Pith. sign in

REVIEW 4 major objections 6 minor 74 references

The Homework Wars: Exploring Emotions, Behaviours, and Conflicts in Parent-Child Homework Interactions

T0 review · 4 major / 6 minor · reviewed 2026-08-09 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims that an LLM-based pipeline can extract and categorize parental behaviors and parent-child conflicts from naturalistic homework audio at expert-level agreement, revealing emotional costs and even positive behaviors tied to…

desk verdict A substantial corpus and credible LLM-coding validation, but the headline behaviour–conflict correlation is confounded by session length. read the letter →

arxiv 2502.01325 v3 pith:TLJKBTNL submitted 2025-02-03 cs.HC

Nan Gao , Yibin Liu , Xin Tang , Yanyan Liu , Chun Yu , Yun Huang , Yuntao Wang , Flora D. Salim
show 3 more authors
Xuhai Orson Xu Jun Wei Yuanchun Shi
This is my paper · ORCID
classification cs.HC
keywords parentalhomeworkinvolvementparent-childconflictlargelanguagemodelsqualitativecodinginsituaudiostudyPADemotionmodelChinesefamiliesubiquitouscomputing
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to show that real, messy homework conversations between Chinese parents and their primary-school children can be analysed at scale by an LLM pipeline, producing behavior and conflict labels that agree with expert human coders. It draws on 475 hours of in-home audio from 78 families over four weeks, plus daily surveys. On this corpus the authors report three substantive results: parents' pleasure and sense of control drop after homework while arousal rises; 18 recurring parental behaviors and seven conflict types can be coded reliably, with Knowledge Conflict the most frequent; and even well-intentioned behaviors such as unlabelled praise correlate with specific conflicts. If true, this would let family research move from retrospective self-reports to moment-by-moment analysis of what actually happens at the kitchen table.

What carries the argument

The load-bearing machinery is a bottom-up, expert-refined codebook combined with a GPT-4o coding loop: transcripts from the Xunfei API are corrected by GPT-4o, speaker roles are assigned by GPT-4o, then open, axial, and selective coding produce 18 behavior codes and seven conflict codes, and GPT-4o applies these codes to all 602 sessions. Validation uses Cohen's Kappa against four human experts, with majority-vote consensus as the gold standard. Emotion analysis uses the PAD/SAM self-report scales for pre-post shifts and the EmoLLaMA model for sentence-level pleasure trajectories.

What would settle it

Have professional transcribers produce gold transcripts for a random sample of the 602 sessions, rerun the GPT-4o coding and the behavior-conflict correlation analysis on those gold transcripts, and check whether Knowledge Conflict remains dominant and praise still correlates with Expectation Conflict.

Watch

Extended reading notes

Core claim

The central claim is that a modular LLM-assisted pipeline—automatic transcription, speaker-role assignment, then GPT-4o coding against an expert-built codebook—can label parental behaviors and parent-child conflicts from unstructured homework conversations with agreement comparable to trained human annotators: substantial for behaviors and moderate for conflicts. Through this lens the paper identifies 18 parental behaviors, seven conflict types, and systematic emotion dynamics, with significant PAD shifts before versus after homework and pervasive positive correlations between behavior frequencies and conflict counts. The counterintuitive headline finding is that positive behaviors are not conflict-free: even Labelled Praise and Unlabelled Praise co-occur with specific conflicts such as Expectation Conflict, which the authors explain through controlling-support and shifting-expectations mechanisms.

Load-bearing premise

The pipeline's outputs inherit whatever errors the automated transcripts contain, and since no word-error rate or speaker-role accuracy is reported, systematic transcription mistakes could reshape the behavior and conflict labels and the correlations built on them.

