Ismail Guennouni
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Ismail Guennouni has a total of 28 co-authors. Here are the top co-authors by number of shared publications:
Georgia Koppe
Interdisciplinary Center for Scientific Computing, Heidelberg University
5 shared publicationsChristoph Korn
2 shared publicationsDaniel Durstewitz
Central Institute of Mental Health (CIMH)
2 shared publicationsMaarten Speekenbrink
2 shared publications- Maarten Speekenbrink
Experimental Psychology, University College London
2 shared publications - Manuel Brenner
Ernst-Strüngmann-Institut (ESI)
2 shared publications
Latest Publications
- Open Access
- 17 November 2026
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 22 (11) , e1014940
Hierarchical data-driven dynamical surrogate models of social decision-making
Social decision making is inherently complex, shaped by dynamic trust building, biases, and adaptive strategies. Yet in laboratory settings, its study is often constrained by small datasets that nonetheless encode rich and sophisticated cognitive processes. This scarcity has historically limited modeling approaches to hand-tailored frameworks that embed strong priors about the underlying mechanisms. Recent advances in data-driven modeling of latent dynamical processes provide an alternative, extracting generative models directly from behavioral data without restrictive assumptions. Building on these methods, we derive hierarchical digital surrogates from sparse, non-Gaussian investment sequences in repeated trust games. Our approach proceeds in three stages. First, we demonstrate that, despite the limited data, the inferred models can accurately predict future investments and outperform theory-driven reinforcement-learning and Bayesian cognitive models, establishing predictive validity and providing a solid baseline for subsequent mechanistic analyses. Second, we analyze the latent dynamics of these models and show that they capture mechanistic structure in how investments and choice uncertainty are organized in state space and how social versus non-social cues are represented, in a way that parallels patterns observed directly in the empirical behavior. Third, we exploit the generative nature of the hierarchical digital surrogates to run in-silico experiments beyond the original data, such as probing how specific cues steer participants into trusting or mistrusting states and simulating responses to novel, unobserved cue-outcome combinations and trustee strategies. Together, this work demonstrates how generative digital surrogates open a new path toward mechanistic, data-driven accounts of social interaction dynamics that both reproduce observed behavior and support hypothesis generation in virtual environments.
- 17 November 2026
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 22 (11) , e1014940
Open AccessHierarchical data-driven dynamical surrogate models of social decision-making
Social decision making is inherently complex, shaped by dynamic trust building, biases, and adaptive strategies. Yet in laboratory settings, its study is often constrained by small datasets that nonetheless encode rich and sophisticated cognitive processes. This scarcity has historically limited modeling approaches to hand-tailored frameworks that embed strong priors about the underlying mechanisms. Recent advances in data-driven modeling of latent dynamical processes provide an alternative, extracting generative models directly from behavioral data without restrictive assumptions. Building on these methods, we derive hierarchical digital surrogates from sparse, non-Gaussian investment sequences in repeated trust games. Our approach proceeds in three stages. First, we demonstrate that, despite the limited data, the inferred models can accurately predict future investments and outperform theory-driven reinforcement-learning and Bayesian cognitive models, establishing predictive validity and providing a solid baseline for subsequent mechanistic analyses. Second, we analyze the latent dynamics of these models and show that they capture mechanistic structure in how investments and choice uncertainty are organized in state space and how social versus non-social cues are represented, in a way that parallels patterns observed directly in the empirical behavior. Third, we exploit the generative nature of the hierarchical digital surrogates to run in-silico experiments beyond the original data, such as probing how specific cues steer participants into trusting or mistrusting states and simulating responses to novel, unobserved cue-outcome combinations and trustee strategies. Together, this work demonstrates how generative digital surrogates open a new path toward mechanistic, data-driven accounts of social interaction dynamics that both reproduce observed behavior and support hypothesis generation in virtual environments.
