pith. sign in

arxiv: 2507.13305 · v2 · submitted 2025-07-17 · 💻 cs.LG

Boosting Team Modeling through Tempo-Relational Representation Learning

Vincenzo Marco De Luca , Giovanna Varni , Andrea Passerini This is my paper

Pith reviewed 2026-05-19 03:39 UTC · model grok-4.3

classification 💻 cs.LG
keywords tempo-relational learningteam modelingtemporal graphsmulti-task learningteam performance predictionexplainable AIsocial dynamics
0
0 comments X

The pith

A tempo-relational neural architecture that jointly models team member interactions and dynamics evolution via temporal graphs outperforms temporal-only and relational-only approaches.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces a tempo-relational neural architecture to jointly capture interactions among team members and how team dynamics evolve over time using temporal graphs. This design incorporates social science insights on temporal interactions that most prior models overlook. It also includes a multi-task version that learns shared embeddings to predict several team constructs at once, cutting down on computation time. Tests on two leading team datasets confirm better accuracy than single-aspect models and show the multi-task version maintains performance while being faster. Explainability methods are added to offer clear recommendations for improving teams.

Core claim

The authors claim that their tempo-relational architecture, by jointly modeling member interactions and temporal evolution of dynamics through temporal graphs, achieves superior results in team performance prediction compared to approaches that handle only temporal or only relational aspects. The multi-task extension learns shared social embeddings for simultaneous prediction of constructs like emergent leadership and teamwork components, reducing training and inference time with no drop in accuracy. Integration of explainability provides interpretable insights and actionable recommendations for team enhancement in human-centered AI applications.

What carries the argument

Tempo-relational neural architecture using temporal graphs to jointly model interactions between team members and the evolution of team dynamics.

Load-bearing premise

The two state-of-the-art team datasets used in the experiments represent a broad enough range of real-world collaborative settings to support the observed outperformance and generalizability.

What would settle it

Testing the architecture on a new independent team dataset from a different context such as online gaming teams or corporate project groups and finding no improvement over simpler temporal or relational baselines would falsify the performance claims.

Figures

Figures reproduced from arXiv: 2507.13305 by Andrea Passerini, Giovanna Varni, Vincenzo Marco De Luca.