Editorial extensions

If this is right

  • Homework interactions in Chinese families can be automatically labeled at a scale that manual expert coding cannot reach, making large-N studies of family dynamics feasible.
  • Parental emotions are measurably worse after homework on average, with pleasure and dominance dropping and average arousal rising.
  • Knowledge Conflict is the most common friction point, consistent with a curse-of-knowledge gap between adult and child perspectives.
  • Intervention designers cannot assume positive-sounding behaviors are safe, because praise and encouragement co-occur with specific conflicts.
  • Pre-session emotional state predicts in-session behavior and conflict, pointing to emotion regulation before homework as an intervention target.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: a testable extension would be to have professional transcribers produce gold transcripts for a random sample of sessions and rerun the coding, since the paper's validation cannot recover information lost at transcription time.
  • Editorial inference: session-level correlations leave causal direction open, so conflict may elicit praise rather than praise causing conflict, and longitudinal or experimental designs are the natural next step.
  • Editorial inference: transferring the pipeline to other emotionally laden interactions such as tutoring, healthcare consultations, or parent-teacher conversations would require rebuilding the codebook, which the paper sketches but does not demonstrate.
  • Editorial inference: adding prosodic features from the audio could sharpen conflict and emotion coding, because the paper itself notes that text-only transcripts lose sarcasm, tone, and other vocal cues.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper reports a four-week in-situ study of 78 Chinese families, yielding 602 audio-recorded homework sessions. The authors use an LLM-based pipeline (Xunfei transcription, GPT-4o error correction, GPT-4o role assignment and coding) to extract 18 parental behaviour categories and 7 conflict types, validate the coding against four human experts, and then report relationships among self-reported emotions, behaviours, and conflicts. The headline empirical claims are that parents' pleasure and dominance drop after homework, that Knowledge Conflict is the most frequent conflict type, and that even positive behaviours such as Labelled Praise and Unlabelled Praise correlate significantly with specific conflicts.

Significance. If the results hold, the paper would make a valuable empirical contribution: a large naturalistic corpus of homework interactions, a codebook tailored to Chinese families, and evidence that LLM coding can approach expert-level reliability on behaviour annotation. The expert validation (Kappa 0.562-0.724 for behaviour coding, 0.410-0.517 for conflict coding) is a genuine strength, as is the transparency of the prompts in the appendix. However, the central correlational finding is currently vulnerable to a session-length confound, and the reliance on LLM-generated labels for the full dataset raises a circularity risk that the paper does not fully address.

major comments (4)
  1. [Section 6, Figure 11] The headline finding that even well-intentioned behaviours correlate with conflicts rests on correlations between raw behaviour counts and raw conflict counts per session. Since session durations vary widely (mean 47.33 minutes, Figure 1), longer sessions mechanically produce more opportunities for both behaviours and conflicts; this can generate pervasive positive correlations even when the per-minute rates are independent. The paper does not report partial correlations controlling for session duration, rate-based analyses, or a multilevel model with session length as a covariate. In addition, the 18x7 = 126 tests are not adjusted for multiple comparisons, and sessions are nested within families (65 families, 511 sessions), so the reported p-values overstate evidence. The authors should re-analyse Figure 11 using rates per minute or partial correlations that include session duration and family as random effects, and should report corrected significance levels.
  2. [Sections 5.1.1, 5.2, and 6] The behaviour and conflict taxonomies were inductively derived with GPT-4o, and GPT-4o was then used to label the entire dataset for the correlation analysis. Human expert validation covers only 200 randomly selected instances per coding task, so the full-dataset labels used in Figure 11 are not independently grounded. If the model has systematic co-labelling tendencies (e.g., it tends to label praise together with expectation conflict because the prompt examples associate them), the correlations could be inflated or induced. The authors should test the robustness of the Figure 11 correlations by recomputing them on the expert-consensus-coded subset or by having human coders adjudicate a random sample of sessions used in the correlation analysis.
  3. [Section 4.1] There is an internal inconsistency in the reported arousal shift. The text says 'parents experienced a decrease in pleasure, a decrease in arousal, and a reduction in their sense of control (dominance)', but the individual-level analysis reports an average arousal shift of +0.535, which is an increase. The group-level and individual-level statements cannot both be correct as written; the authors should correct the sign or clarify whether the PAD arousal scale is reversed. Since the emotional-shift finding is one of the paper's three research questions, this contradiction needs to be resolved before the result can be interpreted.
  4. [Sections 3.4.2 and 3.4.3] The pipeline depends entirely on automatic transcription and automatic speaker-role assignment, yet no word-error rate, speaker-role accuracy, or segmentation accuracy is reported. A human-expert validation of behaviour and conflict coding was performed on the same transcripts, so it cannot recover information lost at transcription time. Systematic transcription errors (homophone confusion, merged or dropped speakers, mis-segmented turns) would propagate into every downstream behaviour and conflict label and could distort the counts and correlations in Section 6. The authors should evaluate transcription and role-assignment quality on a small human-transcribed holdout set and report the impact of transcription error on coding reliability.
minor comments (6)
  1. [Table 3 caption] The caption says the conflict examples were 'synthesised by ChatGPT', while the paper consistently uses GPT-4o elsewhere; please unify the terminology.
  2. [Section 6, reference [10]] The sentence 'This echoes Pomerantz et al.'s [10] findings' cites reference [10], which is Birch and Bloom's 'curse of knowledge' paper, not a Pomerantz study; the citation appears to be incorrect.
  3. [Reference [61]] The title of reference [61] contains a typo: 'deductive doding' should be 'deductive coding'.
  4. [Section 6] The transition from 602 sessions / 78 families to 511 sessions / 65 families is mentioned only as 'excluding one due to insufficient data'; please specify the exclusion criteria and report how many sessions were dropped per family.
  5. [Abstract] The abstract says the pipeline achieved 'high agreement' with experts, but the conflict coding Kappa against consensus is 0.517 (moderate). Consider describing the agreement as 'moderate to substantial' to match the reported values.
  6. [Section 7] The limitations section lists sampling bias, text-only analysis, Hawthorne effect, and privacy, but does not mention the absence of transcription-error evaluation or the session-duration confound; these should be acknowledged or addressed.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: LLM coding is validated against independent expert annotations, and the correlational results are empirical tallies rather than fitted or definitionally forced outputs.

