Balancing safety and efficiency in human decision making DOI Open Access
Pranav Mahajan, Shuangyi Tong, Sang Wan Lee

et al.

Published: Oct. 18, 2024

The safety-efficiency dilemma describes the problem of maintaining safety during efficient exploration and is a special case exploration-exploitation in face potential dangers. Conventional solutions collapse punishment reward into single feedback signal, whereby early losses can be overcome by later gains. However, brain has separate system for Pavlovian fear learning, suggesting possible computational advantage to specific memory exploratory decision-making. In series simulations, we show this promotes safe but learning optimised arbitrating avoidance instrumental decision-making according uncertainty. We provide basic test model simple human approach-withdrawal experiment, that flexible captures choice reaction times. These results more sophisticated role than previously thought, shaping behaviour computationally precise manner.

Language: Английский

Using computational models of learning to advance cognitive behavioral therapy DOI Creative Commons
Isabel M. Berwian,

Peter Hitchock,

Sashank Pisupati

et al.

Communications Psychology, Journal Year: 2025, Volume and Issue: 3(1)

Published: April 27, 2025

Abstract Many psychotherapy interventions have a large evidence base and can help substantial number of people with symptoms mental health conditions. However, we still little understanding why treatments work. Early advances in psychotherapy, such as the development exposure therapy, built on theoretical experimental from Pavlovian instrumental conditioning. More generally, all achieves change through learning. The past 25 years seen developments computational models learning, increased precision focus multiple learning mechanisms their interaction. Now might be good time to formalize improve our psychotherapy. To advance research bring together new joint field theory-driven first review literature cognitive behavioral therapy (exposure restructuring) introduce reinforcement representation We then suggest mapping these algorithms processes presumably underlying effects restructuring. Finally, outline how lens inform intervention research.

Language: Английский

Citations

0
Active reinforcement learning versus action bias and hysteresis: control with a mixture of experts and nonexperts DOI Creative Commons
Jaron T. Colas, John P. O’Doherty, Scott T. Grafton

et al.

PLoS Computational Biology, Journal Year: 2024, Volume and Issue: 20(3), P. e1011950 - e1011950

Published: March 29, 2024

Active reinforcement learning enables dynamic prediction and control, where one should not only maximize rewards but also minimize costs such as of inference, decisions, actions, time. For an embodied agent a human, decisions are shaped by physical aspects actions. Beyond the effects reward outcomes on processes, to what extent can modeling behavior in reinforcement-learning task be complicated other sources variance sequential action choices? What bias (for actions per se) hysteresis determined history chosen previously? The present study addressed these questions with incremental assembly models for choice data from hierarchical structure additional complexity learning. With systematic comparison falsification computational models, human choices were tested signatures parallel modules representing enhanced form generalized hysteresis. We found evidence substantial differences across participants—even comparable magnitude individual Individuals who did learn well revealed greatest biases, those accurately significantly biased. direction varied among individuals repetition or, more commonly, alternation biases persisting multiple previous Considering that button presses trivial motor demands, idiosyncratic forces biasing sequences robust enough suggest ubiquity tasks requiring various In light how function heuristic efficient control adapts uncertainty or low motivation minimizing cost effort, phenomena broaden consilient theory mixture experts encompass expert nonexpert controllers behavior.

Language: Английский

Citations

1
Balancing safety and efficiency in human decision making DOI Open Access
Pranav Mahajan, Shuangyi Tong, Sang Wan Lee

et al.

Published: Oct. 18, 2024

The safety-efficiency dilemma describes the problem of maintaining safety during efficient exploration and is a special case exploration-exploitation in face potential dangers. Conventional solutions collapse punishment reward into single feedback signal, whereby early losses can be overcome by later gains. However, brain has separate system for Pavlovian fear learning, suggesting possible computational advantage to specific memory exploratory decision-making. In series simulations, we show this promotes safe but learning optimised arbitrating avoidance instrumental decision-making according uncertainty. We provide basic test model simple human approach-withdrawal experiment, that flexible captures choice reaction times. These results more sophisticated role than previously thought, shaping behaviour computationally precise manner.