- Open Access
- 27 October 2026
- Published by Center for Open Science in PsyArXiv
Beyond Average Behaviour: Trust and Reciprocity Dynamics Reveal Distinct Behavioural Phenotypes
Cooperation depends not only on how much people trust and reciprocate, but on how it is maintained and repaired after rupture. We analysed a 10-round Trust Game in participants with borderline personality disorder (BPD), an affective-symptom sample (AD), and community controls (N = 1,300 participants contributing 1,310 games; BPD n = 468, AD n = 556, controls n = 286 games), using mixed-effects and hidden Markov models. Average behaviour was broadly similar across groups, although AD trustees returned more than controls. As investors, BPD participants adjusted less to feedback than controls. Among trustees, AD differed most clearly from BPD: AD trustees followed cooperative response patterns more often, resumed cooperation more readily after exploitative returns, and had higher model-implied long-run cooperation at low and moderate partner investments. The occupancy and repair differences persisted when hidden states were estimated without diagnostic information, but not when BPD was restricted to interview-confirmed cases. AD–control differences in cooperative patterns did not persist among participants tested online. BPD–control differences in maintaining and restoring cooperation were smaller and depended on investment and hidden-state specification. Clinical differences were more evident in interaction dynamics than in average behaviour.
- 27 October 2026
- Published by Center for Open Science in PsyArXiv
Open AccessBeyond Average Behaviour: Trust and Reciprocity Dynamics Reveal Distinct Behavioural Phenotypes
Cooperation depends not only on how much people trust and reciprocate, but on how it is maintained and repaired after rupture. We analysed a 10-round Trust Game in participants with borderline personality disorder (BPD), an affective-symptom sample (AD), and community controls (N = 1,300 participants contributing 1,310 games; BPD n = 468, AD n = 556, controls n = 286 games), using mixed-effects and hidden Markov models. Average behaviour was broadly similar across groups, although AD trustees returned more than controls. As investors, BPD participants adjusted less to feedback than controls. Among trustees, AD differed most clearly from BPD: AD trustees followed cooperative response patterns more often, resumed cooperation more readily after exploitative returns, and had higher model-implied long-run cooperation at low and moderate partner investments. The occupancy and repair differences persisted when hidden states were estimated without diagnostic information, but not when BPD was restricted to interview-confirmed cases. AD–control differences in cooperative patterns did not persist among participants tested online. BPD–control differences in maintaining and restoring cooperation were smaller and depended on investment and hidden-state specification. Clinical differences were more evident in interaction dynamics than in average behaviour.
- Open Access
- 30 September 2026
- Published by Center for Open Science in PsyArXiv
When Positive Relational Experiences Backfire: Exposure to Forgiving, Consistently Trusting Agents Lowers Subsequent Reciprocity
Can exposure to a forgiving, consistently trusting partner improve interpersonal cooperation? Attachment theory suggests positive relational experiences can correct negative internal working models, but individuals high in rejection sensitivity (RS), characterized by anxious expectations of rejection, may be resistant to such corrective experiences due to stable negative expectations. We tested this using a randomized experiment (N = 206) in which participants played repeated trust games with HMM-based artificial agents that simulate human-like trust dynamics. After a baseline game, participants were exposed to either an idealized partner (an agent designed never to remain in a low-trust state, free from pre-programmed trust withdrawals, and consequently investing more; Manipulation) or a standard agent maintaining typical trust withdrawal patterns (Control), then played a final game with the standard agent; these features were varied together, so the design tests their joint effect. In the primary, powered contrast, exposure to the idealized partner reduced subsequent cooperation, and participants rated the unchanged post-exposure agent less favorably, a pattern consistent with a negative contrast effect, an interpretation developed after initial analysis. RS moderation was a planned analysis but was not separately powered: we detected no moderation of the reduction by RS, and an apparent RS difference in responsiveness to partner investment was not robust once participant-level variation in responsiveness was modeled. These findings suggest that brief exposure to an idealized partner does not universally promote cooperation and can reduce it, and point to expectation-updating, rather than exposure alone, as one candidate target for future interventions.
- 30 September 2026
- Published by Center for Open Science in PsyArXiv
Open AccessWhen Positive Relational Experiences Backfire: Exposure to Forgiving, Consistently Trusting Agents Lowers Subsequent Reciprocity
Can exposure to a forgiving, consistently trusting partner improve interpersonal cooperation? Attachment theory suggests positive relational experiences can correct negative internal working models, but individuals high in rejection sensitivity (RS), characterized by anxious expectations of rejection, may be resistant to such corrective experiences due to stable negative expectations. We tested this using a randomized experiment (N = 206) in which participants played repeated trust games with HMM-based artificial agents that simulate human-like trust dynamics. After a baseline game, participants were exposed to either an idealized partner (an agent designed never to remain in a low-trust state, free from pre-programmed trust withdrawals, and consequently investing more; Manipulation) or a standard agent maintaining typical trust withdrawal patterns (Control), then played a final game with the standard agent; these features were varied together, so the design tests their joint effect. In the primary, powered contrast, exposure to the idealized partner reduced subsequent cooperation, and participants rated the unchanged post-exposure agent less favorably, a pattern consistent with a negative contrast effect, an interpretation developed after initial analysis. RS moderation was a planned analysis but was not separately powered: we detected no moderation of the reduction by RS, and an apparent RS difference in responsiveness to partner investment was not robust once participant-level variation in responsiveness was modeled. These findings suggest that brief exposure to an idealized partner does not universally promote cooperation and can reduce it, and point to expectation-updating, rather than exposure alone, as one candidate target for future interventions.
