Manuel Lopes
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Manuel Lopes has a total of 102 co-authors. Here are the top co-authors by number of shared publications:
- José Santos-Victor
LARSyS, Institute for Systems and Robotics (ISR)Lisboa, Instituto Superior Tcnico, Universidade de Lisboa, Lisbon, Portugal
16 shared publications Luis Montesano
Instituto de Investigación en Ingeniería de Sistemas (I3A), Universidad de Zaragoza, Zaragoza, Spain
12 shared publications- Alexandre Bernardino
Institute for Systems and Robotics, Instituto Superior Técnico, University of Lisbon, Lisbon, Lisbon, Portugal
8 shared publications Pierre-Yves Oudeyer
8 shared publicationsFrancisco S. Melo
INESC-ID, Instituto Superior Técnico, University of Lisbon, Lisbon, Portugal
7 shared publicationsMarc Toussaint
University of Stuttgart, Machine Learning and Robotics Lab, Germany
7 shared publications
Latest Publications
- 21 April 2026
- Published by The R Foundation
leaf: Learning Equations for Automated Function Discovery
A unified framework for symbolic regression (SR) and multi-view symbolic regression (MvSR) designed for complex, nonlinear systems, with particular applicability to ecological datasets. The package implements a four-stage workflow: data subset generation, functional form discovery, numerical parameter optimization, and multi-objective evaluation. It provides a high-level formula-style interface that abstracts and extends multiple discovery engines: genetic programming (via PySR), Reinforcement Learning with Monte Carlo Tree Search (via RSRM), and exhaustive generalized linear model search. 'leaf' extends these methods by enabling multi-view discovery, where functional structures are shared across groups while parameters are fitted locally, and by supporting the enforcement of domain-specific constraints, such as sign consistency across groups. The framework automatically handles data normalization, link functions, and back-transformation, ensuring that discovered symbolic equations remain interpretable and valid on the original data scale. Implements methods following ongoing work by the authors (2026, in preparation).
- 21 April 2026
- Published by The R Foundation
leaf: Learning Equations for Automated Function Discovery
A unified framework for symbolic regression (SR) and multi-view symbolic regression (MvSR) designed for complex, nonlinear systems, with particular applicability to ecological datasets. The package implements a four-stage workflow: data subset generation, functional form discovery, numerical parameter optimization, and multi-objective evaluation. It provides a high-level formula-style interface that abstracts and extends multiple discovery engines: genetic programming (via PySR), Reinforcement Learning with Monte Carlo Tree Search (via RSRM), and exhaustive generalized linear model search. 'leaf' extends these methods by enabling multi-view discovery, where functional structures are shared across groups while parameters are fitted locally, and by supporting the enforcement of domain-specific constraints, such as sign consistency across groups. The framework automatically handles data normalization, link functions, and back-transformation, ensuring that discovered symbolic equations remain interpretable and valid on the original data scale. Implements methods following ongoing work by the authors (2026, in preparation).
- Open Access
- 5 November 2025
- Published by Elsevier in Digital Applications in Archaeology and Cultural Heritage
- Vol. 39
A sailing simulator for the revival of the Tagus River nautical heritage
- 5 November 2025
- Published by Elsevier in Digital Applications in Archaeology and Cultural Heritage
- Vol. 39
Open AccessA sailing simulator for the revival of the Tagus River nautical heritage
- 1 August 2025
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 158 (2) , 1060-1071
Deep learning–based frameworks for the detection and classification of soniferous fish
Passive acoustic monitoring (PAM) is emerging as a valuable tool for assessing fish populations in natural habitats. This study compares two deep learning–based frameworks: (1) a multi-label segmentation-based classification system (SegClas) combining convolutional neural networks and long short term memory networks and, (2) an object detection approach (ObjDet) using a You Only Look Once based model to detect, classify, and count sounds produced by soniferous fish in the Tagus Estuary, Portugal. The target species—Lusitanian toadfish (Halobatrachus didactylus), meagre (Argyrosomus regius), and weakfish (Cynoscion regalis)—exhibit overlapping vocalization patterns, posing classification challenges. Results show both methods achieve high accuracy (over 96%) and F1 scores above 87% for species-level sound identification, demonstrating their effectiveness under varied noise conditions. ObjDet generally offers slightly higher classification performance (F1 up to 92%) and can annotate each vocalization for more precise counting. However, it requires bounding-box annotations and higher computational costs (inference time of ca. 1.95 s/h of recording). In contrast, SegClas relies on segment-level labels and provides faster inference (ca. 1.46 s/h). This study also compares both counting strategies, each offering distinct advantages for different ecological and operational needs. Our results highlight the potential of deep learning–based PAM for fish population assessment.
