pith. sign in

arxiv: 2605.17468 · v1 · pith:EO55UQASnew · submitted 2026-05-17 · 💻 cs.HC · cs.AI

An Interpretable Closed-Loop Intelligent Tutoring System for Multimodal Affective Feedback in Asynchronous Presentation Training

Hung-Yue Suen , Kuo-En Hung This is my paper

Pith reviewed 2026-05-19 22:29 UTC · model grok-4.3

classification 💻 cs.HC cs.AI
keywords intelligent tutoring systemmultimodal feedbackpresentation skillsBARSexplainable AIdeliberate practiceaffective computingasynchronous training
0
0 comments X p. Extension
pith:EO55UQAS Add to your LaTeX paper What is a Pith Number? 
\usepackage{pith}
\pithnumber{EO55UQAS}

Prints a linked pith:EO55UQAS badge after your title and writes the identifier into PDF metadata. Compiles on arXiv with no extra files. Learn more

The pith

A closed-loop multimodal ITS using traceable BARS feedback produces measurable gains in presentation skills for 204 adult learners.

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

The paper builds and tests an intelligent tutoring system that turns multimodal video features into interpretable coaching for on-camera presentations. It anchors scoring in a seven-dimension Behaviorally Anchored Rating Scale and routes outputs through scoring, diagnostics, and conversational feedback layers. In a 30-day pre-post study, users improved on every dimension, with more frequent practice tied to larger gains after baseline controls. A sympathetic reader would care because the work shows one concrete path for scaling deliberate practice in performance skills without constant human oversight.

Core claim

The paper claims that its three-layer interpretable feedback architecture, built on an XGBoost model trained on 10,360 MOOC segments, maps facial, vocal, textual, and oculomotor inputs into rubric-aligned scores and audience-perceived diagnostics that support deliberate practice, resulting in significant pre-post gains across all seven BARS dimensions with Cohen's d ranging from 0.39 to 0.90 and a positive link between practice frequency and posttest scores.

What carries the argument

The three-layer interpretable feedback architecture that links rubric-aligned multimodal scoring, audience-perceived expressive diagnostics, and retrieval-augmented conversational coaching.

If this is right

  • Feedback can be traced directly to specific observable performance cues in the video.
  • The same architecture can support large-scale asynchronous training for presentation competencies.
  • Higher practice frequency is associated with stronger posttest outcomes after demographic and baseline controls.
  • Multimodal analytic outputs can be converted into observable behavioral change through an integrated coaching loop.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The approach could be extended to other performance domains such as interview or sales pitch training by swapping the BARS rubric.
  • Deployment on consumer video platforms might allow real-time or near-real-time coaching loops beyond the 30-day window tested.
  • Future work could test whether the gains persist after the supported practice period ends.

Load-bearing premise

The observed pre-post gains result from the ITS feedback itself rather than from repeated practice or from self-selection of already-motivated learners.

What would settle it

A randomized trial that assigns one group to the full ITS and another to equivalent practice without the feedback system would show whether the reported gains exceed those from practice alone.

read the original abstract

This paper presents an interpretable closed-loop Intelligent Tutoring System (ITS) that supports feedback-guided practice for developing on-camera oral presentation skills at scale. The system operationalizes a seven-dimensional Behaviorally Anchored Rating Scale (BARS) and implements a three-layer interpretable feedback architecture that connects rubric-aligned multimodal scoring, audience-perceived expressive diagnostics, and retrieval-augmented conversational coaching to support deliberate practice. Built on an XGBoost backbone, the ITS maps multimodal inputs (facial, vocal, textual, and oculomotor features) into evidence-based feedback that can be traced back to observable performance cues. Trained on 10,360 Massive Open Online Course (MOOC) video segments, the system achieved rubric-aligned scoring with performance levels comparable to expert ratings (R2 = 0.48-0.61, Spearman's rho = 0.69-0.78, MAE = 0.43-0.57). In a pre-post validation study with 204 adult learners over a 30-day practice window, participants demonstrated significant improvements across all seven BARS dimensions (Cohen's d = 0.39-0.90), with practice frequency showing a strong positive association with posttest performance after controlling for baseline scores and demographics. The results demonstrate how multimodal analytic outputs can be systematically transformed into observable behavioral change through an integrated feedback architecture, advancing explainable and pedagogically grounded ITS design for performance-based competencies.