- Open Access
- 29 April 2026
- Published by The Royal Society in Royal Society Open Science
- Vol. 13 (4)
The price of dark traits: strategic exploitation and its limitations in repeated interactions
Trust and cooperation are fundamental to human social interaction, with personality traits significantly influencing economic decision making. This study introduces a novel approach using adaptive computational agents to provide the first examination of how the dark factor of personality (D-factor)—the unifying core of malevolent traits—affects trustworthiness and strategic exploitation in dynamic, repeated trust games. After pre-screening 1243 participants, we selected individuals with high and low D-factor scores. Participants (n = 183) played two 25-round trust games as trustees against programmed hidden Markov model (HMM) investors derived from human data but differing in strategic reactivity: one ‘human-like’ HMM exhibited a characteristic ‘sticky’ mid-trust state (potentially vulnerable to exploitation), while a ‘responsive’ HMM reacted more decisively and immediately to trustee returns. Consistent with the D-factor concept, high-D participants returned lower proportions, particularly later in the interactions. High-D individuals also demonstrated sophisticated strategic adaptation, significantly decreasing reciprocity over time only when receiving high investments from the more passively exploitable ‘human-like’ opponent. Crucially, this exploitative strategy proved self-limiting: high-D participants did not achieve higher total payoffs overall, as the ‘responsive’ HMM adaptively reduced investments. These findings uniquely demonstrate how D-factor manifests strategically in dynamic exchanges and reveal the critical role of opponent responsiveness in constraining exploitation, highlighting both the sophistication and the boundaries of dark personality expression in social interactions.
- 29 April 2026
- Published by The Royal Society in Royal Society Open Science
- Vol. 13 (4)
Open AccessThe price of dark traits: strategic exploitation and its limitations in repeated interactions
Trust and cooperation are fundamental to human social interaction, with personality traits significantly influencing economic decision making. This study introduces a novel approach using adaptive computational agents to provide the first examination of how the dark factor of personality (D-factor)—the unifying core of malevolent traits—affects trustworthiness and strategic exploitation in dynamic, repeated trust games. After pre-screening 1243 participants, we selected individuals with high and low D-factor scores. Participants (n = 183) played two 25-round trust games as trustees against programmed hidden Markov model (HMM) investors derived from human data but differing in strategic reactivity: one ‘human-like’ HMM exhibited a characteristic ‘sticky’ mid-trust state (potentially vulnerable to exploitation), while a ‘responsive’ HMM reacted more decisively and immediately to trustee returns. Consistent with the D-factor concept, high-D participants returned lower proportions, particularly later in the interactions. High-D individuals also demonstrated sophisticated strategic adaptation, significantly decreasing reciprocity over time only when receiving high investments from the more passively exploitable ‘human-like’ opponent. Crucially, this exploitative strategy proved self-limiting: high-D participants did not achieve higher total payoffs overall, as the ‘responsive’ HMM adaptively reduced investments. These findings uniquely demonstrate how D-factor manifests strategically in dynamic exchanges and reveal the critical role of opponent responsiveness in constraining exploitation, highlighting both the sophistication and the boundaries of dark personality expression in social interactions.
