Value signals guide abstraction during learning
The human brain excels at constructing and using abstractions, such as rules, or concepts. Here, in two fMRI experiments, we demonstrate a mechanism of abstraction built upon the valuation of sensory features. Human volunteers learned novel association rules based on simple visual features. Reinforc...
Main Authors: | , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
eLife Sciences Publications Ltd
2021-07-01
|
Series: | eLife |
Subjects: | |
Online Access: | https://elifesciences.org/articles/68943 |
_version_ | 1811200778924195840 |
---|---|
author | Aurelio Cortese Asuka Yamamoto Maryam Hashemzadeh Pradyumna Sepulveda Mitsuo Kawato Benedetto De Martino |
author_facet | Aurelio Cortese Asuka Yamamoto Maryam Hashemzadeh Pradyumna Sepulveda Mitsuo Kawato Benedetto De Martino |
author_sort | Aurelio Cortese |
collection | DOAJ |
description | The human brain excels at constructing and using abstractions, such as rules, or concepts. Here, in two fMRI experiments, we demonstrate a mechanism of abstraction built upon the valuation of sensory features. Human volunteers learned novel association rules based on simple visual features. Reinforcement-learning algorithms revealed that, with learning, high-value abstract representations increasingly guided participant behaviour, resulting in better choices and higher subjective confidence. We also found that the brain area computing value signals – the ventromedial prefrontal cortex – prioritised and selected latent task elements during abstraction, both locally and through its connection to the visual cortex. Such a coding scheme predicts a causal role for valuation. Hence, in a second experiment, we used multivoxel neural reinforcement to test for the causality of feature valuation in the sensory cortex, as a mechanism of abstraction. Tagging the neural representation of a task feature with rewards evoked abstraction-based decisions. Together, these findings provide a novel interpretation of value as a goal-dependent, key factor in forging abstract representations. |
first_indexed | 2024-04-12T02:09:24Z |
format | Article |
id | doaj.art-bb2ff8a341ba41dabbc17b578850c6b8 |
institution | Directory Open Access Journal |
issn | 2050-084X |
language | English |
last_indexed | 2024-04-12T02:09:24Z |
publishDate | 2021-07-01 |
publisher | eLife Sciences Publications Ltd |
record_format | Article |
series | eLife |
spelling | doaj.art-bb2ff8a341ba41dabbc17b578850c6b82022-12-22T03:52:26ZengeLife Sciences Publications LtdeLife2050-084X2021-07-011010.7554/eLife.68943Value signals guide abstraction during learningAurelio Cortese0https://orcid.org/0000-0003-4567-0924Asuka Yamamoto1Maryam Hashemzadeh2Pradyumna Sepulveda3https://orcid.org/0000-0003-0159-6777Mitsuo Kawato4Benedetto De Martino5https://orcid.org/0000-0002-3555-2732Computational Neuroscience Labs, ATR Institute International, Kyoto, Japan; Institute of Cognitive Neuroscience, University College London, London, United KingdomComputational Neuroscience Labs, ATR Institute International, Kyoto, Japan; School of Information Science, Nara Institute of Science and Technology, Nara, JapanDepartment of Computing Science, University of Alberta, Edmonton, CanadaInstitute of Cognitive Neuroscience, University College London, London, United KingdomComputational Neuroscience Labs, ATR Institute International, Kyoto, Japan; RIKEN Center for Artificial Intelligence Project, Kyoto, JapanInstitute of Cognitive Neuroscience, University College London, London, United KingdomThe human brain excels at constructing and using abstractions, such as rules, or concepts. Here, in two fMRI experiments, we demonstrate a mechanism of abstraction built upon the valuation of sensory features. Human volunteers learned novel association rules based on simple visual features. Reinforcement-learning algorithms revealed that, with learning, high-value abstract representations increasingly guided participant behaviour, resulting in better choices and higher subjective confidence. We also found that the brain area computing value signals – the ventromedial prefrontal cortex – prioritised and selected latent task elements during abstraction, both locally and through its connection to the visual cortex. Such a coding scheme predicts a causal role for valuation. Hence, in a second experiment, we used multivoxel neural reinforcement to test for the causality of feature valuation in the sensory cortex, as a mechanism of abstraction. Tagging the neural representation of a task feature with rewards evoked abstraction-based decisions. Together, these findings provide a novel interpretation of value as a goal-dependent, key factor in forging abstract representations.https://elifesciences.org/articles/68943reinforcement learningabstractionvmpfcconfidencemultivoxel neural reinforcementvaluation |
spellingShingle | Aurelio Cortese Asuka Yamamoto Maryam Hashemzadeh Pradyumna Sepulveda Mitsuo Kawato Benedetto De Martino Value signals guide abstraction during learning eLife reinforcement learning abstraction vmpfc confidence multivoxel neural reinforcement valuation |
title | Value signals guide abstraction during learning |
title_full | Value signals guide abstraction during learning |
title_fullStr | Value signals guide abstraction during learning |
title_full_unstemmed | Value signals guide abstraction during learning |
title_short | Value signals guide abstraction during learning |
title_sort | value signals guide abstraction during learning |
topic | reinforcement learning abstraction vmpfc confidence multivoxel neural reinforcement valuation |
url | https://elifesciences.org/articles/68943 |
work_keys_str_mv | AT aureliocortese valuesignalsguideabstractionduringlearning AT asukayamamoto valuesignalsguideabstractionduringlearning AT maryamhashemzadeh valuesignalsguideabstractionduringlearning AT pradyumnasepulveda valuesignalsguideabstractionduringlearning AT mitsuokawato valuesignalsguideabstractionduringlearning AT benedettodemartino valuesignalsguideabstractionduringlearning |