- 1 August 2025
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 158 (2) , 1060-1071
Deep learning–based frameworks for the detection and classification of soniferous fish
Passive acoustic monitoring (PAM) is emerging as a valuable tool for assessing fish populations in natural habitats. This study compares two deep learning–based frameworks: (1) a multi-label segmentation-based classification system (SegClas) combining convolutional neural networks and long short term memory networks and, (2) an object detection approach (ObjDet) using a You Only Look Once based model to detect, classify, and count sounds produced by soniferous fish in the Tagus Estuary, Portugal. The target species—Lusitanian toadfish (Halobatrachus didactylus), meagre (Argyrosomus regius), and weakfish (Cynoscion regalis)—exhibit overlapping vocalization patterns, posing classification challenges. Results show both methods achieve high accuracy (over 96%) and F1 scores above 87% for species-level sound identification, demonstrating their effectiveness under varied noise conditions. ObjDet generally offers slightly higher classification performance (F1 up to 92%) and can annotate each vocalization for more precise counting. However, it requires bounding-box annotations and higher computational costs (inference time of ca. 1.95 s/h of recording). In contrast, SegClas relies on segment-level labels and provides faster inference (ca. 1.46 s/h). This study also compares both counting strategies, each offering distinct advantages for different ecological and operational needs. Our results highlight the potential of deep learning–based PAM for fish population assessment.
- Open Access
A Computational Model of Inclusive Pedagogy: From Understanding to Application
Human education transcends mere knowledge transfer, it relies on co-adaptation dynamics -- the mutual adjustment of teaching and learning strategies between agents. Despite its centrality, computational models of co-adaptive teacher-student interactions (T-SI) remain underdeveloped. We argue that this gap impedes Educational Science in testing and scaling contextual insights across diverse settings, and limits the potential of Machine Learning systems, which struggle to emulate and adaptively support human learning processes. To address this, we present a computational T-SI model that integrates contextual insights on human education into a testable framework. We use the model to evaluate diverse T-SI strategies in a realistic synthetic classroom setting, simulating student groups with unequal access to sensory information. Results show that strategies incorporating co-adaptation principles (e.g., bidirectional agency) outperform unilateral approaches (i.e., where only the teacher or the student is active), improving the learning outcomes for all learning types. Beyond the testing and scaling of context-dependent educational insights, our model enables hypothesis generation in controlled yet adaptable environments. This work bridges non-computational theories of human education with scalable, inclusive AI in Education systems, providing a foundation for equitable technologies that dynamically adapt to learner needs.
Open AccessA Computational Model of Inclusive Pedagogy: From Understanding to Application
Human education transcends mere knowledge transfer, it relies on co-adaptation dynamics -- the mutual adjustment of teaching and learning strategies between agents. Despite its centrality, computational models of co-adaptive teacher-student interactions (T-SI) remain underdeveloped. We argue that this gap impedes Educational Science in testing and scaling contextual insights across diverse settings, and limits the potential of Machine Learning systems, which struggle to emulate and adaptively support human learning processes. To address this, we present a computational T-SI model that integrates contextual insights on human education into a testable framework. We use the model to evaluate diverse T-SI strategies in a realistic synthetic classroom setting, simulating student groups with unequal access to sensory information. Results show that strategies incorporating co-adaptation principles (e.g., bidirectional agency) outperform unilateral approaches (i.e., where only the teacher or the student is active), improving the learning outcomes for all learning types. Beyond the testing and scaling of context-dependent educational insights, our model enables hypothesis generation in controlled yet adaptable environments. This work bridges non-computational theories of human education with scalable, inclusive AI in Education systems, providing a foundation for equitable technologies that dynamically adapt to learner needs.