Figure 1
Figure 1. Figure 1: The BFT framework. It is composed of five key components reported in the orange circles (i.e., Adaptability, Back-Up Behavior, Mutual Performance Monitoring, Team Leadership, and Team Orien￾tation) and three additional components reported in the blue squares (i.e., Closed-loop Communication, Mutual Trust, and Shared Mental Models). All of these components contribute to the Team Effectiveness represented in… view at source ↗
Figure 2
Figure 2. Figure 2: A working, healthcare-oriented, HMT scenario for the MT-TRENN architecture. Here we have two main actors: the user and the agent. The user does not need to communicate with the environment to get team information, as the agent can automatically compute several team constructs (i.e., TW, EL, LS). Once the user has assessed the information through an interface, he/she may ask for additional explanation regar… view at source ↗
Figure 3
Figure 3. Figure 3: The possible Neural Network-based architectures for modeling teams. In any subfigure, two main blocks are represented: the rectangle referring to the encoder, and the rounded rectangle referring to the decoder. Any encoder contains the input data X t+i j that it receives as input and its corresponding hidden representation h t+i j . On the other hand, any decoder receives the hidden representation h t+i j … view at source ↗
Figure 4
Figure 4. Figure 4: Team members continuously interact, exchanging verbal and non-verbal messages (a). Our architecture processes such messages, encoding them into a discrete representation through temporal graphs and feature extraction strategies (b). time h3 t h2 t h1 t h3 t+1 h2 t+1 h1 t+1 h3 t+K h2 t+K h1 t+K [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The output of the Relation Learning step in STT TGNN. At each timestep (from t to t + K), the model produces a tensor rep￾resenting the hidden state of each team member row-wise (three in this example). The columns f1 and f2 refer respectively to the first and the second features generated by the Relational Learning module. These timestep-wise representations are then integrated by the Tem￾poral Learning m… view at source ↗
Figure 6
Figure 6. Figure 6: The main steps involved in MPNN to map topological structure into a single tensor. (a) Given a graph composed of three nodes (x i 1 ,x i 2 ,x i 3 ) where each node has three features and two bidirectional edges. This topology encloses both individual node information (i.e., features) and topological information (i.e., edges). (b) Detailed view of the generation of the hidden representation (i.e., h i 1 ) f… view at source ↗
Figure 7
Figure 7. Figure 7: Given a single team member Xi and its representation over time from time-step t until time-step t + K, the input sequence {X t i , X t+1 i ,...,X t+K i } is represented as the blue square taht are the inputs to generate the query tensor Q and the key tensor K (the orange ones). Next, Q and K are combined to generate the Attention tensor A contain￾ing the attention scores. Then, A is combined with the value… view at source ↗
Figure 8
Figure 8. Figure 8: Pairwise Ranking Loss enforces embeddings of the leaders to be closer in the embedding space, while the embedding of non-leaders to be farther from the ones of the leaders often spanning multiple behavioral and social modalities. While data augmentation techniques are commonly used to mitigate the limited availability of data, they alone are not sufficient to ensure effective generalization, especially whe… view at source ↗
Figure 9
Figure 9. Figure 9: Two principal approaches for modeling multiple tasks in Neural Networks: on the left (a):Traditional learning approach: given an input sample (blue square) composed of three features, two parallel encoders composed of two hidden layers having hidden size four and three (orange square), learns different parameters in order to fed it as input to two different linear layers mapping these hidden representation… view at source ↗
Figure 10
Figure 10. Figure 10: The figure refers to the performance of the models under PAVIS. In the first row, they are respectively reported, MSE of TW, MSE of LS, and EL@ACC 1. In the second row on the other hand, they are reported trainable parameters required to achieve performances from the first row, the training time in ms required for a single training epoch, and the inference time required for a single team. RF FFNN LSTM MHA… view at source ↗
Figure 11
Figure 11. Figure 11: The figure refers to the performance of the models under ELEA. In the first row, they are respectively reported, MSE of TW, MSE of LS, and EL@ACC 1. In the second row, on the other hand, they are reported as trainable parameters required to achieve performances from the first row, the training time in ms required for a single training epoch, and the inference time required for a single team. T = (G,Y), a … view at source ↗
Figure 12
Figure 12. Figure 12: How Team Interaction may affect teamwork factual explanations to interpret the relationship between the behavior of individual team members and TW. While this is not an exhaustive analysis, it showcases the potential of MT-TRENN for fine-grained temporal and relational inter￾pretation, while also suggesting possible directions for addi￾tional xAI-based applications [30]. Let us define the expla￾nation sco… view at source ↗
Figure 13
Figure 13. Figure 13: Counterfactual example: In the top panel of the figure, a real world scenario is reported. In the bottom panel of the figure, the suggestion generated as a counterfactual explanation by CoDy is reported. nario. At timestep t, two team members (i.e., the blonde girl and the boy with the hat) discuss which object is more im￾portant to bring with them to solve the task. At timestep t+1, another team member (… view at source ↗
read the original abstract

Team modeling remains a fundamental challenge at the intersection of Artificial Intelligence and Social Sciences. Although a variety of computational models have been proposed in the last two decades, most fail to integrate Social Sciences insights, such as the critical role of temporal interactions in shaping team dynamics, and do not meet key practical requirements for real-world applications, including the ability to provide real-time, actionable recommendations to enhance team performance. To address these limitations, in this paper, we propose a novel tempo-relational neural architecture that jointly models interactions between team members and the evolution of team dynamics through temporal graphs. We additionally propose a multi-task extension of the architecture that learns shared social embeddings for team members enabling the simultaneous prediction of multiple team constructs (e.g., Emergent Leadership, Leadership Style, and Teamwork components). Experiments on two state-of-the-art team datasets show that our tempo-relational architecture out performs temporal-only and relational-only approaches for team performance prediction, and that its multi-task extension substantially reduces training and inference time without loss of predictive performance. Finally, the integration of explainability techniques within the proposed architectures provides interpretable insights and actionable recommendations to support team improvement. These strengths make our approach particularly well-suited for human-centered artificial intelligence applications, such as intelligent decision-support systems in high-stakes collaborative environments.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 3 minor

