Irina Baetu

Are Preventive and Generative Causal Reasoning Symmetrical? Extinction and Competition

Publisher: SAGE Publications

Date: 09-2012

DOI: 10.1080/17470218.2012.667424

Abstract: We tested whether preventive and generative reasoning processes are symmetrical by keeping the training and testing of preventive (inhibitory) and generative (excitatory) causal cues as similar as possible. In Experiment 1, we extinguished excitors and inhibitors in a blocking design, in which each extinguished cause was presented in compound with a novel cause, with the same outcome occurring following the compound and following the novel cause alone. With this novel extinction procedure, the inhibitory cues seemed more likely to lose their properties than the excitatory cues. In Experiment 2, we investigated blocking of excitatory and inhibitory causes and found similar blocking effects. Taken together, these results suggest that acquisition of excitation and inhibition is similar, but that inhibition is more liable to extinguish with our extinction procedure. In addition, we used a variable outcome, and this enabled us to test the predictions of an inferential reasoning account about what happens when the outcome level is at its minimum or maximum (De Houwer, Beckers, & Glautier, 2002). We discuss the predictions of this inferential account, Rescorla and Wagner's (1972) model, and a connectionist model—the auto-associator.

Polymorphisms in dopaminergic genes predict proactive processes of response inhibition

Publisher: Wiley

Date: 09-01-2019

DOI: 10.1111/EJN.14323

Abstract: The ability to inhibit a prepared emotional or motor action is difficult but critical to everyday functioning. It is well-established that response inhibition relies on the dopaminergic system in the basal ganglia. However, response inhibition is often measured imprecisely due to a process which slows our responses and increases subsequent inhibition success known as proactive inhibition. As the role of the dopamine system in proactive inhibition is unclear, we investigated the contribution of dopaminergic genes to proactive inhibition. We operationalised proactive inhibition as slower responses after failures to inhibit a response in a Go/No-Go paradigm and investigated its relationship to rs686/A at DRD1 (associated with increased gene expression) and rs1800497/T at DRD2 (associated with reduced D2 receptor availability). Even though our s le (N = 264) was relatively young (18-40 years), we found that proactive inhibition improves the ability to withhold erroneous responses in older participants (p = 0.002) and those with lower fluid intelligence scores (p < 0.001), indicating that proactive inhibition is likely a naturally occurring compensatory mechanism. Critically, we found that a polygenic risk score consisting of the number of rs686 A and rs1800497 T alleles predicts higher engagement of proactive inhibition (p = 0.040), even after controlling for age (p = 0.011). Furthermore, age seemed to magnify these genetic effects (p < 0.001). This suggests that the extent to which proactive inhibition is engaged depends on increased dopamine D1 and decreased D2 neurotransmission. These results provide important considerations for future work investigating disorders of the dopaminergic system.

Age-related differences in sequence learning: Findings from two visuo-motor sequence learning tasks

Publisher: Wiley

Date: 24-03-2018

DOI: 10.1111/BJOP.12299

Abstract: The Serial Reaction Time Task (SRTT) is thought to assess implicit learning, which seems to be preserved with age. However, the reaction time (RT) measures employed on implicit-like tasks might be too unreliable to detect in idual differences. We investigated whether RT-based measures mask age effects by comparing the performance of 43 younger and 35 older adults on SRTT and an explicit-like Predictive Sequence Learning Task (PSLT). RT-based measures (difference scores and a ratio) were collected for both tasks, and accuracy was additionally measured for PSLT. We also measured fluid abilities. The RT-difference scores indicated preserved SRTT and PSLT performance with age and did not correlate with fluid abilities, while ratio RT and the accuracy-based measures indicated age-related decline and correlated with fluid abilities. Therefore, RT-difference scores might mask in idual differences, which compromises the interpretation of previous studies using SRTT.

Accuracy-based measures provide a better measure of sequence learning than reaction time-based measures

Publisher: Frontiers Media SA

Date: 11-08-2015

DOI: 10.3389/FPSYG.2015.01158

Measurement of the response inhibition network in a brief battery of tasks: Introducing an assessment of reactive and proactive inhibition processes.