- Open Access
- 16 January 2026
- Published by Ubiquity Press, Ltd. in Computational Psychiatry
- Vol. 10 (1) , 184-201
(Re)building Cooperation: Effects of a Cognitive Intervention on Cooperative Behaviour in Games
Cooperation often breaks down when one party appears to withdraw trust, yet most research has focused on initial trust formation rather than how the aggrieved party can resist the impulse to retaliate. In a randomised controlled online trial (N = 318), we tested whether a brief cognitive intervention inspired by Dialectical Behaviour Therapy principles could promote resilient cooperation in the face of trust withdrawals. Participants played a Repeated Trust Game as trustees against adaptive computerised investors driven by hidden Markov models (HMMs) trained on human data, providing behavioural realism under full experimental control. Following the intervention, participants returned a significantly higher proportion of the amount received than an active control group across the post-manipulation game. The groups diverged most sharply at the pre-programmed low investment, where the control group retaliated by reducing returns while the intervention group maintained the same level of returns. This shift occurred without a change in self-reported emotional reactivity, dissociating the preserved emotional reaction from the retaliatory behaviour it drives. HMM analysis of participants’ own return patterns revealed that participants in the intervention condition were less likely to become stuck in low-return behavioural states following the pre-programmed low investment, though a notable minority did not respond to the intervention. The intervention effect did not transfer to a Prisoner’s Dilemma game, consistent with context-specific effects. These findings provide proof-of-concept evidence that targeted cognitive strategies can support cooperation after trust violations, and show that HMM-based agents are a useful tool for studying social behaviour under controlled conditions.
- 16 January 2026
- Published by Ubiquity Press, Ltd. in Computational Psychiatry
- Vol. 10 (1) , 184-201
Open Access(Re)building Cooperation: Effects of a Cognitive Intervention on Cooperative Behaviour in Games
Cooperation often breaks down when one party appears to withdraw trust, yet most research has focused on initial trust formation rather than how the aggrieved party can resist the impulse to retaliate. In a randomised controlled online trial (N = 318), we tested whether a brief cognitive intervention inspired by Dialectical Behaviour Therapy principles could promote resilient cooperation in the face of trust withdrawals. Participants played a Repeated Trust Game as trustees against adaptive computerised investors driven by hidden Markov models (HMMs) trained on human data, providing behavioural realism under full experimental control. Following the intervention, participants returned a significantly higher proportion of the amount received than an active control group across the post-manipulation game. The groups diverged most sharply at the pre-programmed low investment, where the control group retaliated by reducing returns while the intervention group maintained the same level of returns. This shift occurred without a change in self-reported emotional reactivity, dissociating the preserved emotional reaction from the retaliatory behaviour it drives. HMM analysis of participants’ own return patterns revealed that participants in the intervention condition were less likely to become stuck in low-return behavioural states following the pre-programmed low investment, though a notable minority did not respond to the intervention. The intervention effect did not transfer to a Prisoner’s Dilemma game, consistent with context-specific effects. These findings provide proof-of-concept evidence that targeted cognitive strategies can support cooperation after trust violations, and show that HMM-based agents are a useful tool for studying social behaviour under controlled conditions.
- Open Access
- 14 January 2026
- Published by Center for Open Science in PsyArXiv
Data-Driven Hierarchical Digital Twins of Social Interactions
Social interactions are inherently complex, shaped by dynamic trust building, biases, and adaptive strategies. Yet in laboratory settings, their study is often constrained by small datasets that nonetheless encode rich and sophisticated cognitive processes. This scarcity has historically limited modeling approaches to hand-tailored frameworks that embed strong priors about the underlying mechanisms. Recent advances in data-driven modeling of latent dynamical processes provide an alternative, extracting generative models directly from behavioral data without restrictive assumptions. Building on these methods, we derive hierarchical digital twins from sparse, non-Gaussian investment sequences in repeated trust games. Our approach proceeds in three stages. First, we demonstrate that, despite the limited data, the inferred twins can accurately predict future investments, establishing predictive validity and providing a solid baseline for subsequent mechanistic analyses. Second, we analyze the latent dynamics of these models and show that they capture mechanistic structure in how investments and choice uncertainty are organized in state space and how social versus non-social cues are represented, in a way that parallels patterns observed directly in the empirical behavior. Third, we exploit the generative nature of the hierarchical digital twins to run in-silico experiments beyond the original data, such as probing how specific cues steer participants into trusting or mistrusting states and simulating responses to novel, unobserved cue-outcome combinations and trustee strategies. Together, this work demonstrates how generative digital twins open a new path toward mechanistic, data-driven accounts of social interaction dynamics that both reproduce observed behavior and support hypothesis generation in virtual environments.