Summary. The paper proposes a tempo-relational neural architecture that jointly models team member interactions and the temporal evolution of team dynamics via temporal graphs. A multi-task extension learns shared social embeddings to predict multiple constructs (e.g., Emergent Leadership, Leadership Style) simultaneously. Experiments on two state-of-the-art team datasets demonstrate that the architecture outperforms temporal-only and relational-only baselines for team performance prediction, while the multi-task variant reduces training and inference time with no loss in accuracy. Explainability techniques are integrated to yield interpretable, actionable recommendations for team improvement.

Significance. If the empirical results hold, the work offers a meaningful advance in team modeling at the AI-social sciences intersection by explicitly incorporating temporal interactions, a factor highlighted in social science literature but often omitted in prior computational models. The multi-task efficiency gains and focus on real-time actionable outputs are practical strengths for human-centered applications. Credit is due for supplying dataset descriptions, architecture diagrams, and quantitative experimental comparisons that ground the central claims.

major comments (1)
  1. [Experiments] Experiments section: the outperformance claims rest on comparisons to temporal-only and relational-only baselines, but the manuscript does not specify whether these baselines were given equivalent model capacity, hyperparameter search budgets, or training regimes; without this, the superiority could be attributable to implementation differences rather than the tempo-relational design.
minor comments (3)
  1. [Abstract] Abstract: 'out performs' appears as two words and should be corrected to 'outperforms' for standard English usage.
  2. [§3] Architecture diagrams and §3: node and edge feature definitions for the temporal graphs are described at a high level; explicit equations or pseudocode for how temporal snapshots are constructed and aggregated would improve reproducibility.
  3. [Experiments] Results tables: inclusion of standard deviations or confidence intervals across multiple runs (or at least across data splits) would strengthen the quantitative support for the efficiency and accuracy claims.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback and positive recommendation for minor revision. We address the major comment below and will incorporate the necessary clarifications in the revised manuscript.

read point-by-point responses

  1. Referee: [Experiments] Experiments section: the outperformance claims rest on comparisons to temporal-only and relational-only baselines, but the manuscript does not specify whether these baselines were given equivalent model capacity, hyperparameter search budgets, or training regimes; without this, the superiority could be attributable to implementation differences rather than the tempo-relational design.

    Authors: We appreciate the referee's emphasis on experimental fairness. The baselines were implemented following the architectures and hyperparameter configurations reported in their original publications, with training regimes aligned to those used for our model (including the same optimizer, learning rate schedule, and early stopping criteria). To make this explicit and eliminate any ambiguity, we will add a dedicated subsection to the Experiments section that details model capacities (e.g., parameter counts), the hyperparameter search procedure applied uniformly across all methods, and the training protocols. This revision will include a comparison table confirming that the observed gains stem from the tempo-relational integration rather than unequal resources. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper introduces a tempo-relational neural architecture for modeling team dynamics and evaluates its performance through standard supervised training and comparison against temporal-only and relational-only baselines on two external datasets. No equations, derivations, or first-principles results are presented that reduce claimed predictions to inputs by construction; the architecture is defined independently, trained on held-out data, and assessed via empirical metrics without self-definitional loops, fitted-input predictions, or load-bearing self-citations. The multi-task extension and explainability components are likewise presented as architectural choices validated externally rather than tautological renamings or ansatzes smuggled via prior work.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on standard neural network training assumptions plus the domain premise that temporal interactions are critical for team dynamics; no free parameters or new entities are explicitly introduced in the abstract.

axioms (1)
  • domain assumption Temporal interactions play a critical role in shaping team dynamics
    Stated in the abstract as a key social science insight that prior models fail to integrate.
invented entities (1)
  • tempo-relational neural architecture no independent evidence
    purpose: Jointly model team member interactions and temporal evolution of dynamics via temporal graphs
    New proposed model whose independent evidence is limited to the described experiments.

pith-pipeline@v0.9.0 · 5760 in / 1290 out tokens · 27584 ms · 2026-05-19T03:39:58.952993+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Actionable Real-Time Modeling of Surgical Team Dynamics via Time-Expanded Interaction Graphs

    cs.AI 2026-05 unverdicted novelty 6.0

    Time-expanded interaction graphs with graph neural networks enable real-time prediction of surgical procedure duration deviations from team communication patterns and support counterfactual identification of beneficia...