Publisher: Center for Open Science

Date: 08-11-2022

DOI: 10.31234/OSF.IO/WX5AE

Abstract: Traditional response inhibition tasks are assumed to capture one’s ability to inhibit a response. This ability, however, requires a reactive process and a proactive process, post-error slowing (PES). Recent evidence shows that Stop-Signal Tasks (SSTs) measure the reactive process, and while the Sustained Attention to Response Task measures overall response inhibition, that measure is confounded by a proactive process. Since the diseases associated with response inhibition deficits often co-occur with symptoms that diminish the capacity for lengthy behavioural testing, and, since it is unknown to which process such decrements can be attributed and where in the brain these processes are generated, rapid and precise measurement of reactive and proactive processes is important. To address these issues, we administered a battery of four response inhibition tasks to healthy young adults (N = 123), two SSTs and two Go/No-Go tasks. In three tasks, we implemented adaptations to allow direct observation of proactive inhibition, reactive inhibition, and overall response inhibition. We introduce a novel cueing procedure to investigate the possibility of a predictive mechanism of proactive inhibition, arguing that slower response times on trials with a higher Stop/No-Go probability indicate predictive proactive inhibition. Based on these findings, we propose a novel demarcation to proactive inhibition: remedial proactive inhibition (PES), and predictive proactive inhibition. Additionally, we provide empirical support for a Bayesian adaptive staircase (Livesey & Livesey, 2016) that allows rapid convergence on estimates of reactive inhibition in SSTs in as few as 20 trials that are robust against potential predictive proactive inhibition confounds.

Propositional learning is a useful research heuristic but it is not a theoretical algorithm

Publisher: Cambridge University Press (CUP)

Date: 04-2009

DOI: 10.1017/S0140525X09000879

Abstract: Mitchell et al.'s claim, that their propositional theory is a single-process theory, is illusory because they relegate some learning to a secondary memory process. This renders the single-process theory untestable. The propositional account is not a process theory of learning, but rather, a heuristic that has led to interesting research.

Extinction and blocking of conditioned inhibition in human causal learning

Publisher: Springer Science and Business Media LLC

Date: 11-2010

DOI: 10.3758/LB.38.4.394

Maybe this old dinosaur isn’t extinct: What does Bayesian modeling add to associationism?

Publisher: Cambridge University Press (CUP)

Date: 08-2011

DOI: 10.1017/S0140525X11000203

Abstract: We agree with Jones & Love (J& L) that much of Bayesian modeling has taken a fundamentalist approach to cognition but we do not believe in the potential of Bayesianism to provide insights into psychological processes. We discuss the advantages of associative explanations over Bayesian approaches to causal induction, and argue that Bayesian models have added little to our understanding of human causal reasoning.

Post-Error Slowing May Not Be Proactive: Electrophysiological evidence favours a disorienting account

Publisher: Center for Open Science

Date: 03-06-2023

DOI: 10.31234/OSF.IO/V3SRX

Abstract: The slowing down of a response after committing an error in speeded response tasks has been reliably observed over the last 60 years, but no explanation has yet been articulated to account for it. Post-error slowing (PES) is thought to reflect a proactive mechanism to improve one’s chances of successfully inhibiting a response or selecting the correct response from an array of possibilities. Recently, Dutilh and colleagues (2012a) used computational modelling to compare how well several accounts of PES fit real and simulated data. They concluded that PES is the result of participants widening their response boundaries, which they assumed corresponds to increased caution. This explanation supports a proactive account of PES. We used EEG to test the same four accounts modelled by Dutilh and colleagues to provide direct neural evidence to supplement their simulated data. In a Go/NoGo task administered to N = 100 healthy young adults (24.3 ± 4.8 yrs), we mapped ERP parameters to the theoretical drift parameters established by Dutilh and colleagues. Their hypothesis would predict larger N2 after errors and that the litude of the N2 should correlate with magnitude of PES. Our results did not support these predictions (N2 litude was smaller after errors, p = .015, and there was no correlation between N2 litude and PES, p = .523). Our findings support another common account of PES, a disorienting account, that supposes errors disrupt attentional processing. The post-error anterior N1 was significantly disrupted by errors (p = .020) and was correlated with the magnitude of PES (p = .016). We, therefore, suggest that PES is not completely proactive, but rather is partially the consequence of disruptions to attentional processing that only incidentally improve response inhibition by offsetting the initiation of response execution. Interestingly, the post-error N1 in older adults was diminished (p = .0008), but higher general intelligence rescued such disruptions to attention (p & .0001), indicating a partial compensatory mechanism in ageing that is supported by general intelligence.

