obtained outcomes. These findings affirm the importance of uncertainty in multiple facets of behavior and help delineate its effects on decision making.
Functional magnetic resonance imaging (fMRI) is a standard tool to investigate the neural correlates of cognition. fMRI noninvasively measures brain activity, allowing identification of patterns evoked by tasks performed during scanning. Despite the long history of this technique, the idiosyncrasies of each dataset have led to the use of ad-hoc preprocessing protocols customized for nearly every different study. This approach is time-consuming, error-prone, and unsuitable for combining datasets from many sources. Here we showcase fMRIPrep ( http://fmriprep.org ), a robust tool to prepare human fMRI data for statistical analysis. This software instrument addresses the reproducibility concerns of the established protocols for fMRI preprocessing. By leveraging the Brain Imaging Data Structure (BIDS) to standardize both the input datasets —MRI data as stored by the scanner— and the outputs —data ready for modeling and analysis—, fMRIPrep is capable of preprocessing a diversity of datasets without manual intervention. In support of the growing popularity of fMRIPrep , this protocol describes how to integrate the tool in a task-based fMRI investigation workflow.
1 December 2023 in The Annals of Applied Probability
We study a discrete-time Markov process Xn∈Rd for which the distribution of the future increments depends only on the relative ranking of its components (descending order by value). We endow the process with a rich-get-richer assumption and show that, together with a finite second moments assumption, it is enough to guarantee almost sure convergence of Xn/n. We characterize the possible limits if one is free to choose the initial state and we give a condition under which the initial state is irrelevant. Finally, we show how our framework can account for ranking-based Pólya urns and can be used to study ranking algorithms for web interfaces.