full rationale

The paper's derivation chain is not circular. The central method claim is that a GPT-4o pipeline extracts parental behaviours and conflicts from transcripts; this is evaluated against four human experts using Cohen's Kappa (Section 5.2, Table 4), an external benchmark that does not derive from the model itself. The behaviour and conflict codebooks were developed with GPT-4o assistance and then reviewed and refined by human education experts (Section 5.1.1), so the taxonomy is not a direct model output pulled through to the results. The prevalence statistics (Figures 6-10) and correlation analyses (Figures 11-12) are computed from the coded transcripts; they are descriptive tallies, not predictions from fitted parameters. No equation or fitted value is renamed as a finding. The session-duration confound raised about Figure 11 is a statistical validity concern, not a circularity by construction: it does not make the correlation equal to its input. The one self-citation (reference [26] on self-report bias) appears in related work and is not load-bearing for any central claim. The human-expert validation provides independent support for the label-reliability claim, and the correlation claims, while potentially affected by shared model bias, are not equivalent to the coding inputs by definition. Therefore no significant circularity is present.

Assumptions & free parameters 3 free parameters · 5 assumptions · 2 invented entities

The central claims rest on transcription quality, the validity of a self-built codebook, and self-reported emotion measures. No numeric parameters are fit to data, but several analysis thresholds are chosen by hand. The taxonomies are the paper's main invented constructs; they are well-specified enough to be applied by others, which gives them some independent evidential handle.