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

2 major / 2 minor

Summary. The manuscript presents an interpretable closed-loop Intelligent Tutoring System (ITS) for asynchronous on-camera presentation training. It operationalizes a seven-dimensional Behaviorally Anchored Rating Scale (BARS) and deploys a three-layer feedback architecture that links multimodal scoring (facial, vocal, textual, oculomotor features via XGBoost), audience-perceived expressive diagnostics, and retrieval-augmented conversational coaching. The system is trained on 10,360 MOOC video segments and reports rubric-aligned performance (R² = 0.48-0.61, Spearman's rho = 0.69-0.78, MAE = 0.43-0.57). In a pre-post validation study with 204 adult learners over 30 days, participants showed significant gains across all BARS dimensions (Cohen's d = 0.39-0.90), with practice frequency positively associated with posttest scores after covariate adjustment.

Significance. If the pre-post gains can be attributed to the ITS feedback architecture rather than repeated practice or selection effects, the work would advance explainable ITS design for performance competencies by demonstrating a traceable pipeline from multimodal analytics to observable behavioral change. The large training corpus, emphasis on interpretability, and integration of rubric-aligned scoring with coaching are clear strengths that could support scalable deliberate practice tools.

major comments (2)
  1. [Abstract / validation study] Abstract and validation study description: the pre-post design reports significant BARS improvements (d = 0.39-0.90) and a practice-frequency association after controlling for baseline and demographics, yet supplies no control arm, randomization, or yoked practice-only condition. This leaves the central claim that the three-layer closed-loop architecture produces behavioral change vulnerable to alternative explanations such as repeated recording practice or motivated self-selection; a randomized or controlled comparison is required to isolate the system's contribution.
  2. [Abstract] Abstract: the reported scoring metrics (R² = 0.48-0.61, rho = 0.69-0.78) are presented without accompanying details on feature engineering, cross-validation strategy, inter-rater reliability baselines, or ablation against simpler models. These omissions limit evaluation of whether the XGBoost backbone reliably supports the downstream interpretable feedback claims.
minor comments (2)
  1. [Abstract] The seven BARS dimensions are referenced but not enumerated in the abstract; listing them explicitly would improve immediate readability.
  2. [Discussion] The manuscript would benefit from a dedicated limitations subsection that directly addresses the pre-post design constraints and their implications for causal inference.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We appreciate the referee's insightful comments on our manuscript. We address the major concerns point by point below, making revisions to the manuscript where appropriate to strengthen the presentation of our work.

read point-by-point responses

  1. Referee: [Abstract / validation study] Abstract and validation study description: the pre-post design reports significant BARS improvements (d = 0.39-0.90) and a practice-frequency association after controlling for baseline and demographics, yet supplies no control arm, randomization, or yoked practice-only condition. This leaves the central claim that the three-layer closed-loop architecture produces behavioral change vulnerable to alternative explanations such as repeated recording practice or motivated self-selection; a randomized or controlled comparison is required to isolate the system's contribution.

    Authors: We agree that the pre-post design without a control arm or randomization limits the ability to fully isolate the contribution of the three-layer ITS architecture from repeated practice or self-selection effects. The study was designed as an initial naturalistic validation to assess feasibility and observable skill gains in a scalable setting. We have revised the manuscript to expand the limitations and discussion sections, explicitly addressing these alternative explanations and noting the positive association between practice frequency and posttest scores (after covariate adjustment) as consistent with a dose-response pattern aligned with deliberate practice. A dedicated randomized controlled trial would be needed for stronger causal inference and is identified as future work. revision: partial

  2. Referee: [Abstract] Abstract: the reported scoring metrics (R² = 0.48-0.61, rho = 0.69-0.78) are presented without accompanying details on feature engineering, cross-validation strategy, inter-rater reliability baselines, or ablation against simpler models. These omissions limit evaluation of whether the XGBoost backbone reliably supports the downstream interpretable feedback claims.