- Open Access
- 15 November 2024
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment
- Vol. 20 (1) , 179-188
Generating Game Levels by Defining Player Experiences
This paper proposes a novel pipeline for generating game levels that elicit predefined emotional experiences from players. Our approach uses evolutionary algorithms alongside data-driven persona agents, predictive emotional models, a PCG parametric level generator, and a newly defined language for the clear and computable definition of player emotional experiences: ExpREx (Experience Regular Expressions). Using these components, we evolve game levels to match the player experience goals specified using the ExpREx language, aiming to create levels that evoke specific emotional experiences for different subsets of players. The efficacy of our method was validated through a user study involving 101 participants, whose continuous annotations of emotional experience were collected and analyzed to assess the congruence between the actual emotional responses elicited and those targeted by our pipeline. We found that 93.73% of the ExpREx goals targeted were also reported by the user study subjects.
- 15 November 2024
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment
- Vol. 20 (1) , 179-188
Open AccessGenerating Game Levels by Defining Player Experiences
This paper proposes a novel pipeline for generating game levels that elicit predefined emotional experiences from players. Our approach uses evolutionary algorithms alongside data-driven persona agents, predictive emotional models, a PCG parametric level generator, and a newly defined language for the clear and computable definition of player emotional experiences: ExpREx (Experience Regular Expressions). Using these components, we evolve game levels to match the player experience goals specified using the ExpREx language, aiming to create levels that evoke specific emotional experiences for different subsets of players. The efficacy of our method was validated through a user study involving 101 participants, whose continuous annotations of emotional experience were collected and analyzed to assess the congruence between the actual emotional responses elicited and those targeted by our pipeline. We found that 93.73% of the ExpREx goals targeted were also reported by the user study subjects.
- 3 January 2024
- Published by Springer Nature
- p. 3-18
Data Driven Agents for User Experience Testing
- Open Access
- 6 October 2023
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment
- Vol. 19 (1) , 215-223
Persona Agents from Playtraces and Emotion
This paper proposes a novel pipeline for generating agents that simulate player behaviour. By clustering player traces and using evolutionary algorithms to evolve parametric agents to best represent those clusters, our pipeline creates persona agents that represent the behavioural space of players. We here propose clustering playtraces based on behaviour, emotional experience and a mixture of both. We implement the pipeline on a test bed game and using 182 collected player traces with both behavioural and emotional information, we demonstrate that our persona agents can generate diverse player-like behaviour both in the level used to evolve them but also in a previously unseen level. We further find that using emotional information leads to better behavioural coverage on both levels. Although on its early stages, our approach offers a new perspective on how game developers and testers can gather insights on player behaviour without having to rely on extensive user testing.
- 6 October 2023
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment
- Vol. 19 (1) , 215-223
Open AccessPersona Agents from Playtraces and Emotion
This paper proposes a novel pipeline for generating agents that simulate player behaviour. By clustering player traces and using evolutionary algorithms to evolve parametric agents to best represent those clusters, our pipeline creates persona agents that represent the behavioural space of players. We here propose clustering playtraces based on behaviour, emotional experience and a mixture of both. We implement the pipeline on a test bed game and using 182 collected player traces with both behavioural and emotional information, we demonstrate that our persona agents can generate diverse player-like behaviour both in the level used to evolve them but also in a previously unseen level. We further find that using emotional information leads to better behavioural coverage on both levels. Although on its early stages, our approach offers a new perspective on how game developers and testers can gather insights on player behaviour without having to rely on extensive user testing.
- Open Access
- 16 September 2021
- Published by Frontiers Media SA in Frontiers in Artificial Intelligence
- Vol. 4
Teaching Multiple Inverse Reinforcement Learners
In this paper, we propose the first machine teaching algorithm for multiple inverse reinforcement learners. As our initial contribution, we formalize the problem of optimally teaching a sequential task to a heterogeneous class of learners. We then contribute a theoretical analysis of such problem, identifying conditions under which it is possible to conduct such teaching using the same demonstration for all learners. Our analysis shows that, contrary to other teaching problems, teaching a sequential task to a heterogeneous class of learners with a single demonstration may not be possible, as the differences between individual agents increase. We then contribute two algorithms that address the main difficulties identified by our theoretical analysis. The first algorithm, which we dub SplitTeach, starts by teaching the class as a whole until all students have learned all that they can learn as a group; it then teaches each student individually, ensuring that all students are able to perfectly acquire the target task. The second approach, which we dub JointTeach, selects a single demonstration to be provided to the whole class so that all students learn the target task as well as a single demonstration allows. While SplitTeachensures optimal teaching at the cost of a bigger teaching effort, JointTeachensures minimal effort, although the learners are not guaranteed to perfectly recover the target task. We conclude by illustrating our methods in several simulation domains. The simulation results agree with our theoretical findings, showcasing that indeed class teaching is not possible in the presence of heterogeneous students. At the same time, they also illustrate the main properties of our proposed algorithms: in all domains, SplitTeachguarantees perfect teaching and, in terms of teaching effort, is always at least as good as individualized teaching (often better); on the other hand, JointTeachattains minimal teaching effort in all domains, even if sometimes it compromises the teaching performance.