Commonly-occurring polymorphisms in the COMT, DRD1 and DRD2 genes influence different aspects of motor sequence learning in humans

Publisher: Elsevier BV

Date: 11-2015

DOI: 10.1016/J.NLM.2015.09.009

Abstract: Performing sequences of movements is a ubiquitous skill that involves dopamine transmission. However, it is unclear which components of the dopamine system contribute to which aspects of motor sequence learning. Here we used a genetic approach to investigate the relationship between different components of the dopamine system and specific aspects of sequence learning in humans. In particular, we investigated variations in genes that code for the catechol-O-methyltransferase (COMT) enzyme, the dopamine transporter (DAT) and dopamine D1 and D2 receptors (DRD1 and DRD2). COMT and the DAT regulate dopamine availability in the prefrontal cortex and the striatum, respectively, two key regions recruited during learning, whereas dopamine D1 and D2 receptors are thought to be involved in long-term potentiation and depression, respectively. We show that polymorphisms in the COMT, DRD1 and DRD2 genes differentially affect behavioral performance on a sequence learning task in 161 Caucasian participants. The DRD1 polymorphism predicted the ability to learn new sequences, the DRD2 polymorphism predicted the ability to perform a previously learnt sequence after performing interfering random movements, whereas the COMT polymorphism predicted the ability to switch flexibly between two sequences. We used computer simulations to explore potential mechanisms underlying these effects, which revealed that the DRD1 and DRD2 effects are possibly related to neuroplasticity. Our prediction-error algorithm estimated faster rates of connection strengthening in genotype groups with presumably higher D1 receptor densities, and faster rates of connection weakening in genotype groups with presumably higher D2 receptor densities. Consistent with current dopamine theories, these simulations suggest that D1-mediated neuroplasticity contributes to learning to select appropriate actions, whereas D2-mediated neuroplasticity is involved in learning to inhibit incorrect action plans. However, the learning algorithm did not account for the COMT effect, suggesting that prefrontal dopamine availability might affect sequence switching via other, non-learning, mechanisms. These findings provide insight into the function of the dopamine system, which is relevant to the development of treatments for disorders such as Parkinson's disease. Our results suggest that treatments targeting dopamine D1 receptors may improve learning of novel sequences, whereas those targeting dopamine D2 receptors may improve the ability to initiate previously learned sequences of movements.

Falling for phishing attempts: An investigation of individual differences that are associated with behavior in a naturalistic phishing simulation

Publisher: Center for Open Science

Date: 08-11-2022

DOI: 10.31234/OSF.IO/XDK53

Abstract: Social engineering cyber-attacks such as phishing emails pose a serious threat to the safety of many organizations. Given that the effectiveness of these attacks heavily relies on poor human decision making, an improved understanding of the in idual characteristics that increase cybersecurity vulnerability could inform more targeted training. The current study aimed to identify whether several factors, including phishing email detection ability, confidence in one’s phishing identification decisions, attitudes toward one’s level of responsibility and efficacy, and employee satisfaction and loyalty to the organization, can predict behavior in a naturalistic phishing simulation in an employment setting. We followed up employees of a large organization who had been recently targeted by a phishing simulation and asked them to complete a survey that included a phishing detection task. The employee’s behavior in the phishing simulation was ranked according to its safety: reporting the suspicious email, neither reporting nor clicking on the embedded link, and clicking on the link. We found that fewer years of employment at the organization and lower employee satisfaction and loyalty predicted increasingly unsafe behavior in the simulation. This suggests that newer and unsatisfied employees are most vulnerable to phishing attempts and might benefit most from targeted cybersecurity training.