Reference graph

Works this paper leans on

146 extracted references · 146 canonical work pages · cited by 1 Pith paper · 4 internal anchors

  1. [1]

    Nida Itrat Abbasi, Micol Spitale, Joanna Anderson, Tamsin Ford, Peter B Jones, and Hatice Gunes. 2022. Can robots help in the evaluation of mental wellbeing in children? an empirical study. In2022 31st IEEE in- ternational conference on robot and human interactive communication (RO-MAN). IEEE, 1459–1466

    work page 2022

  2. [2]

    Allen, Colin Fisher, Mohamed Chetouani, Ming Ming Chiu, Hatice Gunes, Marc Mehu, and Hay- ley Hung

    Joseph A. Allen, Colin Fisher, Mohamed Chetouani, Ming Ming Chiu, Hatice Gunes, Marc Mehu, and Hay- ley Hung. 2017. Comparing Social Science and Com- puter Science Workflow Processes for Studying Group Interactions.Small Group Research48, 5 (2017), 568– 590

    work page 2017

  3. [3]

    Multimodal Prediction of Tearing Instabilities in a Tokamak

    Ahmed Amer, Chirag Bhuvaneshwara, Gowtham K. Addluri, Mohammed M. Shaik, Vedant Bonde, and Philipp Müller. 2023. Backchannel Detection and Agreement Estimation from Video with Transformer Networks. In2023 International Joint Conference on Neural Networks (IJCNN). 1–8.https://doi.org/ 10.1109/IJCNN54540.2023.10191640

    work page doi:10.1109/ijcnn54540.2023.10191640 2023

  4. [4]

    2000.Small groups as complex systems: For- mation, coordination, development, and adaptation

    Holly Arrow, Joseph E McGrath, and Jennifer L Berdahl. 2000.Small groups as complex systems: For- mation, coordination, development, and adaptation. Sage Publications

    work page 2000

  5. [5]

    Umut Avci and Oya Aran. 2016. Predicting the perfor- mance in decision-making tasks: From individual cues to group interaction.IEEE Transactions on Multime- dia18, 4 (2016), 643–658. 21

    work page 2016

  6. [6]

    Chongyang Bai, Maksim Bolonkin, Viney Regunath, and VS Subrahmanian. 2023. DIPS: a dyadic impres- sion prediction system for group interaction videos. ACM Transactions on Multimedia Computing, Com- munications and Applications19, 1s (2023), 1–24

    work page 2023

  7. [7]

    Robert Freed Bales. 1980. SYMLOG: Case study kit with instructions for a group self study.(No Title) (1980)

    work page 1980

  8. [8]

    Alex Bavelas. 1950. Communication patterns in task- oriented groups.The journal of the acoustical society of America22, 6 (1950), 725–730

    work page 1950

  9. [9]

    Cigdem Beyan, Francesca Capozzi, Cristina Becchio, and Vittorio Murino. 2017. Multi-task learning of so- cial psychology assessments and nonverbal features for automatic leadership identification. InProceedings of the 19th ACM International Conference on Multi- modal Interaction. 451–455

    work page 2017

  10. [10]

    Cigdem Beyan, Nicolò Carissimi, Francesca Capozzi, Sebastiano Vascon, Matteo Bustreo, Antonio Pierro, Cristina Becchio, and Vittorio Murino. 2016. Detect- ing emergent leader in a meeting environment using nonverbal visual features only. InProceedings of the 18th ACM international conference on multimodal in- teraction. 317–324

    work page 2016

  11. [11]

    Cigdem Beyan, Vasiliki-Maria Katsageorgiou, and Vittorio Murino. 2019. A Sequential Data Analy- sis Approach to Detect Emergent Leaders in Small Groups.IEEE Transactions on Multimedia21, 8 (2019), 2107–2116.https://doi.org/10.1109/ TMM.2019.2895505

    work page arXiv 2019

  12. [12]