free parameters (3)
  • 2-second pause segmentation threshold
    Chosen by hand as the turn-boundary rule for Xunfei transcription; it defines what counts as a conversational unit and affects all subsequent coding (Section 3.4.2).
  • 15-second emotion binning window
    Selected as the averaging interval for pleasure trajectories; the choice trades detail against sparsity and shapes the emotion curves in Figure 5 (Section 4.2.2).
  • First-10-minutes analysis window
    Used to standardize sessions of different lengths; longer sessions are truncated, so the emotion trajectory results cover only the start of homework (Section 4.2.2).
assumptions (5)
  • domain assumption Automated transcription with Xunfei API and GPT-4o error correction accurately preserves the dialogue, speaker roles, and turn structure.
    Invoked in Sections 3.4.2 and 3.4.3 without a word-error-rate or role-accuracy evaluation; all behavior and conflict labels are derived from these transcripts.
  • domain assumption Text transcripts alone, without prosody or video, contain enough information to identify parental behaviors and parent-child conflicts.
    Acknowledged as a limitation in Section 7; the coding is performed on text, so tone, pitch, and non-verbal actions are absent.
  • domain assumption Expert consensus, formed by majority voting among four human experts, is a valid gold standard for behavior and conflict coding.
    Used in Section 5.2 to evaluate GPT-4o; human disagreement is substantial for conflict coding (kappa range 0.330 to 0.738).
  • ad hoc to paper The inductively derived codebook of 18 behaviors and 7 conflict types is a complete and valid taxonomy for Chinese parent-child homework interactions.
    Developed through GPT-4o open coding and expert refinement (Section 5.1.1); no external validation against an established instrument is provided.
  • domain assumption Self-reported SAM/PAD ratings before and after homework measure parents' emotional states without systematic bias.
    Used in Section 4.1; self-report can be biased, and the paper itself criticizes self-reports in related work.
invented entities (2)
  • 18-category parental behavior taxonomy independent evidence
    purpose: Classifies each parental action during homework (praise, instruction, commands, criticism, etc.) for quantitative analysis.
    Codified in Table 2 with usage guidelines in Table 6, so other researchers can apply it to new transcripts; however, the categories were initially generated by GPT-4o.
  • 7-type parent-child conflict taxonomy independent evidence
    purpose: Classifies conflict episodes (knowledge, expectation, rules, time, etc.) from homework dialogue.
    Codified in Table 3 with guidelines in Table 7 and example dialogues; the categories are co-constructed with GPT-4o and refined by educational experts.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Homework Wars: Exploring Emotions, Behaviours, and Conflicts in Parent-Child Homework Interactions." pith.science (2026). https://pith.science/paper/TLJKBTNL

@misc{pith2026250201325,
  author       = {Pith},
  title        = {Pith review of: The Homework Wars: Exploring Emotions, Behaviours, and Conflicts in Parent-Child Homework Interactions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TLJKBTNL}},
  note         = {Machine review of arXiv:2502.01325}
}
read the original abstract

Parental involvement in homework is a crucial aspect of family education, but it often triggers emotional strain and conflicts. Despite growing concern over its impact on family well-being, prior research has lacked access to fine-grained, real-time dynamics of these interactions. To bridge this gap, we present a framework that leverages naturalistic parent-child interaction data and large language models (LLMs) to analyse homework conversations at scale. In a four-week in situ study with 78 Chinese families, we collected 475 hours of audio recordings and accompanying daily surveys, capturing 602 homework sessions in everyday home settings. Our LLM-based pipeline reliably extracted and categorised parental behaviours and conflict patterns from transcribed conversations, achieving high agreement with expert annotations. The analysis revealed significant emotional shifts in parents before and after homework, 18 recurring parental behaviours and seven common conflict types, with Knowledge Conflict being the most frequent. Notably, even well-intentioned behaviours were significantly positively correlated with specific conflicts. This work advances ubiquitous computing methods for studying complex family dynamics and offers empirical insights to enrich family education theory and inform more effective parenting strategies and interventions in the future.

Figures

Figures reproduced from arXiv: 2502.01325 by the authors.