Human Learning About Causation

Publisher: Wiley

Date: 24-06-2016

DOI: 10.1002/9781118650813.CH15

Falling for phishing attempts: An investigation of individual differences that are associated with behavior in a naturalistic phishing simulation

Publisher: Elsevier BV

Date: 08-2023

DOI: 10.1016/J.COSE.2023.103313

Individual differences in anxiety and fear learning: The role of working memory capacity

Publisher: Elsevier BV

Date: 02-2019

DOI: 10.1016/J.ACTPSY.2018.12.006

Abstract: Anxiety disorders are characterised by the perception of fear and threat in the presence of stimuli that are neutral or ambiguous. Attempts in previous research to explain the relationship between anxiety and fear learning have been inconsistent, possibly due to the influence of an unmeasured mechanism that mediates the relationship between them. Working memory capacity has been suggested as one such mechanism. The current study investigated the influence of anxiety-based in idual differences upon associative fear learning, while accounting for in idual differences in working memory. We hypothesised that in iduals high in both anxiety and working memory would show unimpaired fear learning whereas in iduals high in anxiety and low in working memory would exhibit dysfunctional fear learning. Sixty participants completed a battery of anxiety and working memory tests, as well as a fear conditioning experiment that tested for blocking, conditioned inhibition and fear discrimination. We found that anxious in iduals were more likely to show impaired fear discrimination only if they also had a low working memory capacity. Furthermore, anxiety was particularly associated with poorer learning about safety cues. Such relationships were not observed for blocking and conditioned inhibition. These results suggest that the relationship between anxiety and fear learning is complex and warrants further investigation of the potential mediating role of higher-order cognitive faculties.

Reasoning about redundant and non-redundant alternative causes of a single outcome: Blocking or enhancement caused by the stronger cause

Publisher: SAGE Publications

Date: 2018

DOI: 10.1080/17470218.2017.1338302

Abstract: Perceptions of the effectiveness of a moderate probabilistic cause are influenced by the presence of stronger alternative causes. One important idea is that this influence occurs because the strong cause renders the weaker one statistically redundant. Alternatively, the causes might be contrasted to each other, so the stronger cause may simply overpower perceptions of the weaker one. Causes may have the same polarity (e.g., two generative/excitatory causes or two preventive/inhibitory causes) or be of opposite polarity (e.g., a generative cause versus a preventive or inhibitory cause). Previously, we found that the presence of a stronger redundant alternative cause of the same polarity reduces causal judgements of the moderate cause (i.e., blocking occurs) but a stronger cause of the opposite polarity enhances judgements of the moderate cause (i.e., enhancement). Experiments 1 and 2 further explored these cue competition effects with redundant and non-redundant alternative causes (i.e., correlated versus independent alternatives). We generally found that blocking and enhancement occur with both redundant and non-redundant alternative causes. This is inconsistent with an information processing view of cue competition that relies on statistical redundancy to account for blocking. Although these results are inconsistent with a redundancy information processing account of cue competition and are consistent with our earlier contrast account, we demonstrate here that a simple associative model can account for the sometimes apparently contradictory effects of cue competition.

Competition between multiple causes of a single outcome in causal reasoning.

Publisher: American Psychological Association (APA)

Date: 2009

DOI: 10.1037/A0012699

Abstract: A strong positive predictor of an outcome modulates the causal judgments of a moderate predictor. To study the empirical basis of this modulation, we compared treatments with one and with two strong competing (i.e., modulating) causes. This allowed us to vary the frequency of outcome occurrences or effects paired with the predictors. We investigated causal competition between positive predictors (those signaling the occurrence of the outcome), between negative predictors (those signaling the absence of the outcome) and between predictors of opposite polarity (positive and negative). The results are consistent with a contrast rather than a reduced associative strength or conditional contingency account, because a strong predictor of opposite polarity enhances rather than reduces causal estimates of moderate predictors. In addition, we found competition effects when the strong predictor predicted fewer outcome occurrences than the moderate predictor, thus implying that cue competition is, at least sometimes, a consequence of contingency rather than total cue-outcome pairings.

Human judgments of positive and negative causal chains.

