Response outcomes gate the impact of expectations on perceptual decisions

Hermoso-Mendizabal, Ainhoa; Hyafil, Alexandre; Rueda-Orozco, Pavel E.; Jaramillo, Santiago; Robbe, David; Rocha, Jaime de la

doi:10.1038/s41467-020-14824-w

Cited by 48 publications

(132 citation statements)

References 45 publications

Supporting

Mentioning

115

Contrasting

Order By: Relevance

“…Fig. 1) (Hermoso-Mendizabal et al 2020). Importantly, as with the reset strategy, this disregard of any information involving error trials is not optimal, since in the 2AFC task all types of transitions are equally informative (Supp.…”

Section: Resultsmentioning

confidence: 99%

Ecologically pre-trained RNNs explain suboptimal animal decisions

Molano-Mazón

Duque

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

When faced with a new task, animals′ cognitive capabilities are determined both by individual experience and by structural priors evolved to leverage the statistics of natural environments. Rats can quickly learn to capitalize on the trial sequence correlations of two-alternative forced choice (2AFC) tasks after correct trials, but consistently deviate from optimal behavior after error trials, when they waive the accumulated evidence. To understand this outcome-dependent gating, we first show that Recurrent Neural Networks (RNNs) trained in the same 2AFC task outperform animals as they can readily learn to use previous trials′ information both after correct and error trials. We hypothesize that, while RNNs can optimize their behavior in the 2AFC task without a priori restrictions, rats′ strategy is constrained by a structural prior adapted to a natural environment in which rewarded and non-rewarded actions provide largely asymmetric information. When pre-training RNNs in a more ecological task with more than two possible choices, networks develop a strategy by which they gate off the across-trial evidence after errors, mimicking rats′ behavior. Our results suggest that the observed suboptimal behavior reflects the influence of a structural prior that, adaptive in a natural multi-choice environment, constrains performance in a 2AFC laboratory task.

show abstract

Section: Resultsmentioning

confidence: 99%

Ecologically pre-trained RNNs explain suboptimal animal decisions

Molano-Mazón

Duque

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Accumulation (possibly temporally discounted) of action outcomes also underlies several reinforcement learning algorithms ( 4, 27 – 29 ). Although less attention has been devoted to reset-like computations (but see ( 30 )), they are also essential for inference when certain observations specify a state unambiguously ( 2 ). Our results show that these computations can be cast into a single parametric generative model capable of producing a set of decision variables embodying a variety of stimulus-bound and inference-based strategies.…”

Section: Discussionmentioning

confidence: 99%

A repertoire of foraging decision variables in the mouse brain

Cazettes

Murakami

Renart

et al. 2021

Preprint

View full text Add to dashboard Cite

Decision making strategies guided by observable stimuli and those that also require inferences about unobserved states have been linked to distinct computational requirements and neural substrates. Here, we formulate a model based on temporal integration and reset that incorporates both strategies into a unified family of decision algorithms. We show, using recordings from the frontal cortex of mice performing a foraging task, that the entire family of algorithms can be simultaneously decoded from the same neural ensemble, regardless of the one concurrently executed by the mice. Thus, using multiplexed integration, the cortex may avoid premature commitment to a single algorithm and maintain multiple decision strategies in parallel.

show abstract

“…A source for choice biases and their dependence on recent history could involve stable working memory representations ( Fischer and Whitney, 2014 ; Fritsche et al, 2017 ; Urai et al, 2019 ). History effects on choice have been observed in humans ( Fischer and Whitney, 2014 ; Abrahamyan et al, 2016 ) and rodents ( Busse et al, 2011 ; Akrami et al, 2018 ) and are present in forced-choice ( Gold et al, 2008 ; Fischer and Whitney, 2014 ), memory-guided ( Akrami et al, 2018 ; Hermoso-Mendizabal et al, 2018 ), and free-choice tasks ( Sugrue et al, 2004 ; Lau and Glimcher, 2005 ). We also found that mice and humans adapted their behavioral strategies depending on the discriminability trajectories established during training ( Ahissar et al, 2009 ; Trevino et al, 2013 ).…”

Section: Discussionmentioning

confidence: 99%

Adaptive Choice Biases in Mice and Humans

Treviño

León

Haro

2020

Front. Behav. Neurosci.

View full text Add to dashboard Cite

The contribution of non-sensory information processing to perceptual decision making is not fully understood. Choice biases have been described for mice and humans and are highly prevalent even if they decrease rewarding outcomes. Choice biases are usually reduced by discriminability because stimulus strength directly enables the adjustments in the decision strategies used by decision-makers. However, choice biases could also derive from functional asymmetries in sensory processing, decision making, or both. Here, we tested how particular experimental contingencies influenced the production of choice biases in mice and humans. Our main goal was to establish the tasks and methods to jointly characterize psychometric performance and innate side-choice behavior in mice and humans. We implemented forced and un-forced visual tasks and found that both species displayed stable levels of side-choice biases, forming continuous distributions from low to high levels of choice stereotypy. Interestingly, stimulus discriminability reduced the side-choice biases in forced-choice, but not in free-choice tasks. Choice biases were stable in appearance and intensity across experimental days and could be employed to identify mice and human participants. Additionally, side- and alternating choices could be reinforced for both mice and humans, implying that choice biases were adaptable to non-visual manipulations. Our results highlight the fact that internal and external elements can influence the production of choice biases. Adaptations of our tasks could become a helpful diagnostic tool to detect aberrant levels of choice variability.

show abstract

Response outcomes gate the impact of expectations on perceptual decisions

Cited by 48 publications

References 45 publications

Ecologically pre-trained RNNs explain suboptimal animal decisions

Ecologically pre-trained RNNs explain suboptimal animal decisions

A repertoire of foraging decision variables in the mouse brain

Adaptive Choice Biases in Mice and Humans

Contact Info

Product

Resources

About