- 14 January 2026
- Published by Center for Open Science in PsyArXiv
Open AccessData-Driven Hierarchical Digital Twins of Social Interactions
Social interactions are inherently complex, shaped by dynamic trust building, biases, and adaptive strategies. Yet in laboratory settings, their study is often constrained by small datasets that nonetheless encode rich and sophisticated cognitive processes. This scarcity has historically limited modeling approaches to hand-tailored frameworks that embed strong priors about the underlying mechanisms. Recent advances in data-driven modeling of latent dynamical processes provide an alternative, extracting generative models directly from behavioral data without restrictive assumptions. Building on these methods, we derive hierarchical digital twins from sparse, non-Gaussian investment sequences in repeated trust games. Our approach proceeds in three stages. First, we demonstrate that, despite the limited data, the inferred twins can accurately predict future investments, establishing predictive validity and providing a solid baseline for subsequent mechanistic analyses. Second, we analyze the latent dynamics of these models and show that they capture mechanistic structure in how investments and choice uncertainty are organized in state space and how social versus non-social cues are represented, in a way that parallels patterns observed directly in the empirical behavior. Third, we exploit the generative nature of the hierarchical digital twins to run in-silico experiments beyond the original data, such as probing how specific cues steer participants into trusting or mistrusting states and simulating responses to novel, unobserved cue-outcome combinations and trustee strategies. Together, this work demonstrates how generative digital twins open a new path toward mechanistic, data-driven accounts of social interaction dynamics that both reproduce observed behavior and support hypothesis generation in virtual environments.
- Open Access
- 14 November 2025
- Published by Center for Open Science in PsyArXiv
The Price of Dark Traits: Strategic Exploitation and Its Limitations in Repeated Interactions
Trust and cooperation are fundamental to human social interaction, with personality traits significantly influencing economic decision-making.This study introduces a novel approach using adaptive computational agents to provide the first examination of how the Dark Factor of Personality (D-factor)—the unifying core of malevolent traits—affects trustworthiness and strategic exploitation in dynamic, repeated trust games.After pre-screening 1,243 participants, we selected individuals with high and low D-factor scores.Participants (N=183) played two 25-round trust games as trustees against programmed Hidden Markov Model (HMM) investors derived from human data but differing in strategic reactivity: one 'human-like' HMM exhibited a characteristic 'sticky' mid-trust state (potentially vulnerable to exploitation), while a 'responsive' HMM reacted more decisively and immediately to trustee returns.Consistent with the D-factor concept, high-D participants returned lower proportions, particularly later in the interactions.High-D individuals also demonstrated sophisticated strategic adaptation, significantly decreasing reciprocity over time only when receiving high investments from the more passively exploitable 'human-like' opponent.Crucially, this exploitative strategy proved self-limiting: high-D participants did not achieve higher total payoffs overall, as the 'responsive' HMM adaptively reduced investments.These findings uniquely demonstrate how D-factor manifests strategically in dynamic exchanges and reveal the critical role of opponent responsiveness in constraining exploitation, highlighting both the sophistication and the boundaries of dark personality expression in social interactions.
- 14 November 2025
- Published by Center for Open Science in PsyArXiv
Open AccessThe Price of Dark Traits: Strategic Exploitation and Its Limitations in Repeated Interactions
Trust and cooperation are fundamental to human social interaction, with personality traits significantly influencing economic decision-making.This study introduces a novel approach using adaptive computational agents to provide the first examination of how the Dark Factor of Personality (D-factor)—the unifying core of malevolent traits—affects trustworthiness and strategic exploitation in dynamic, repeated trust games.After pre-screening 1,243 participants, we selected individuals with high and low D-factor scores.Participants (N=183) played two 25-round trust games as trustees against programmed Hidden Markov Model (HMM) investors derived from human data but differing in strategic reactivity: one 'human-like' HMM exhibited a characteristic 'sticky' mid-trust state (potentially vulnerable to exploitation), while a 'responsive' HMM reacted more decisively and immediately to trustee returns.Consistent with the D-factor concept, high-D participants returned lower proportions, particularly later in the interactions.High-D individuals also demonstrated sophisticated strategic adaptation, significantly decreasing reciprocity over time only when receiving high investments from the more passively exploitable 'human-like' opponent.Crucially, this exploitative strategy proved self-limiting: high-D participants did not achieve higher total payoffs overall, as the 'responsive' HMM adaptively reduced investments.These findings uniquely demonstrate how D-factor manifests strategically in dynamic exchanges and reveal the critical role of opponent responsiveness in constraining exploitation, highlighting both the sophistication and the boundaries of dark personality expression in social interactions.