    Authors: We thank the referee for this observation. We have revised the manuscript to include expanded details in the methods section on the multimodal feature engineering pipeline, the cross-validation strategy used for model training and evaluation, inter-rater reliability baselines for the BARS annotations, and ablation comparisons against simpler baseline models. These additions provide greater transparency on the reliability of the scoring component and its role in enabling the interpretable feedback architecture. revision: yes

standing simulated objections not resolved
  • The request for a randomized or controlled comparison to isolate the ITS contribution from repeated practice or selection effects, which would require new data collection and study design beyond the current work.

Circularity Check

0 steps flagged

No significant circularity in empirical model training or pre-post validation

full rationale

The paper describes training an XGBoost model on 10,360 MOOC segments to produce rubric-aligned scores (R2 = 0.48-0.61, rho = 0.69-0.78) and reports pre-post gains in a 204-learner study (Cohen's d = 0.39-0.90) with a practice-frequency association after covariates. No equations, first-principles derivations, or predictions are presented that reduce by construction to fitted inputs, self-definitions, or self-citation chains. The central claims rest on observable empirical metrics and an observational validation design rather than any load-bearing step that is equivalent to its own inputs.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The claim rests on the untested assumption that the chosen multimodal features and BARS dimensions capture the causal drivers of audience-perceived quality, plus standard ML assumptions about i.i.d. data and stable feature distributions across MOOC videos.

free parameters (1)
  • XGBoost hyperparameters and feature weights
    Model is trained on 10,360 segments; exact regularization, tree depth, and feature selection choices are fitted to data but not reported.
axioms (2)
  • domain assumption The seven-dimensional Behaviorally Anchored Rating Scale validly measures presentation quality as perceived by audiences.
    The entire feedback architecture is built on operationalizing this scale.
  • domain assumption Multimodal features (facial, vocal, textual, oculomotor) are sufficient to predict the BARS scores without major missing variables.
    The mapping from raw video to rubric scores assumes these channels capture the relevant variance.

pith-pipeline@v0.9.0 · 5792 in / 1476 out tokens · 46285 ms · 2026-05-19T22:29:13.207699+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

50 extracted references · 50 canonical work pages

  1. [1]

    Controlled evaluation of a multimodal system to improve oral presentation skills in a real learning setting,

    X. Ochoa and F. Domínguez, “Controlled evaluation of a multimodal system to improve oral presentation skills in a real learning setting,” Br. J. Educ. Technol., vol. 51, no. 5, pp. 1615 –1630, 2020, doi: 10.1111/bjet.12987

    work page doi:10.1111/bjet.12987 2020

  2. [2]

    OpenOPAF: An open source multimodal system for automated feedback for oral presentations,

    X. Ochoa and H. Zhao, “OpenOPAF: An open source multimodal system for automated feedback for oral presentations,” J. Learn. Anal., vol. 11, no. 3, pp. 224–248, 2024, doi: 10.18608/jla.2024.8411

    work page doi:10.18608/jla.2024.8411 2024

  3. [3]

    Evaluation of presentation skills in the context of online learning: A literature review,

    S. Suroto, E. Y. Haenilah, H. Hariri, Pargito, and N. Trenggono, “Evaluation of presentation skills in the context of online learning: A literature review,” Int. J. Inf. Educ. Technol., vol. 13, no. 5, pp. 855 –860, 2023, doi: 10.18178/ijiet.2023.13.5.1879

    work page doi:10.18178/ijiet.2023.13.5.1879 2023

  4. [4]

    Developing a computer -based tutor utilizing generative artificial intelligence (GAI) and retrieval-augmented generation (RAG),

    Y. Lee, “Developing a computer -based tutor utilizing generative artificial intelligence (GAI) and retrieval-augmented generation (RAG),” Educ. Inf. Technol., vol. 30, pp. 7841–7862, 2025, doi: 10.1007/s10639-024-13129- 5

    work page doi:10.1007/s10639-024-13129- 2025

  5. [5]

