A theory of learning to infer

Dasgupta, Ishita; Schulz, Eric; Tenenbaum, Joshua B.; Gershman, Samuel J.

doi:10.1101/644534

Cited by 23 publications

(28 citation statements)

References 148 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…More relevant to our work is the approach of Dasgupta et al (2020), who taught neural networks to approximate Bayesian inference, given some information about an inference problem's prior and likelihood. Restricting the size of the network allowed them to account for a large amount of cognitive biases, including base rate neglect and conservatism.…”

Section: Resource Rationalitymentioning

confidence: 99%

“…This approach shares its core principles with our theory: resource rationality and meta-learning. However, BMLI does not approximate Bayesian inference explicitly as done by Dasgupta et al (2020). Instead, it attempts to infer distributions that are optimal for making future predictions (which may or may not correspond to Bayesian inference).…”

Section: Resource Rationalitymentioning

confidence: 99%

See 1 more Smart Citation

Heuristics From Bounded Meta-Learned Inference

Binz¹,

Gershman²,

Schulz

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Numerous researchers have put forward heuristics as models of human decision making. However, where such heuristics come from is still a topic of ongoing debates. In this work we propose a novel computational model that advances our understanding of heuristic decision making by explaining how different heuristics are discovered and how they are selected. This model, called bounded meta-learned inference, is based on the idea that people make environment-specific inferences about which strategies to use, while being efficient in terms of how they use computational resources. We show that our approach discovers two previously suggested types of heuristics -- one reason decision making and equal weighting -- in specific environments. Furthermore, the model provides clear and precise predictions about when each heuristic should be applied: knowing the correct ranking of attributes leads to one reason decision making, knowing the directions of the attributes leads to equal weighting, and not knowing about either leads to strategies that use weighted combinations of multiple attributes. This allows us to gain new insights on mixed results of prior empirical work on heuristic decision making. In three empirical paired comparison studies with continuous features, we verify predictions of our theory, and show that it captures several characteristics of human decision making not explained by alternative theories.

show abstract

Section: Resource Rationalitymentioning

confidence: 99%

Section: Resource Rationalitymentioning

confidence: 99%

Heuristics From Bounded Meta-Learned Inference

Binz¹,

Gershman²,

Schulz

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“… 1 Another family of approximation methods, known as variational Bayes ( Blei & Jordan, 2006 ; Blei, Kucukelbir, & McAuliffe, 2017 ), optimizes an approximate, simplified model of the probability distribution of interest, rather than working with a sample from that distribution. This approach may also be the starting point for neuroscientific and psychological hypotheses, although we do not consider it further here ( Dasgupta, Schulz, Tenenbaum, & Gershman, 2019 ; Gershman & Beck, 2017 ; Ma, Beck, Latham, & Pouget, 2006 ; Sanborn, 2017 ). …”

mentioning

confidence: 99%

The Bayesian sampler: Generic Bayesian inference causes incoherence in human probability judgments.

Zhu

Sanborn

Chater

2020

Psychological Review

170

View full text Add to dashboard Cite

Human probability judgments are systematically biased, in apparent tension with Bayesian models of cognition. But perhaps the brain does not represent probabilities explicitly, but approximates probabilistic calculations through a process of sampling, as used in computational probabilistic models in statistics. Naïve probability estimates can be obtained by calculating the relative frequency of an event within a sample, but these estimates tend to be extreme when the sample size is small. We propose instead that people use a generic prior to improve the accuracy of their probability estimates based on samples, and we call this model the Bayesian sampler. The Bayesian sampler trades off the coherence of probabilistic judgments for improved accuracy, and provides a single framework for explaining phenomena associated with diverse biases and heuristics such as conservatism and the conjunction fallacy. The approach turns out to provide a rational reinterpretation of "noise" in an important recent model of probability judgment, the probability theory plus noise model (Costello & Watts, 2014, 2016a, 2019Costello, Watts, & Fisher, 2018), making equivalent average predictions for simple events, conjunctions, and disjunctions. The Bayesian sampler does, however, make distinct predictions for conditional probabilities, and we show in a new experiment that this model better captures these judgments both qualitatively and quantitatively.

show abstract

“…Thus, consumers make approximations about the most essential features of the environment and then compare the data to their mental model according to the Bayesian principle. At least in part, the Bayesian approach solves the curse of dimensionality ( Dasgupta et al, 2020 ) if it is assumed that inductive inference addresses and learns the most important aspects of the individuals’ environment. In this sense, cultural habits, social norms, and attitudes play an important role in consumer decision-making.…”

Section: Consumers Control Contexts By Expectationsmentioning

confidence: 99%

The Consumer Contextual Decision-Making Model

Suomala

2020

Front. Psychol.

View full text Add to dashboard Cite

Consumers can have difficulty expressing their buying intentions on an explicit level. The most common explanation for this intention-action gap is that consumers have many cognitive biases that interfere with rational decision-making. The current resource-rational approach to understanding human cognition, however, suggests that brain environment interactions lead consumers to minimize the expenditure of cognitive energy according to the principle of Occam's Razor. This means that the consumer seeks as simple of a solution as possible for a problem requiring decision-making. In addition, this resourcerational approach to decision-making emphasizes the role of inductive inference and Bayesian reasoning. Together, the principle of Occam's Razor, inductive inference, and Bayesian reasoning illuminate the dynamic human-environment relationship. This paper analyzes these concepts from a contextual perspective and introduces the Consumer Contextual Decision-Making Model (CCDMM). Based on the CCDMM, two hypothetical strategies of consumer decision-making will be presented. First, the SIMilarity-Strategy (SIMS) is one in which most of a consumer's decisions in a real-life context are based on prior beliefs about the role of a commodities specific to real-life situation being encountered. Because beliefs are based on previous experiences, consumers are already aware of the most likely consequences of their actions. At the same time, they do not waste time on developing contingencies for what, based on previous experience, is unlikely to happen. Second, the What-is-Out-therein the World Strategy (WOWS) is one in which prior beliefs do not work in a real-life situation, requiring consumers to update their beliefs. The principle argument being made is that most experimental consumer research describes decisionmaking based on the WOWS, when participants cannot apply their previous knowledge and situation-based strategy to problems. The article analyzes sensory and cognitive biases described by behavioral economists from a CCDMM perspective, followed by a description and explanation of the typical intention-action gap based on the model. Prior to a section dedicated to discussion, the neuroeconomic approach will be described along with the valuation network of the brain, which has evolved to solve problems that the human has previously encountered in an information-rich environment. The principles of brain function will also be compared to CCDMM. Finally, different approaches and the future direction of consumer research from a contextual point of view will be presented.

show abstract

A theory of learning to infer

Cited by 23 publications

References 148 publications

Heuristics From Bounded Meta-Learned Inference

Heuristics From Bounded Meta-Learned Inference

The Bayesian sampler: Generic Bayesian inference causes incoherence in human probability judgments.

The Consumer Contextual Decision-Making Model

Contact Info

Product

Resources

About