Figure 2
Figure 2. Number of audio recordings for different partici [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Distribution of parental emotions before and after homework involvement. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Mean emotion shifts after homework involvement. The error bars represent the 95% confidence interval. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figures from the paper (11 more)
Figure 5
Figure 5. Figure 5: Average pleasure for the first 10 minutes of sessions. The dark grey band indicates the standard error. [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Average number of positive, neutral, and negative be [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 8
Figure 8. Figure 8: Average frequency of positive, neutral, and negative behaviours per homework session for each parent. [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: Proportion of various negative behaviours per homework session for each parent. [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]
Figure 10
Figure 10. Figure 10: Average frequency of various conflict types per homework session for each parent. [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]
Figure 11
Figure 11. Figure 11: Correlation between parental behaviours and parent-child conflicts (* p<0.05, ** p<0.01, *** p<0.001). [PITH_FULL_IMAGE:figures/full_fig_p017_11.png]
Figure 12
Figure 12. Figure 12: Correlation between emotions, behaviours and parent-child conflicts (* p<0.05, ** p<0.01, *** p<0.001). [PITH_FULL_IMAGE:figures/full_fig_p019_12.png]
Figure 13
Figure 13. Figure 13: Different impacts of recording on educational behaviours (self-reported). [PITH_FULL_IMAGE:figures/full_fig_p020_13.png]
Figure 14
Figure 14. Figure 14: Number of responses for different participants. Linguistics 26, 3 (2000), 339–373. https://doi.org/10.1162/089120100561737 [56] Anselm Strauss and Juliet Corbin. 1994. Grounded theory methodology: An overview. (1994). [57] Natalia Suárez Fernández, María Estrella Fern…
Figure 15
Figure 15. Figure 15: Confusion matrix of parental behaviour classification by GPT-4o. Diagonal cells indicate correct classifications, and [PITH_FULL_IMAGE:figures/full_fig_p025_15.png]
Figure 16
Figure 16. Figure 16: Confusion matrix of parent-child conflict classification by GPT-4o. Diagonal cells indicate correct matches and [PITH_FULL_IMAGE:figures/full_fig_p025_16.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

74 extracted references · 64 canonical work pages

  1. [1]

    [n. d.]. China - Overview of the education system (EAG 2023). http://gpseducation.oecd.org/

    work page 2023

  2. [2]

    [n. d.]. China Family Panel Studies. https://www.isss.pku.edu.cn/cfps/index.htm

    work page

  3. [3]

    [n. d.]. Speech Transcription API Documentation | iFLYTEK Open Platform Documentation Center. https://www.xfyun.cn/doc/asr/ lfasr/API.html#%E6%8E%A5%E5%8F%A3%E8%B0%83%E7%94%A8%E6%B5%81%E7%A8%8B

    work page

  4. [4]

    John G Adair. 1984. The Hawthorne effect: a reconsideration of the methodological artifact. Journal of Applied Psychology 69, 2 (1984),

    work page 1984

  5. [5]

    Mohammed Aldeen, Jeffrey Young, Song Liao, Tsu-Yao Chang, Long Cheng, Haipeng Cai, Xiapu Luo, and Hongxin Hu. 2024. End-Users Know Best: Identifying Undesired Behavior of Alexa Skills Through User Review Analysis. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 8, 3 (2024), 1–28. https://doi.org/10.1145/3678517

    work page doi:10.1145/3678517 2024

  6. [6]

    Willem A Arrindell, Ezio Sanavio, Guido Aguilar, Claudio Sica, Chryse Hatzichristou, Martin Eisemann, Luis A Recinos, Peter Gaszner, Monika Peter, Giuseppe Battagliese, et al. 1999. The development of a short form of the EMBU: Its appraisal with students in Greece, Guatemala, Hungary and Italy.Personality and Individual Differences 27, 4 (1999), 613–628. ...

    work page doi:10.1016/s0191-8869(98)00192-5 1999

  7. [7]

    Diana Baumrind. 1991. The influence of parenting style on adolescent competence and substance use. The Journal of Early Adolescence 11, 1 (1991), 56–95

    work page 1991

  8. [8]

    Sharon Begley. 1998. Homework doesn’t help. Newsweek 131, 13 (1998), 50–52

    work page 1998

Show all 74 references
  1. [9]

    Lisa Benckwitz, Katharina Kohl, Janina Roloff, Oliver Lüdtke, and Karin Guill. 2023. Reciprocal relationships between parental and scholastic homework assistance and students’ academic functioning at elementary school. Frontiers in Psychology 14 (2023), 1106362. https://doi.or...

    2023

  2. [10]

    Susan AJ Birch and Paul Bloom. 2007. The curse of knowledge in reasoning about false beliefs.Psychological Science 18, 5 (2007), 382–386. https://doi.org/10.1111/j.1467-9280.2007.01909.x

    2007

  3. [11]

    Ljubiša Bojić, Olga Zagovora, Asta Zelenkauskaite, Vuk Vuković, Milan Čabarkapa, Selma Veseljević Jerković, and Ana Jovančević. 2025. Comparing large Language models and human annotators in latent content analysis of sentiment, political leaning, emotional intensity and sarcas...