    Predicting presentation skill of a speaker using automatic speaker and audience measurement,

    C. Thomas and D. Jayagopi, “Predicting presentation skill of a speaker using automatic speaker and audience measurement,” IEEE Trans. Learn. Technol., vol. 15, pp. 350–363, 2022, doi: 10.1109/TLT.2022.3171601

    work page doi:10.1109/tlt.2022.3171601 2022

  6. [6]

    Multimodal transfer learning for oral presentation assessment,

    S. S. Y. Tun, S. Okada, H. -H. Huang, and C. W. Leong, “Multimodal transfer learning for oral presentation assessment,” IEEE Access, vol. 11, pp. 84013–84026, 2023, doi: 10.1109/ACCESS.2023.3295832

    work page doi:10.1109/access.2023.3295832 2023

  7. [7]

    The first steps for adapting an artificial intelligence emotion expression recognition software for emotional management in the educational context,

    J. Herrero, F. Gomez -Donoso, and R. Roig -Vila, “The first steps for adapting an artificial intelligence emotion expression recognition software for emotional management in the educational context,” Br. J. Educ. Technol., vol. 54, pp. 1939–1963, 2023, doi: 10.1111/bjet.13326

    work page doi:10.1111/bjet.13326 1939

  8. [8]

    Learning through AI -clones: Enhancing self-perception and presentation performance,

    Q. Zheng, Z. Chen, and Y. Huang, “Learning through AI -clones: Enhancing self-perception and presentation performance,” Comput. Hum. Behav.: Artif. Humans, vol. 3, p. 100117, 2025, doi: 10.1016/j.chbah.2025.100117

    work page doi:10.1016/j.chbah.2025.100117 2025

  9. [9]

    Effect of video styles on learner engagement in MOOCs,

    R. Deng, “Effect of video styles on learner engagement in MOOCs,” Technol. Pedagog. Educ., vol. 33, no. 1, pp. 1 –21, 2023, doi: 10.1080/1475939X.2023.2246981

    work page doi:10.1080/1475939x.2023.2246981 2023

  10. [10]

    Towards a set of design principles for developing oral presentation competence: A synthesis of research in higher education,

    H. van Ginkel, J. Gulikers, H. Biemans, and M. Mulder, “Towards a set of design principles for developing oral presentation competence: A synthesis of research in higher education,” Educ. Res. Rev., vol. 14, pp. 62–80, 2015, doi: 10.1016/j.edurev.2015.02.002

    work page doi:10.1016/j.edurev.2015.02.002 2015

  11. [11]

    Card sorting with fewer cards and the same mental models? a reexamination of an established practice.International Journal of Human–Computer Interaction, 0 (0):1–25, 2026

    D. J. Neufeld, M. M. Roghanizad, and R. E. White, "The impact of video- mediated communication on social predictions and theory of mind activation," Int. J. Hum. –Comput. Interact. , pp. 1 –14, 2025, doi: 10.1080/10447318.2025.2493374

    work page doi:10.1080/10447318.2025.2493374 2025

  12. [12]

    Evaluating recent advances in affective intelligent tutoring systems: A scoping review of educational impacts and future prospects,

    J. Ferná ndez-Herrero, “Evaluating recent advances in affective intelligent tutoring systems: A scoping review of educational impacts and future prospects,” Educ. Sci., vol. 14, no. 8, p. 839, 2024, doi: 10.3390/educsci14080839

    work page doi:10.3390/educsci14080839 2024

  13. [13]

    XGBoost to enhance learner performance prediction,

    S. Hakkal and A. A. Lahcen, "XGBoost to enhance learner performance prediction," Comput. Educ.: Artif . Intell., vol. 7, p. 100254, 2024. doi: 10.1016/j.caeai.2024.100254

    work page doi:10.1016/j.caeai.2024.100254 2024

  14. [14]

    A review of automated feedback systems for learners: Classification framework, challenges and opportunities,