- Open Access
- 5 November 2025
- Published by Center for Open Science in PsyArXiv
(Re)building cooperation: Effects of a cognitive intervention on cooperative behavior in games
Restoring cooperation after trust breaches remains a major hurdle in social, clinical, and organizational contexts. Yet, most interventions focus on initial trust formation rather than repair. We introduce a novel, online cognitive intervention—rooted in Dialectical Behavior Therapy—to reduce psychological reactivity and promote resilient cooperation. In a randomized controlled online trial (N = 318), participants played a Repeated Trust Game against an HMM-based adaptive agent trained on human data, ensuring dynamic, human-like responses under full experimental control. Compared to an active control, the cognitive intervention prevented retaliation following programmed trust violations and yielded significantly higher cooperative returns. Mixed-effects analysis revealed that while control participants became increasingly reciprocal over time (approaching tit-for-tat strategies), intervention participants maintained more stable cooperative behavior that was less contingent on partner investment levels. This change did not generalize to Prisoner's Dilemma, underscoring the intervention's specificity. These results validate HMM agents as a powerful experimental tool and suggest that targeted cognitive strategies may help support cooperative behavior; clinical efficacy and real-world deployment remain to be established.
- 5 November 2025
- Published by Center for Open Science in PsyArXiv
Open Access(Re)building cooperation: Effects of a cognitive intervention on cooperative behavior in games
Restoring cooperation after trust breaches remains a major hurdle in social, clinical, and organizational contexts. Yet, most interventions focus on initial trust formation rather than repair. We introduce a novel, online cognitive intervention—rooted in Dialectical Behavior Therapy—to reduce psychological reactivity and promote resilient cooperation. In a randomized controlled online trial (N = 318), participants played a Repeated Trust Game against an HMM-based adaptive agent trained on human data, ensuring dynamic, human-like responses under full experimental control. Compared to an active control, the cognitive intervention prevented retaliation following programmed trust violations and yielded significantly higher cooperative returns. Mixed-effects analysis revealed that while control participants became increasingly reciprocal over time (approaching tit-for-tat strategies), intervention participants maintained more stable cooperative behavior that was less contingent on partner investment levels. This change did not generalize to Prisoner's Dilemma, underscoring the intervention's specificity. These results validate HMM agents as a powerful experimental tool and suggest that targeted cognitive strategies may help support cooperative behavior; clinical efficacy and real-world deployment remain to be established.
- Top Cited PapersOpen Access
Fatigue outcomes following COVID-19: a systematic review and meta-analysis
Objectives: Fatigue is a pervasive clinical symptom in coronaviruses and may continue beyond the acute phase, lasting for several months or years. This systematic review and meta-analysis aimed to incorporate the current evidence for postinfection fatigue among survivors of SARS-CoV-2 and investigate associated factors. Methods: Embase, PsyINFO, Medline, CINAHL, CDSR, Open Grey, BioRxiv and MedRxiv were systematically searched from January 2019 to December 2021. Eligible records included all study designs in English. Outcomes were fatigue or vitality in adults with a confirmed diagnosis of SARS-CoV-2 measured at >30 days post infection. Non-confirmed cases were excluded. JBI risk of bias was assessed by three reviewers. Random effects model was used for the pooled proportion with 95% CIs. A mixed effects meta-regression of 35 prospective articles calculated change in fatigue overtime. Subgroup analyses explored specific group characteristics of study methodology. Heterogeneity was assessed using Cochran’s Q and I 2 statistic. Egger’s tests for publication bias. Results: Database searches returned 14 262 records. Following deduplication and screening, 178 records were identified. 147 (n=48 466 participants) were included for the meta-analyses. Pooled prevalence was 41% (95% CI: 37% to 45%, k=147, I 2 =98%). Fatigue significantly reduced over time (−0.057, 95% CI: −107 to −0.008, k=35, I 2 =99.3%, p=0.05). A higher proportion of fatigue was found in studies using a valid scale (51%, 95% CI: 43% to 58%, k=36, I 2 =96.2%, p=0.004). No significant difference was found for fatigue by study design (p=0.272). Egger’s test indicated publication bias for all analyses except valid scales. Quality assessments indicated 4% at low risk of bias, 78% at moderate risk and 18% at high risk. Frequently reported associations were female gender, age, physical functioning, breathlessness and psychological distress. Conclusion: This study revealed that a significant proportion of survivors experienced fatigue following SARS-CoV-2 and their fatigue reduced overtime. Non-modifiable factors and psychological morbidity may contribute to ongoing fatigue and impede recovery. PROSPERO registration number: CRD42020201247.