- 16 September 2021
- Published by Frontiers Media SA in Frontiers in Artificial Intelligence
- Vol. 4
Open AccessTeaching Multiple Inverse Reinforcement Learners
In this paper, we propose the first machine teaching algorithm for multiple inverse reinforcement learners. As our initial contribution, we formalize the problem of optimally teaching a sequential task to a heterogeneous class of learners. We then contribute a theoretical analysis of such problem, identifying conditions under which it is possible to conduct such teaching using the same demonstration for all learners. Our analysis shows that, contrary to other teaching problems, teaching a sequential task to a heterogeneous class of learners with a single demonstration may not be possible, as the differences between individual agents increase. We then contribute two algorithms that address the main difficulties identified by our theoretical analysis. The first algorithm, which we dub SplitTeach, starts by teaching the class as a whole until all students have learned all that they can learn as a group; it then teaches each student individually, ensuring that all students are able to perfectly acquire the target task. The second approach, which we dub JointTeach, selects a single demonstration to be provided to the whole class so that all students learn the target task as well as a single demonstration allows. While SplitTeachensures optimal teaching at the cost of a bigger teaching effort, JointTeachensures minimal effort, although the learners are not guaranteed to perfectly recover the target task. We conclude by illustrating our methods in several simulation domains. The simulation results agree with our theoretical findings, showcasing that indeed class teaching is not possible in the presence of heterogeneous students. At the same time, they also illustrate the main properties of our proposed algorithms: in all domains, SplitTeachguarantees perfect teaching and, in terms of teaching effort, is always at least as good as individualized teaching (often better); on the other hand, JointTeachattains minimal teaching effort in all domains, even if sometimes it compromises the teaching performance.
- Open Access
Norms for Beneficial A.I.: A Computational Analysis of the Societal Value Alignment Problem
The rise of artificial intelligence (A.I.) based systems is already offering substantial benefits to the society as a whole. However, these systems may also enclose potential conflicts and unintended consequences. Notably, people will tend to adopt an A.I. system if it confers them an advantage, at which point non-adopters might push for a strong regulation if that advantage for adopters is at a cost for them. Here we propose an agent-based game-theoretical model for these conflicts, where agents may decide to resort to A.I. to use and acquire additional information on the payoffs of a stochastic game, striving to bring insights from simulation to what has been, hitherto, a mostly philosophical discussion. We frame our results under the current discussion on ethical A.I. and the conflict between individual and societal gains: the societal value alignment problem. We test the arising equilibria in the adoption of A.I. technology under different norms followed by artificial agents, their ensuing benefits, and the emergent levels of wealth inequality. We show that without any regulation, purely selfish A.I. systems will have the strongest advantage, even when a utilitarian A.I. provides significant benefits for the individual and the society. Nevertheless, we show that it is possible to develop A.I. systems following human conscious policies that, when introduced in society, lead to an equilibrium where the gains for the adopters are not at a cost for non-adopters, thus increasing the overall wealth of the population and lowering inequality. However, as shown, a self-organised adoption of such policies would require external regulation.