    G. Deeva, D. Bogdanova, E. Serral, M. Snoeck, and J. De Weerdt, “A review of automated feedback systems for learners: Classification framework, challenges and opportunities,” Comput. Educ., vol. 162, p. 104094, 2021, doi: 10.1016/j.compedu.2020.104094

    work page doi:10.1016/j.compedu.2020.104094 2021

  15. [15]

    Intelligent tutoring systems and learning outcomes: A meta-analysis,

    W. Ma, O. Adesope, J. Nesbit, and Q. Liu, “Intelligent tutoring systems and learning outcomes: A meta-analysis,” J. Educ. Psychol., vol. 106, no. 4, pp. 901–918, 2014, doi: 10.1037/a0037123

    work page doi:10.1037/a0037123 2014

  16. [16]

    Effectiveness of intelligent tutoring systems,

    J. A. Kulik and J. D. Fletcher, “Effectiveness of intelligent tutoring systems,” Rev. Educ. Res., vol. 86, no. 1, pp. 42 –78, 2016, doi: 10.3102/0034654315581420

    work page doi:10.3102/0034654315581420 2016

  17. [17]

    Data - driven artificial intelligence in education: A comprehensive review,

    K. Ahmad, H. Ullah, A. Al -Barakati, M. Al -Shehri, and F. Alam, “Data - driven artificial intelligence in education: A comprehensive review,” IEEE Trans. Learn. Technol., vol. 17, no. 1, pp. 12 –31, 2024, doi: 10.1109/TLT.2023.3323123

    work page doi:10.1109/tlt.2023.3323123 2024

  18. [18]

    Rubric formats for the formative assessment of oral presentation skills acquisition in secondary education,

    R. Nadolski, H. Hummel, E. Rusman, and K. Ackermans, “Rubric formats for the formative assessment of oral presentation skills acquisition in secondary education,” Educ. Technol. Res. Dev., vol. 69, pp. 2663 –2682, 2021, doi: 10.1007/s11423-021-10030-7. Accepted manuscript. Published version available in IEEE Transactions on Learning Technologies. DOI: 10....

    work page doi:10.1007/s11423-021-10030-7 2021

  19. [19]

    Attaining self -regulation: A social cognitive perspective,

    B. J. Zimmerman, “Attaining self -regulation: A social cognitive perspective,” in Handb. Self-Regulation, M. Boekaerts, P. R. Pintrich, and M. Zeidner, Eds. Academic Press, 2000, pp. 13 –39, doi: 10.1016/B978 - 012109890-2/50031-7

    work page doi:10.1016/b978 2000

  20. [20]

    The added benefit of an extra practice session in virtual reality on the development of presentation skills: A randomized control trial,

    J. Boetje and S. Ginkel, “The added benefit of an extra practice session in virtual reality on the development of presentation skills: A randomized control trial,” J. Comput. Assist. Learn., vol. 37, pp. 253 –264, 2020, doi: 10.1111/jcal.12484

    work page doi:10.1111/jcal.12484 2020

  21. [21]

    DBATES: Dataset for discerning benefits of audio, textual, and facial expression features in comp etitive debate speeches,

    T. Sen, G. Naven, L. Gerstner, D. Bagley, R. Baten, W. Rahman, M. Hasan, K. Haut, A. Mamun, S. Samrose, A. Solbu, R. Barnes, G. Mark, F. Metze, and E. Hoque, “DBATES: Dataset for discerning benefits of audio, textual, and facial expression features in comp etitive debate speeches,” IEEE Trans. Affect. Comput., vol. 14, pp. 1028 –1043, 2023, doi: 10.1109/T...

    work page doi:10.1109/taffc.2021.3103442 2023

  22. [22]

    Automatic gaze analysis: A survey of deep learning based approaches,

    S. Ghosh, A. Dhall, M. Hayat, J. Knibbe, and Q. Ji, “Automatic gaze analysis: A survey of deep learning based approaches,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 46, no. 1, pp. 61 –84, Jan. 2024, doi: 10.1109/TPAMI.2023.3321337

    work page doi:10.1109/tpami.2023.3321337 2024

  23. [23]