    Indrani Bhattacharya, Michael Foley, Christine Ku, Ni Zhang, Tongtao Zhang, Cameron Mine, Manling Li, Heng Ji, Christoph Riedl, Brooke Foucault Welles, et al. 2019. The unobtrusive group interaction (UGI) corpus. InProceedings of the 10th ACM Multimedia Systems Conference. 249–254

    work page 2019

  13. [13]

    McKenzie Braley and Gabriel Murray. 2018. The group affect and performance (GAP) corpus. (2018), 1–9

    work page 2018

  14. [14]

    Aaron Brown and David A Patterson. 2001. To err is human. InProceedings of the First Workshop on evaluating and architecting system dependability (EASY’01)

    work page 2001

  15. [15]

    C Shawn Burke, Kevin C Stagl, Cameron Klein, Ger- ald F Goodwin, Eduardo Salas, and Stanley M Halpin

    work page

  16. [16]

    What type of leadership behaviors are functional in teams? A meta-analysis.The leadership quarterly 17, 3 (2006), 288–307

    work page 2006

  17. [17]

    Janis A Cannon-Bowers and Eduardo Salas. 1997. Teamwork competencies: The interaction of team member knowledge, skills, and attitudes.Workforce readiness: Competencies and assessment(1997), 151– 174

    work page 1997

  18. [18]

    K Catchpole, A Mishra, A Handa, and P McCulloch

    work page

  19. [19]

    Teamwork and error in the operating room: analysis of skills and roles.Annals of surgery247, 4 (2008), 699–706

    work page 2008

  20. [20]

    Albert Chai, Joshua P Le, Andrew S Lee, and Stan- ley M Lo. 2019. Applying graph theory to exam- ine the dynamics of student discussions in small-group learning.CBE—Life Sciences Education18, 2 (2019), ar29

    work page 2019

  21. [21]

    Guillaume Chanel and Christian Mühl. 2015. Con- necting brains and bodies: applying physiological computing to support social interaction.Interacting with Computers27, 5 (2015), 534–550

    work page 2015

  22. [22]

    Woan-Shiuan Chien, Huang-Cheng Chou, and Chi- Chun Lee. 2021. Belongingness and satisfaction recognition from physiological synchrony with a group-modulated attentive BLSTM under small-group conversation. InCompanion Publication of the 2021 International Conference on Multimodal Interaction. 220–229

    work page 2021

  23. [23]

    Woan-Shiuan Chien, Huang-Cheng Chou, and Chi- Chun Lee. 2021. Self-assessed emotion classification from acoustic and physiological features within small- group conversation. InCompanion Publication of the 2021 International Conference on Multimodal Inter- action. 230–239

    work page 2021

  24. [24]

    Marina Chugunova and Daniela Sele. 2022. We and it: an interdisciplinary review of the experimental evi- dence on how humans interact with machines.Journal of Behavioral and Experimental Economics99 (2022), 101897

    work page 2022

  25. [25]

    Andrea Cini, Ivan Marisca, Daniele Zambon, and Ce- sare Alippi. 2023. Graph deep learning for time series forecasting.arXiv preprint arXiv:2310.15978(2023)

    work page arXiv 2023

  26. [26]

    Nancy J Cooke, Jamie C Gorman, Carolyn W Myers, and Jacqueline L Duran. 2013. Interactive team cog- nition.Cognitive science37, 2 (2013), 255–285

    work page 2013

  27. [27]

    Nicola Corbellini, Eleonora Ceccaldi, Giovanna Varni, and Gualtiero V olpe. 2022. An exploratory study on group potency classification from non-verbal so- cial behaviours. InInternational Conference on Pat- tern Recognition. Springer, 240–255

    work page 2022

  28. [28]

    Stephen H Courtright, Gary R Thurgood, Greg L Stew- art, and Abigail J Pierotti. 2015. Structural interdepen- dence in teams: An integrative framework and meta- analysis.Journal of Applied Psychology100, 6 (2015), 1825

    work page 2015

  29. [29]