Open AccessNorms for Beneficial A.I.: A Computational Analysis of the Societal Value Alignment Problem
The rise of artificial intelligence (A.I.) based systems is already offering substantial benefits to the society as a whole. However, these systems may also enclose potential conflicts and unintended consequences. Notably, people will tend to adopt an A.I. system if it confers them an advantage, at which point non-adopters might push for a strong regulation if that advantage for adopters is at a cost for them. Here we propose an agent-based game-theoretical model for these conflicts, where agents may decide to resort to A.I. to use and acquire additional information on the payoffs of a stochastic game, striving to bring insights from simulation to what has been, hitherto, a mostly philosophical discussion. We frame our results under the current discussion on ethical A.I. and the conflict between individual and societal gains: the societal value alignment problem. We test the arising equilibria in the adoption of A.I. technology under different norms followed by artificial agents, their ensuing benefits, and the emergent levels of wealth inequality. We show that without any regulation, purely selfish A.I. systems will have the strongest advantage, even when a utilitarian A.I. provides significant benefits for the individual and the society. Nevertheless, we show that it is possible to develop A.I. systems following human conscious policies that, when introduced in society, lead to an equilibrium where the gains for the adopters are not at a cost for non-adopters, thus increasing the overall wealth of the population and lowering inequality. However, as shown, a self-organised adoption of such policies would require external regulation.
- Open Access
- 24 March 2020
- Published by SAGE Publications in Journal of Psychopharmacology
- Vol. 34 (8) , 891-900
Oxytocin and vasopressin modulation of prisoner’s dilemma strategies
Background: The neuropeptides oxytocin and vasopressin have been repeatedly implicated in social decision making by enhancing social salience and, generally, cooperation. The iterated and sequential version of the prisoner’s dilemma (PD) game is a social dilemma paradigm eliciting strategies of cooperation versus competition. Aims: We aimed to characterise the role of PD players’ sex, game partner type (computer vs. human) and oxytocin or vasopressin inhalation on the player’s strategy preference. Methods: Participants (153 men; 151 women) were randomised to intranasal 24 IU oxytocin, 20 IU vasopressin or placebo, double-blind, and played the PD. We examined main and interactive effects of sex, drug and partner type on strategy preference. Results: We found a pervasive preference for a tit-for-tat strategy (i.e. general sensitivity to the partner’s choices) over unconditional cooperation, particularly when against a human rather than a computer partner. Oxytocin doubled this sensitivity in women (i.e. the preference for tit-for-tat over unconditional cooperation strategies) when playing against computers, which suggests a tendency to anthropomorphise them, and doubled women’s unconditional cooperation preference when playing against humans. Vasopressin doubled sensitivity to the partner’s previous choices (i.e. for tit-for-tat over unconditional cooperation) across sexes and partner types. Conclusions: These findings suggest that women may be more sensitive to oxytocin’s social effects of anthropomorphism of non-humans and of unconditional cooperation with humans, which may be consistent with evolutionary pressures for maternal care, and that vasopressin, irrespective of sex and partner type, may be generally sensitising humans to others’ behaviour.
- 24 March 2020
- Published by SAGE Publications in Journal of Psychopharmacology
- Vol. 34 (8) , 891-900
Open AccessOxytocin and vasopressin modulation of prisoner’s dilemma strategies
Background: The neuropeptides oxytocin and vasopressin have been repeatedly implicated in social decision making by enhancing social salience and, generally, cooperation. The iterated and sequential version of the prisoner’s dilemma (PD) game is a social dilemma paradigm eliciting strategies of cooperation versus competition. Aims: We aimed to characterise the role of PD players’ sex, game partner type (computer vs. human) and oxytocin or vasopressin inhalation on the player’s strategy preference. Methods: Participants (153 men; 151 women) were randomised to intranasal 24 IU oxytocin, 20 IU vasopressin or placebo, double-blind, and played the PD. We examined main and interactive effects of sex, drug and partner type on strategy preference. Results: We found a pervasive preference for a tit-for-tat strategy (i.e. general sensitivity to the partner’s choices) over unconditional cooperation, particularly when against a human rather than a computer partner. Oxytocin doubled this sensitivity in women (i.e. the preference for tit-for-tat over unconditional cooperation strategies) when playing against computers, which suggests a tendency to anthropomorphise them, and doubled women’s unconditional cooperation preference when playing against humans. Vasopressin doubled sensitivity to the partner’s previous choices (i.e. for tit-for-tat over unconditional cooperation) across sexes and partner types. Conclusions: These findings suggest that women may be more sensitive to oxytocin’s social effects of anthropomorphism of non-humans and of unconditional cooperation with humans, which may be consistent with evolutionary pressures for maternal care, and that vasopressin, irrespective of sex and partner type, may be generally sensitising humans to others’ behaviour.