    Face direction estimation based on MediaPipe landmarks,

    A. Al -Nuimi and G. Mohammed, “Face direction estimation based on MediaPipe landmarks,” in Proc. 7th Int. Conf. Contemp. Inf. Technol. Math. (ICCITM), 2021, pp. 185 –190, doi: 10.1109/ICCITM53167.2021.9677878

    work page doi:10.1109/iccitm53167.2021.9677878 2021

  24. [24]

    Online presentations for instruction: An overview,

    F. Ruth, C. Lipphardt, M. Schickel, E. Ruth -Herbein, and T. Ringeisen, “Online presentations for instruction: An overview,” Front. Educ., vol. 10, p. 1450222, 2025, doi: 10.3389/feduc.2025.145022

    work page doi:10.3389/feduc.2025.145022 2025

  25. [25]

    Chen and C

    T. Chen and C. Guestrin, “XGBoost: A scalable tree boosting system,” Proc. 22nd ACM SIGKDD Int. Conf. Knowledge Discovery and Data Mining, 2016, pp. 785–794, doi: 10.1145/2939672.2939785

    work page doi:10.1145/2939672.2939785 2016

  26. [26]

    A review on gradient boosting decision trees,

    Z. Feng, C. Chen, and Z. Li, “A review on gradient boosting decision trees,” IEEE Trans. Neural Netw. Learn. Syst., early access, pp. 1 –20, 2023, doi: 10.1109/TNNLS.2023.3242933

    work page doi:10.1109/tnnls.2023.3242933 2023

  27. [27]

    Does using virtual reality to enhance students' presentation skills work? The role of feedback and presence,

    R. Di Palma, S. Beausaert, D. Mahr, J. Heller, and T. Hilken, “Does using virtual reality to enhance students' presentation skills work? The role of feedback and presence,” J. Comput. Assist. Learn., vol. 41, no. 5, p. e70097, 2025, doi: 10.1111/jcal.70097

    work page doi:10.1111/jcal.70097 2025

  28. [28]

    Heuristics for supporting cooperative dashboard design,

    V. Setlur, M. Correll, A. Satyanarayan, and M. Tory, “Heuristics for supporting cooperative dashboard design,” IEEE Trans. Vis. Comput. Graph., vol. 30, pp. 370–380, 2023, doi: 10.1109/TVCG.2023.3327158

    work page doi:10.1109/tvcg.2023.3327158 2023

  29. [29]

    Examining the applications of intelligent tutoring systems in real educational contexts: A systematic literature review from the social experiment perspective,

    H. Wang, A. Tlili, R. Huang, et al., “Examining the applications of intelligent tutoring systems in real educational contexts: A systematic literature review from the social experiment perspective,” Educ. Inf. Technol., vol. 28, pp. 9113–9148, 2023, doi: 10.1007/s10639-022-11555- x

    work page doi:10.1007/s10639-022-11555- 2023

  30. [30]

    The comparison of two automated feedback approaches based on automated analysis of the online asynchronous interaction: a case of massive online teacher training,

    N. Ma, Y. L. Zhang, C. P. Liu, and L. Du, “The comparison of two automated feedback approaches based on automated analysis of the online asynchronous interaction: a case of massive online teacher training,” Interact. Learn. Environ., vol. 32, no. 7, pp. 38 18–3839, 2023, doi: 10.1080/10494820.2023.2191252

    work page doi:10.1080/10494820.2023.2191252 2023

  31. [31]

    Designing an automated assessment of public speaking skills using multimodal cues,

    L. Chen, G. Feng, C. W. Leong, J. Joe, C. Kitchen, and C. M. Lee, “Designing an automated assessment of public speaking skills using multimodal cues,” J. Learn. Anal., vol. 3, no. 2, pp. 261 –281, 2016, doi: 10.18608/jla.2016.32.13

    work page doi:10.18608/jla.2016.32.13 2016

  32. [32]