    Marco Cristani, Ramachandra Raghavendra, Alessio Del Bue, and Vittorio Murino. 2013. Human behavior analysis in video surveillance: A social signal process- ing perspective.Neurocomputing100 (2013), 86–97

    work page 2013

  30. [30]

    Fabio Cuzzolin, Alice Morelli, Bogdan Cirstea, and Barbara J Sahakian. 2020. Knowing me, knowing you: 22 theory of mind in AI.Psychological medicine50, 7 (2020), 1057–1061

    work page 2020

  31. [31]

    Abhinav Dahiya, Alexander M Aroyo, Kerstin Daut- enhahn, and Stephen L Smith. 2023. A sur- vey of multi-agent human–robot interaction sys- tems.Robotics and Autonomous Systems161 (2023), 104335

    work page 2023

  32. [32]

    Vincenzo Marco De Luca, Antonio Longa, Pietro Lio, and Andrea Passerini. [n. d.]. xAI-Drop: Don’t use what you cannot explain. InThe Third Learning on Graphs Conference

    work page

  33. [33]

    Ewart J de Visser, Yigit Topoglu, Shawn Joshi, Frank Krueger, Elizabeth Phillips, Jonathan Gratch, Chad C Tossell, and Hasan Ayaz. 2022. Designing man’s new best friend: Enhancing human-robot dog interaction through dog-like framing and appearance.Sensors22, 3 (2022), 1287

    work page 2022

  34. [34]

    Fabrice Delice, Moira Rousseau, and Jennifer Feitosa

    work page

  35. [35]

    Advancing teams research: What, when, and how to measure team dynamics over time.Frontiers in psychology10 (2019), 1324

    work page 2019

  36. [36]

    D Scott DeRue and Susan J Ashford. 2010. Who will lead and who will follow? A social process of leader- ship identity construction in organizations.Academy of management review35, 4 (2010), 627–647

    work page 2010

  37. [37]

    Jessica E Dinh, Robert G Lord, William L Gardner, Jeremy D Meuser, Robert C Liden, and Jinyu Hu

    work page

  38. [38]

    Leadership theory and research in the new mil- lennium: Current theoretical trends and changing per- spectives.The leadership quarterly25, 1 (2014), 36– 62

    work page 2014

  39. [39]

    Kevin J Dooley. 1997. A complex adaptive systems model of organization change.Nonlinear dynamics, psychology, and life sciences1 (1997), 69–97

    work page 1997

  40. [40]

    Dario Dotti, Mirela Popa, and Stylianos Asteriadis

    work page

  41. [41]

    Interact

    Being the Center of Attention: A Person- Context CNN Framework for Personality Recogni- tion.ACM Trans. Interact. Intell. Syst.10, 3, Article 19 (Nov. 2020), 20 pages.https://doi.org/10. 1145/3338245

    work page 2020

  42. [42]

    Jean L Dyer. 1984. Team research and team training: A state-of-the-art review.Human factors review26 (1984), 285–323

    work page 1984

  43. [43]

    Florian Eyben, Klaus R Scherer, Björn W Schuller, Johan Sundberg, Elisabeth André, Carlos Busso, Laurence Y Devillers, Julien Epps, Petri Laukka, Shrikanth S Narayanan, et al. 2015. The Geneva min- imalistic acoustic parameter set (GeMAPS) for voice research and affective computing.IEEE transactions on affective computing7, 2 (2015), 190–202

    work page 2015

  44. [44]

    Florian Eyben, Martin Wöllmer, and Björn Schuller

    work page

  45. [45]

    InProceedings of the 18th ACM international conference on Multimedia

    Opensmile: the munich versatile and fast open- source audio feature extractor. InProceedings of the 18th ACM international conference on Multimedia. 1459–1462

    work page

  46. [46]

    David L FORD and M Ismail KIRAN. 2008. Per- ceived leader effectiveness across organizational roles: Exploratory evidence from Central Eurasia.Eurasian Journal of Business and Economics1, 2 (2008), 131– 156

    work page 2008

  47. [47]