    2025 doi

  4. [12]

    Margaret M Bradley and Peter J Lang. 1994. Measuring emotion: the self-assessment manikin and the semantic differential. Journal of Behavior Therapy and Experimental Psychiatry 25, 1 (1994), 49–59. https://doi.org/10.1016/0005-7916(94)90063-9

    1994 doi

  5. [13]

    Sven Buechel and Udo Hahn. 2022. EmoBank: Studying the Impact of Annotation Perspective and Representation Format on Dimensional Emotion Analysis. arXiv:2205.01996 [cs.CL] https://arxiv.org/abs/2205.01996

    2022 arXiv

  6. [14]

    Kathy Charmaz. 2006. Constructing grounded theory: A practical guide through qualitative analysis . SAGE

    2006

  7. [15]

    Cecilia S Cheung, Eva M Pomerantz, Meifang Wang, and Yang Qu. 2016. Controlling and autonomy-supportive parenting in the United States and china: Beyond children’s reports. Child Development 87, 6 (2016), 1992–2007

    2016

  8. [16]

    Harris Cooper. 1989. Synthesis of research on homework. Educational Leadership 47, 3 (1989), 85–91

    1989

  9. [17]

    Lyn Corno and Jianzhong Xu. 2004. Homework as the job of childhood. Theory into Practice 43, 3 (2004), 227–233. https://doi.org/10. 1207/s15430421tip4303_9

    2004

  10. [18]

    Jennifer Cunha, Pedro Rosário, Lúcia Macedo, Ana Rita Nunes, Sonia Fuentes, Ricardo Pinto, and Natalia Suárez. 2015. Parents’ concep- tions of their homework involvement in elementary school.Psicothema 27, 2 (2015), 159–165. https://doi.org/10.7334/psicothema2014.210

    2015 doi

  11. [19]

    Shih-Chieh Dai, Aiping Xiong, and Lun-Wei Ku. 2023. LLM-in-the-loop: Leveraging Large Language Model for Thematic Analysis. arXiv:2310.15100 [cs.CL] https://arxiv.org/abs/2310.15100

    2023 arXiv

  12. [20]

    Susanne A Denham. 1998. Emotional development in young children . Guilford Press

    1998

  13. [21]

    Swantje Dettmers, Sittipan Yotyodying, and Kathrin Jonkmann. 2019. Antecedents and outcomes of parental homework involvement: How do family-school partnerships affect parental homework involvement and student outcomes? Frontiers in Psychology 10 (2019),

    2019

  14. [22]

    Hanna Dumont, Ulrich Trautwein, Gabriel Nagy, and Benjamin Nagengast. 2014. Quality of parental homework involvement: predictors and reciprocal relations with academic functioning in the reading domain. Journal of Educational Psychology 106, 1 (2014), 144. https://doi.org/10.1...

    2014 doi

  15. [23]

    Zackary Okun Dunivin. 2025. Scaling hermeneutics: a guide to qualitative coding with LLMs for reflexive content analysis. EPJ Data Science 14, 1 (2025), 28. https://doi.org/10.1140/epjds/s13688-025-00548-8

    2025 doi

  16. [24]

    Jacquelynne S Eccles and Rena D Harold. 2013. Family involvement in children’s and adolescents’ schooling. In Family-school links. Routledge, 3–34

    2013

  17. [25]

    Yongtao Gan and Sude Bilige. 2019. Parental involvement in home-based education and children’s academic achievement in China. Social Behavior and Personality 47, 12 (2019), 1–15. https://doi.org/10.2224/sbp.8491

    2019 doi

  18. [26]

    Nan Gao, Mohammad Saiedur Rahaman, Wei Shao, and Flora D Salim. 2021. Investigating the reliability of self-report data in the wild: The quest for ground truth. In Adjunct Proceedings of the 2021 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proc...