Publisher: American Psychological Association (APA)

Date: 2009

DOI: 10.1037/A0013764

Abstract: Three experiments investigated the way participants construct causal chains from experience with the in idual links that make up those chains. Participants were presented with contingency information about the relationship between events A and B, as well as events B and C, using trial-by-trial presentations. The A-B and B-C contingencies could be positive, negative, or zero. Although participants had never experienced A and C together, A-C ratings were a multiplicative function of the A-B and B-C contingencies. These findings can be generated by an auto-associator using the delta rule. This explanation is also useful for understanding sensory preconditioning and second-order conditioning.

Neural indices of associative learning in pre-adolescents: An event-related potential study

Publisher: Elsevier BV

Date: 03-2019

DOI: 10.1016/J.BANDC.2018.12.006

Abstract: This study investigated electroencephalography (EEG) correlates of prediction error during probabilistic learning in pre-adolescents. The detection of prediction errors, the discrepancies between experienced and anticipated outcomes, is thought to be a critical mechanism that drives new learning. Thirty-three typically developing pre-adolescents (mean age = 10.62 years) participated in an associative learning task in which they learned the probabilistic relationships between cues and outcome stimuli in the absence of explicit feedback. We investigated whether three outcome-locked event-related potentials (ERPs) could reflect prediction error processing: the P3, the late positive potential (LPP), and the feedback-related negativity (FRN). All ERP components investigated were sensitive to the magnitude of hypothetical prediction errors that were estimated based on each in idual's learning performance. Higher estimated prediction errors generated larger P3 and LPP components, and a more negative FRN. These findings indicate that pre-adolescents are capable of undergoing probabilistic learning in the absence of explicit feedback, much in the same way as adults, and that prediction error processing is physiologically indexed via the FRN, P3 and LPP following outcome stimuli.

Polymorphisms that affect GABA neurotransmission predict processing of aversive prediction errors in humans

Publisher: Elsevier BV

Date: 08-2018

DOI: 10.1016/J.NEUROIMAGE.2018.04.058

Abstract: Learning is one of our most adaptive abilities, allowing us to adjust our expectations about future events. Aberrant learning processes may underlie disorders such as anxiety, motivating the search for the neural mechanisms that underpin learning. Animal studies have shown that the neurotransmitter GABA is required for the computation of prediction errors, the mismatches between anticipated and experienced outcomes, which drive new learning. Given that evidence from human studies is lacking, we sought to determine whether these findings extend to humans. Here, in two s les of Caucasian in iduals, we investigated whether genetically determined in idual differences in GABA neurotransmission predict the P3 event-related potential, an EEG component known to reflect prediction error processing. Consistent with the results of animal studies, we show that a weighted genetic risk score computed from the number of GABRB2 rs1816072 A alleles (associated with increased expression of the GABA

Maladaptive avoidance patterns in Parkinson’s disease are exacerbated by symptoms of depression

Publisher: Elsevier BV

Date: 03-2020

DOI: 10.1016/J.BBR.2020.112473

New Directions in Human Associative Learning

Publisher: Psychology Press

Date: 15-01-2005

DOI: 10.4324/9781410612113

Fluid Abilities and Rule Learning: Patterning and Biconditional Discriminations

Publisher: MDPI AG

Date: 27-02-2018

DOI: 10.3390/JINTELLIGENCE6010007

McGill University

Location: Canada

View Organisation

University Of Adelaide

Location: Australia

View Organisation

University Of Cambridge

Location: United Kingdom of Great Britain and Northern Ireland

View Organisation

Investigating Differences In Decision-Making Ability In Older Adults: Computational Modelling And Neurogenetics Of The Basal Ganglia

Start Date: 2019

End Date: 2022

Funder: Australian Research Council

Associative Learning And Fluid Intelligence: Computational And Neurogenetic Analyses

Start Date: 2014

End Date: 2017

Funder: Australian Research Council

Discovery Projects - Grant ID: DP190103600

Start Date: 2019

End Date: 12-2023

Amount: $443,000.00

Funder: Australian Research Council

Discovery Early Career Researcher Award - Grant ID: DE140100750

Start Date: 2014

End Date: 02-2017

Amount: $395,106.00

Funder: Australian Research Council