    Daniel Gatica-Perez. 2009. Automatic nonverbal anal- ysis of social interaction in small groups: A review. Image and vision computing27, 12 (2009), 1775– 1787

    work page 2009

  48. [48]

    Shreya Ghosh, Abhinav Dhall, Nicu Sebe, and Tom Gedeon. 2020. Automatic prediction of group cohe- siveness in images.IEEE Transactions on Affective Computing13, 3 (2020), 1677–1690

    work page 2020

  49. [49]

    Justin Gilmer, Samuel S Schoenholz, Patrick F Riley, Oriol Vinyals, and George E Dahl. 2017. Neural mes- sage passing for quantum chemistry. InInternational conference on machine learning. PMLR, 1263–1272

    work page 2017

  50. [50]

    2006.Social intelligence: The new sci- ence of human relationships

    D Goleman. 2006.Social intelligence: The new sci- ence of human relationships. Bantam Books

    work page 2006

  51. [51]

    right to explanation

    Bryce Goodman and Seth Flaxman. 2017. European Union regulations on algorithmic decision-making and a “right to explanation”.AI magazine38, 3 (2017), 50–57

    work page 2017

  52. [52]

    Marco Gori, Gabriele Monfardini, and Franco Scarselli. 2005. A new model for learning in graph do- mains. InProceedings. 2005 IEEE international joint conference on neural networks, 2005., V ol. 2. IEEE, 729–734

    work page 2005

  53. [53]

    Jamie C Gorman, Terri A Dunbar, David Grimm, and Christina L Gipson. 2017. Understanding and model- ing teams as dynamical systems.Frontiers in psychol- ogy8 (2017), 1053

    work page 2017

  54. [54]

    Alex Graves and Jürgen Schmidhuber. 2005. Frame- wise phoneme classification with bidirectional LSTM and other neural network architectures.Neural net- works18, 5-6 (2005), 602–610

    work page 2005

  55. [55]

    Hayley Hung and Daniel Gatica-Perez. 2010. Esti- mating Cohesion in Small Groups Using Audio-Visual Nonverbal Behavior.IEEE Transactions on Multime- dia12, 6 (2010), 563–575

    work page 2010

  56. [56]

    Norton Jr, Monique L

    William I. Norton Jr, Monique L. Ueltschy Murfield, and Melissa S. Baucus. 2014. Leader emergence: the development of a theoretical framework.Leadership & Organization Development Journal35, 6 (2014), 513–529

    work page 2014

  57. [57]

    Michele H Jackson. 2024. Modeling of Small Groups in Computational Sciences: A Prospecting Review. Small Group Research(2024), 10464964241279164

    work page 2024

  58. [58]

    Julio CS Jacques Junior, Ya ˘gmur Güçlütürk, Marc Pérez, Umut Güçlü, Carlos Andujar, Xavier Baró, Hugo Jair Escalante, Isabelle Guyon, Marcel AJ 23 Van Gerven, Rob Van Lier, et al. 2019. First impres- sions: A survey on vision-based apparent personality trait analysis.IEEE Transactions on Affective Com- puting13, 1 (2019), 75–95

    work page 2019

  59. [59]

    Eija Kaasinen, Anu-Hanna Anttila, Päivi Heikkilä, Jari Laarni, Hanna Koskinen, and Antti Väätänen. 2022. Smooth and resilient human–machine teamwork as an industry 5.0 design challenge.Sustainability14, 5 (2022), 2773

    work page 2022

  60. [60]

    Tharindu Kaluarachchi, Andrew Reis, and Suranga Nanayakkara. 2021. A review of recent deep learn- ing approaches in human-centered machine learning. Sensors21, 7 (2021), 2514

    work page 2021

  61. [61]

    Joseph Kim and Julie A Shah. 2016. Improv- ing team’s consistency of understanding in meetings. IEEE Transactions on Human-Machine Systems46, 5 (2016), 625–637

    work page 2016

  62. [62]

    Thomas N Kipf and Max Welling. 2016. Semi- supervised classification with graph convolutional net- works.arXiv preprint arXiv:1609.02907(2016)

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  63. [63]