Top Cited PapersOpen AccessFatigue outcomes following COVID-19: a systematic review and meta-analysis
Objectives: Fatigue is a pervasive clinical symptom in coronaviruses and may continue beyond the acute phase, lasting for several months or years. This systematic review and meta-analysis aimed to incorporate the current evidence for postinfection fatigue among survivors of SARS-CoV-2 and investigate associated factors. Methods: Embase, PsyINFO, Medline, CINAHL, CDSR, Open Grey, BioRxiv and MedRxiv were systematically searched from January 2019 to December 2021. Eligible records included all study designs in English. Outcomes were fatigue or vitality in adults with a confirmed diagnosis of SARS-CoV-2 measured at >30 days post infection. Non-confirmed cases were excluded. JBI risk of bias was assessed by three reviewers. Random effects model was used for the pooled proportion with 95% CIs. A mixed effects meta-regression of 35 prospective articles calculated change in fatigue overtime. Subgroup analyses explored specific group characteristics of study methodology. Heterogeneity was assessed using Cochran’s Q and I 2 statistic. Egger’s tests for publication bias. Results: Database searches returned 14 262 records. Following deduplication and screening, 178 records were identified. 147 (n=48 466 participants) were included for the meta-analyses. Pooled prevalence was 41% (95% CI: 37% to 45%, k=147, I 2 =98%). Fatigue significantly reduced over time (−0.057, 95% CI: −107 to −0.008, k=35, I 2 =99.3%, p=0.05). A higher proportion of fatigue was found in studies using a valid scale (51%, 95% CI: 43% to 58%, k=36, I 2 =96.2%, p=0.004). No significant difference was found for fatigue by study design (p=0.272). Egger’s test indicated publication bias for all analyses except valid scales. Quality assessments indicated 4% at low risk of bias, 78% at moderate risk and 18% at high risk. Frequently reported associations were female gender, age, physical functioning, breathlessness and psychological distress. Conclusion: This study revealed that a significant proportion of survivors experienced fatigue following SARS-CoV-2 and their fatigue reduced overtime. Non-modifiable factors and psychological morbidity may contribute to ongoing fatigue and impede recovery. PROSPERO registration number: CRD42020201247.
- Open Access
- 31 May 2022
- Published by Springer Nature in Computational Brain & Behavior
- Vol. 5 (3) , 326-342
Transfer of Learned Opponent Models in Zero Sum Games
Human learning transfer abilities take advantage of important cognitive building blocks such as an abstract representation of concepts underlying tasks and causal models of the environment. One way to build abstract representations of the environment when the task involves interactions with others is to build a model of the opponent that may inform what actions they are likely to take next. In this study, we explore opponent modelling and its transfer in games where human agents play against computer agents with human-like limited degrees of iterated reasoning. In two experiments, we find that participants deviate from Nash equilibrium play and learn to adapt to their opponent’s strategy to exploit it. Moreover, we show that participants transfer their learning to new games. Computational modelling shows that players start each game with a model-based learning strategy that facilitates between-game transfer of their opponent’s strategy, but then switch to behaviour that is consistent with a model-free learning strategy in the latter stages of the interaction.
- 31 May 2022
- Published by Springer Nature in Computational Brain & Behavior
- Vol. 5 (3) , 326-342
Open AccessTransfer of Learned Opponent Models in Zero Sum Games
Human learning transfer abilities take advantage of important cognitive building blocks such as an abstract representation of concepts underlying tasks and causal models of the environment. One way to build abstract representations of the environment when the task involves interactions with others is to build a model of the opponent that may inform what actions they are likely to take next. In this study, we explore opponent modelling and its transfer in games where human agents play against computer agents with human-like limited degrees of iterated reasoning. In two experiments, we find that participants deviate from Nash equilibrium play and learn to adapt to their opponent’s strategy to exploit it. Moreover, we show that participants transfer their learning to new games. Computational modelling shows that players start each game with a model-based learning strategy that facilitates between-game transfer of their opponent’s strategy, but then switch to behaviour that is consistent with a model-free learning strategy in the latter stages of the interaction.