Language: Английский

Citations

1
Interindividual differences in Pavlovian influence on learning are consistent DOI Creative Commons
Sepehr Saeedpour,

Mostafa Minadari Hossein,

Ophélia Deroy

et al.

Royal Society Open Science, Journal Year: 2023, Volume and Issue: 10(9)

Published: Sept. 1, 2023

Pavlovian influences impair instrumental learning. It is easier to learn approach reward-predictive signals and avoid punishment-predictive cues than their contrary. Whether the interindividual variability in this influence consistent across time has been examined by a number of recent studies met with mixed results. Here we introduce an open-source, web-based instance well-established Go-NoGo paradigm for measuring influence. We closely replicated previous laboratory-based Moreover, differences were two-week window at level (i) raw measures learning (i.e. performance accuracy), (ii) linear, descriptive estimates bias (test-retest reliability: 0.40), (iii) parameters obtained from reinforcement model fitting selection 0.25). Nonetheless, correlations reported here are still lower standards 0.7) employed psychometrics self-reported measures. Our results provide support trusting as relatively stable individual characteristic using its measure computational understanding human mental health.

Language: Английский

Citations

3
Balancing safety and efficiency in human decision making DOI Creative Commons
Pranav Mahajan, Shuangyi Tong, Sang Wan Lee

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Jan. 24, 2024

ABSTRACT The safety-efficiency dilemma describes the problem of maintaining safety during efficient exploration and is a special case exploration-exploitation in face potential dangers. Conventional solutions collapse punishment reward into single feedback signal, whereby early losses can be overcome by later gains. However, brain has separate system for Pavlovian fear learning, suggesting possible computational advantage to specific memory exploratory decision-making. In series simulations, we show this promotes safe but learning optimised arbitrating avoidance instrumental decision-making according uncertainty. We provide basic test model simple human approach-withdrawal experiment, that flexible captures choice reaction times. These results more sophisticated role than previously thought, shaping behaviour computationally precise manner.

Language: Английский

Citations

0
Bayesian Priors in Active Avoidance DOI Open Access

Tobias Granwald,

Peter Dayan, Máté Lengyel

et al.

Published: Aug. 8, 2024

Failing to make decisions that would actively avoid negative outcomes is central helplessness. In a Bayesian framework, deciding whether act informed by beliefs about the world can be characterised as priors. However, these priors have not been previously quantified. Here we administered two tasks in which participants decided attempt active avoidance actions. The differed framing and valence, allowing us test prior generating biases behaviour problem-specific or task-independent general. We performed extensive comparisons of models offering different structural explanations data, finding model with task-invariant for provided best fit participants’ trial-by-trial behaviour. parameters this were reliable, an optimistic also reported higher levels positive affect. These results show individual differences explain engage outcomes, providing evidence conceptualization

Language: Английский

Citations

0
Balancing safety and efficiency in human decision making DOI Open Access
Pranav Mahajan, Shuangyi Tong, Sang Wan Lee

et al.

Published: Oct. 18, 2024

The safety-efficiency dilemma describes the problem of maintaining safety during efficient exploration and is a special case exploration-exploitation in face potential dangers. Conventional solutions collapse punishment reward into single feedback signal, whereby early losses can be overcome by later gains. However, brain has separate system for Pavlovian fear learning, suggesting possible computational advantage to specific memory exploratory decision-making. In series simulations, we show this promotes safe but learning optimised arbitrating avoidance instrumental decision-making according uncertainty. We provide basic test model simple human approach-withdrawal experiment, that flexible captures choice reaction times. These results more sophisticated role than previously thought, shaping behaviour computationally precise manner.

Language: Английский

Citations

0