    Florian Klonek, Melissa Twemlow, Maria Tims, and Sharon K Parker. 2024. It’s About Time! Un- derstanding the Dynamic Team Process-Performance Relationship Using Micro-and Macroscale Time Lenses.Group & Organization Management(2024), 10596011241278556

    work page 2024

  64. [64]

    Robert J Koenigs, SE Hare, and AP Hare. 2002. SYM- LOG reliability and validity.San Diego: Symlog Con- sulting Group(2002)

    work page 2002

  65. [65]

    Steve WJ Kozlowski and Georgia T Chao. 2018. Un- packing team process dynamics and emergent phe- nomena: Challenges, conceptual advances, and inno- vative methods.American Psychologist73, 4 (2018), 576

    work page 2018

  66. [66]

    Steve WJ Kozlowski, SM Gully, PP McHugh, E Salas, and JA Cannon-Bowers. 1996. A dynamic theory of leadership and team effectiveness: Developmental and task contingent leader roles.Research in personnel and human resources management14 (1996), 253– 306

    work page 1996

  67. [67]

    Steve WJ Kozlowski and Daniel R Ilgen. 2006. En- hancing the effectiveness of work groups and teams. Psychological science in the public interest7, 3 (2006), 77–124

    work page 2006

  68. [68]

    S. W. J. Kozlowski and D. R. Ilgen. 2006. Enhancing the Effectiveness of Work Groups and Teams.Psy- chological Science in the Public Interest7, 3 (2006), 77–124

    work page 2006

  69. [69]

    Uliyana Kubasova, Gabriel Murray, and McKenzie Braley. 2019. Analyzing verbal and nonverbal fea- tures for predicting group performance.arXiv preprint arXiv:1907.01369(2019)

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  70. [70]

    Colin Lea, Michael D Flynn, Rene Vidal, Austin Re- iter, and Gregory D Hager. 2017. Temporal convolu- tional networks for action segmentation and detection. Inproceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 156–165

    work page 2017

  71. [71]

    RTAJ Leenders, NS Contractor, and LA DeChurch

    work page

  72. [72]

    Organizational Psychology Review, 6 (1), 92–115

    Once upon a time: Understanding team pro- cesses as relational event networks. Organizational Psychology Review, 6 (1), 92–115

    work page

  73. [73]

    Kurt Lewin. 1948. Field theory

    work page 1948

  74. [74]

    Jia Li, Yangchen Yu, Yin Chen, Yu Zhang, Peng Jia, Yunbo Xu, Ziqiang Li, Meng Wang, and Richang Hong. 2024. DAT: Dialogue-Aware Transformer with Modality-Group Fusion for Human Engagement Esti- mation. InProceedings of the 32nd ACM International Conference on Multimedia. 11397–11403

    work page 2024

  75. [75]

    Yun-Shao Lin and Chi-Chun Lee. 2020. Predict- ing performance outcome with a conversational graph convolutional network for small group interactions. In ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 8044–8048

    work page 2020

  76. [76]

    Yun-Shao Lin, Yi-Ching Liu, and Chi-Chun Lee

    work page

  77. [77]

    An Interaction-process-guided Framework for Small-group Performance Prediction.ACM Transac- tions on Multimedia Computing, Communications and Applications19, 2 (2023), 1–25

    work page 2023

  78. [78]

    Antonio Longa, Veronica Lachi, Gabriele Santin, Monica Bianchini, Bruno Lepri, Pietro Lio, Franco Scarselli, and Andrea Passerini. 2023. Graph neu- ral networks for temporal graphs: State of the art, open challenges, and opportunities.arXiv preprint arXiv:2302.01018(2023)

    work page arXiv 2023

  79. [79]

    Robert G Lord, Roseanne J Foti, and Christy L De Vader. 1984. A test of leadership categorization theory: Internal structure, information processing, and leadership perceptions.Organizational behavior and human performance34, 3 (1984), 343–378

    work page 1984

  80. [80]

    Martin A Makary and Michael Daniel. 2016. Medical error—the third leading cause of death in the US.Bmj 353 (2016)

    work page 2016

Showing first 80 references.