    AutoTutor: A tutor with dialogue in natural language,

    A. C. Graesser, S. Lu, G. T. Jackson, H. H. Mitchell, and A. Olney, “AutoTutor: A tutor with dialogue in natural language,” Behav. Res. Methods Instrum. Comput., vol. 36, no. 2, pp. 180 –192, 2004, doi: 10.3758/BF03195563

    work page doi:10.3758/bf03195563 2004

  33. [33]

    ATTENDEE: an affective tutoring system based on facial emotion recognition and head pose estimation to personalize e-learning environment,

    M. Pourmirzaei, G. A. Montazer, and E. Mousavi, “ATTENDEE: an affective tutoring system based on facial emotion recognition and head pose estimation to personalize e-learning environment,” J. Comput. Educ., vol. 12, pp. 65–92, 2025, doi: 10.1007/s40692-023-00303-w

    work page doi:10.1007/s40692-023-00303-w 2025

  34. [34]

    Thin slices of expressive behavior as predictors of interpersonal consequences: A meta-analysis,

    N. Ambady and R. Rosenthal, “Thin slices of expressive behavior as predictors of interpersonal consequences: A meta-analysis,” Psychol. Bull., vol. 111, pp. 256–274, 1992, doi: 10.1037/0033-2909.111.2.256

    work page doi:10.1037/0033-2909.111.2.256 1992

  35. [35]

    The expressive balance effect: Perception and physiological responses of prosody and gestures,

    E. Rodero, O. Larrea, I. Rodrí guez-De-Dios, and I. Lucas, “The expressive balance effect: Perception and physiological responses of prosody and gestures,” J. Lang. Soc. Psychol., vol. 41, pp. 659 –684, 2022, doi: 10.1177/0261927X221078317

    work page doi:10.1177/0261927x221078317 2022

  36. [36]

    KHFA: Knowledge -Driven Hierarchical Feature Alignment Framework for Subject -Invariant Facial Action Unit Detection,

    H. Zhao, S. He, C. Du, L. Liu, and L. Yu, “KHFA: Knowledge -Driven Hierarchical Feature Alignment Framework for Subject -Invariant Facial Action Unit Detection,” IEEE Trans. Instrum. Meas., vol. 73, pp. 1 –14,

    work page

  37. [37]

    0.1109/TIM.2024.3361596

    work page arXiv 2024

  38. [38]

    A computational modeling approach to investigating mind wandering-related adjustments to gaze behavior during scene viewing,

    K. Krasich, K. O’Neill, S. Murray, J. R. Brockmole, F. De Brigard, and A. Nuthmann, “A computational modeling approach to investigating mind wandering-related adjustments to gaze behavior during scene viewing,” Cognition, vol. 242, p. 105624, 2023

    work page 2023

  39. [39]

    Recent developments in openSMILE, the Munich open -source multimedia feature extraction toolkit,

    F. Eyben, F. Weninger, F. Gross, and B. Schuller, "Recent developments in openSMILE, the Munich open -source multimedia feature extraction toolkit," in Proc. 21st ACM Int. Conf. Multimedia, Oct. 2013, pp. 835–838. doi: 10.1145/2502081.2502224

    work page doi:10.1145/2502081.2502224 2013

  40. [40]

    Sentence -BERT: Sentence Embeddings using Siamese BERT-Networks,

    N. Reimers and I. Gurevych, "Sentence -BERT: Sentence Embeddings using Siamese BERT-Networks," in Proc. 2019 Conf. Empirical Methods Natural Language Process. 9th Int. Joint Conf. Natural Language Process. (EMNLP-IJCNLP), Nov. 2019, pp. 3982 –3992. doi: 10.1 8653/v1/D19- 1410

    work page 2019

  41. [41]

    Adams, and Nando de Freitas

    B. Shahriari, K. Swersky, Z. Wang, R. P. Adams, and N. de Freitas, “Taking the human out of the loop: A review of Bayesian optimization,” Proc. IEEE, vol. 104, no. 1, pp. 148 –175, Jan. 2016, doi: 10.1109/JPROC.2015.2494218

    work page doi:10.1109/jproc.2015.2494218 2016

  42. [42]

