SSRIs modulate asymmetric learning from reward and punishment

Michely, Jochen; Eldar, Eran; Erdman, Alon; Martin, Ingrid M.; Dolan, Raymond J.

doi:10.1101/2020.05.21.108266

Cited by 11 publications

(14 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Future research should examine how valence biases in learning and memory, as well as the reward statistics of an individual's realworld environment, relate to vulnerability or resilience to psychopathology across adolescent development. Finally, given an extensive literature demonstrating the pronounced influence of neuromodulatory systems on both valence biases in reinforcement learning (Cox et al, 2015;Frank et al, 2004Frank et al, , 2007Michely et al, 2020) and value-guided memory (Lisman & Grace, 2005;Sara, 2009), future studies might examine how developmental changes in these systems relate to the age-related shifts in valence biases observed here.…”

Section: Discussionmentioning

confidence: 95%

Valence biases in reinforcement learning shift across adolescence and modulate subsequent memory

Rosenbaum¹,

Grassie²,

Hartley³

2020

Preprint

View full text Add to dashboard Cite

In new or uncertain situations, we must often learn through experience to make choices likely to result in good outcomes and to avoid those that may yield bad outcomes. Past research suggests that individuals differ in this trial-and-error learning, with some preferentially incorporating good outcomes into their subsequent evaluations, and others weighting bad outcomes more heavily. Such valence biases in learning may additionally influence what we later remember about our experiences. Adolescents have been proposed to be particularly sensitive to rewarding experiences, a tendency which may foster a propensity for risky choice. In the present study, we examined whether valence asymmetries in reinforcement learning change across adolescent development, conferring age differences in risk preferences, and whether individual learning asymmetries might also bias subsequent memory. Participants ages 8-27 chose between pairs of probabilistic and deterministic “point machines” to learn their associated reward outcomes. Trial-unique images were presented with the outcome of each choice, and incidental memory for these pictures was tested immediately following the learning task. Counterintuitively, adolescents exhibited less risk taking than children and adults, corresponding to a bias toward overweighting worse- relative to better-than-expected outcomes during reinforcement learning. Incidental memory was also modulated by valence biases in learning, such that individuals who learned more from worse-than-expected outcomes were more likely to remember the trial-unique images that coincided with them, and vice versa. Collectively, these results highlight age-related changes across adolescence in the computation of subjective value, and demonstrate that a valence asymmetric valuation process also influences how information is prioritized in episodic memory.

show abstract

Section: Discussionmentioning

confidence: 95%

Valence biases in reinforcement learning shift across adolescence and modulate subsequent memory

Rosenbaum¹,

Grassie²,

Hartley³

2020

Preprint

View full text Add to dashboard Cite

show abstract

“…One individual has what we denote as trait depression, defined by a greater learning rate for positive surprises than for negative surprises. This sounds counter-intuitive, but we consider depression emerges as a function of more slowly learning about negative outcomes [23]. That is, the individual fails to update their expectations about negative events and therefore believes their co-workers to be more trustworthy than they truly are.…”

Section: Simulating How Persistent Low Mood Can Results From Intrinsicmentioning

confidence: 96%

Towards formal models of psychopathological traits that explain symptom trajectories

Sharp

Miller

Dolan

et al. 2020

BMC Med

Self Cite

View full text Add to dashboard Cite

Background A dominant methodology in contemporary clinical neuroscience is the use of dimensional self-report questionnaires to measure features such as psychological traits (e.g., trait anxiety) and states (e.g., depressed mood). These dimensions are then mapped to biological measures and computational parameters. Researchers pursuing this approach tend to equate a symptom inventory score (plus noise) with some latent psychological trait. Main text We argue this approach implies weak, tacit, models of traits that provide fixed predictions of individual symptoms, and thus cannot account for symptom trajectories within individuals. This problem persists because (1) researchers are not familiarized with formal models that relate internal traits to within-subject symptom variation and (2) rely on an assumption that trait self-report inventories accurately indicate latent traits. To address these concerns, we offer a computational model of trait depression that demonstrates how parameters instantiating a given trait remain stable while manifest symptom expression varies predictably. We simulate patterns of mood variation from both the computational model and the standard self-report model and describe how to quantify the relative validity of each model using a Bayesian procedure. Conclusions Ultimately, we would urge a tempering of a reliance on self-report inventories and recommend a shift towards developing mechanistic trait models that can explain within-subject symptom dynamics.

show abstract

“…Models that dynamically adjust learning rates based on experienced outcomes could also yield further insight into the block order effects we observed by allowing for learned information about the environment's reward statistics to be carried over into new contexts. Biologically, the flexible adjustment of valence biases may be implemented by dopaminergic and serotonergic mechanisms (Collins & Frank, 2014;Cox et al, 2015;Daw et al, 2002;Frank et al, 2007;Michely et al, 2020), which undergo pronounced changes from childhood to early adulthood (Doremus-Fitzwater & Spear, 2016;Li, 2013). Thus, future studies should explicitly test different, biologically plausible algorithms through which learning rates may be dynamically updated -as well as how their underlying parameters may change across age.…”

Section: Discussionmentioning

confidence: 99%

Flexibility in valenced reinforcement learning computations across development

Nussenbaum¹,

Velez²,

Washington³

et al. 2021

Preprint

View full text Add to dashboard Cite

Optimal integration of positive and negative outcomes during learning varies depending on an environment’s reward statistics. The present study investigated the extent to which children, adolescents, and adults (N = 142 8 - 25 year-olds, 55% female, 42% White, 31% Asian, 17% mixed race, and 8% Black) adapt their weighting of better-than-expected and worse-than-expected outcomes when learning from reinforcement. Participants made a series of choices across two contexts: one in which weighting positive outcomes more heavily than negative outcomes led to better performance, and one in which the reverse was true. Reinforcement learning modeling revealed that across age, participants shifted their valence biases in accordance with the structure of the environment. Exploratory analyses revealed increases in context-dependent flexibility with age.

show abstract

SSRIs modulate asymmetric learning from reward and punishment

Cited by 11 publications

References 31 publications

Valence biases in reinforcement learning shift across adolescence and modulate subsequent memory

Valence biases in reinforcement learning shift across adolescence and modulate subsequent memory

Towards formal models of psychopathological traits that explain symptom trajectories

Flexibility in valenced reinforcement learning computations across development

Contact Info

Product

Resources

About