    The Impact of Speaker -Independent Experiments on the Validity of Speech-Based Affective Computing,

    G. Mezgec and S. Seljak, "The Impact of Speaker -Independent Experiments on the Validity of Speech-Based Affective Computing," IEEE Access, vol. 12, pp. 15432 -15450, 2024. doi: 10.1109/ACCESS.2024.335678

    work page doi:10.1109/access.2024.335678 2024

  43. [43]

    Comparing the Pearson and Spearman correlation coefficients across distributions and sample sizes: A tutorial using simulations and empirical data,

    J. C. F. de Winter, S. D. Gosling, and J. Potter, “Comparing the Pearson and Spearman correlation coefficients across distributions and sample sizes: A tutorial using simulations and empirical data,” Psychol. Methods, vol. 21, no. 3, pp. 273–290, 2016, doi: 10.1037/met0000079

    work page doi:10.1037/met0000079 2016

  44. [44]

    Enhancing learner affective engagement: The impact of instructor emotional expressions and vocal charisma in asynchronous video-based online learning,

    H.-Y. Suen and K.-E. Hung, “Enhancing learner affective engagement: The impact of instructor emotional expressions and vocal charisma in asynchronous video-based online learning,” Educ. Inf. Technol., vol. 30, pp. 4033–4060, 2025, doi: 10.1007/s10639-024-12956-w

    work page doi:10.1007/s10639-024-12956-w 2025

  45. [45]

    Card sorting with fewer cards and the same mental models? a reexamination of an established practice.International Journal of Human–Computer Interaction, 0 (0):1–25, 2026

    H.-Y. Suen and Y. -S. Su, “Teachers’ vocal expressions and student engagement in asynchronous video learning,” Int. J. Hum. -Comput. Interact., pp. 1–12, 2025, doi: 10.1080/10447318.2025.2474469

    work page doi:10.1080/10447318.2025.2474469 2025

  46. [46]

    Whitmore and T

    S. Whitmore and T. Gaskell, Coaching for performance: The principles and practice of coaching and leadership, 6th ed. London, UK: John Murray Business, 2024

    work page 2024

  47. [47]

    Intraclass correlations: Uses in assessing rater reliability,

    P. E. Shrout and J. L. Fleiss, "Intraclass correlations: Uses in assessing rater reliability," Psychol. Bull., vol. 86, no. 2, pp. 420 –428, 1979. doi : 10.1037/0033-2909.86.2.420

    work page doi:10.1037/0033-2909.86.2.420 1979

  48. [48]

    WIP: Adaptive presentation training powered by AI,

    P. Lai, C. Chan, J. Chen, and C. Chan, “Enhancing English oral presentation skills through a rubric -based hybrid AI –peer feedback platform,” in Proc. IEEE Int. Conf. Teaching, Assessment, and Learning for Engineering (TALE), Macao, China, 2025, pp. 1 –8, do i: 10.1109/TALE66047.2025.11346692

    work page doi:10.1109/tale66047.2025.11346692 2025

  49. [49]

    WIP: Adaptive presentation training powered by AI,

    Y. Guo, H. L. Li, and H. Y. J. Lai, “WIP: Adaptive presentation training powered by AI,” in Proc. IEEE Int. Conf. Teaching, Assessment, and Learning for Engineering (TALE), Macao, China, 2025, pp. 1 –3, doi: 10.1109/TALE66047.2025.11346676

    work page doi:10.1109/tale66047.2025.11346676 2025

  50. [50]

    Evaluation of artificial intelligence as a tool for assessing presentation skills among first -year medical students at Ain Shams University,

    N. N. A. Abdel Fatah, A. S. Mohamed Bakr, H. A. M. Shaaban, S. K. Ashry, and M. A. A. Abdel -Hamid Elzahry, “Evaluation of artificial intelligence as a tool for assessing presentation skills among first -year medical students at Ain Shams University,” QJM: An International Journal of Medicine, vol. 118, suppl. 1, p. hcaf224-138, 2025. Hung-Yue Suen receiv...

    work page 2025