    2021

  19. [27]

    Barney Glaser and Anselm Strauss. 2017. Discovery of grounded theory: Strategies for qualitative research . Routledge

    2017

  20. [28]

    Marjory Roberts Gray and Laurence Steinberg. 1999. Unpacking authoritative parenting: Reassessing a multidimensional construct. Journal of Marriage and the Family (1999), 574–587. https://doi.org/10.2307/353561

    1999 doi

  21. [29]

    Wendy S Grolnick and Eva M Pomerantz. 2009. Issues and challenges in studying parental control: Toward a new conceptualization. Child Development Perspectives 3, 3 (2009), 165–170. https://doi.org/10.1111/j.1750-8606.2009.00099.x

    2009

  22. [30]

    Suzanne T Gurland and Wendy S Grolnick. 2005. Perceived threat, controlling parenting, and children’s achievement orientations. Motivation and Emotion 29 (2005), 103–121. https://doi.org/10.1007/s11031-005-7956-2 Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 9, N...

    2005 doi

  23. [31]

    Nguyen Thi Hong Hanh. 2023. The Current Situation of Conflict between Parents and Teenage Children. EAS Journal of Psychology and Behavioural Sciences 5, 4 (2023), 70–76. https://doi.org/10.36349/easjpbs.2023.v05i04.001

    2023 doi

  24. [32]

    S Heimgartner-Moroni, H Dumont, U Trautwein, A Niggli, and F Baeriswyl. 2012. More chameleon effects in homework research: How the choice of instruments affects the associations between family background, parental homework involvement, and reading achievement. Dissertation, Eb...

    2012

  25. [33]

    Kathleen V Hoover-Dempsey and Howard M Sandler. 2005. The social context of parental involvement: A path to enhanced achievement. Vanderbilt University, Institute of Educational Sciences Department of Education (2005)

    2005

  26. [34]

    Sungwon Kim and Vanessa L Fong. 2013. How parents help children with homework in China: Narratives across the life span. Asia Pacific Education Review 14 (2013), 581–592. https://doi.org/10.1007/s12564-013-9284-7

    2013 doi

  27. [35]

    Wonjung Kim, Seungchul Lee, Seonghoon Kim, Sungbin Jo, Chungkuk Yoo, Inseok Hwang, Seungwoo Kang, and Junehwa Song. 2020. Dyadic mirror: Everyday second-person live-view for empathetic reflection upon parent-child interaction. Proceedings of the ACM on Interactive, Mobile, Wea...

    2020 doi

  28. [36]

    J Richard Landis and Gary G Koch. 1977. The measurement of observer agreement for categorical data. Biometrics (1977), 159–174

    1977

  29. [37]

    Siddique Latif, Muhammad Usama, Mohammad Ibrahim Malik, and Björn W Schuller. 2023. Can large language models aid in annotating speech emotional data? uncovering new frontiers. arXiv preprint arXiv:2307.06090 (2023)

    2023 arXiv

  30. [38]

    Diane Litman and Kate Forbes. 2003. Recognizing emotions from student speech in tutoring dialogues. In 2003 IEEE Workshop on Automatic Speech Recognition and Understanding . IEEE, 25–30. https://doi.org/10.1109/ASRU.2003.1318398

    2003 arXiv

  31. [39]

    Xiner Liu, Andres Felipe Zambrano, Ryan S Baker, Amanda Barany, Jaclyn Ocumpaugh, Jiayi Zhang, Maciej Pankiewicz, Nidhi Nasiar, and Zhanlan Wei. 2025. Qualitative coding with GPT-4: Where it works better. Journal of Learning Analytics 12, 1 (2025), 169–185. https://doi.org/10....

    2025

  32. [40]

    Xiner Liu, Jiayi Zhang, Amanda Barany, Maciej Pankiewicz, and Ryan S. Baker. 2024. Assessing the Potential and Limits of Large Language Models in Qualitative Coding. In Advances in Quantitative Ethnography , Yoon Jeon Kim and Zachari Swiecki (Eds.). Springer Nature Switzerland...

    2024 doi

  33. [41]

    Zhiwei Liu, Kailai Yang, Qianqian Xie, Tianlin Zhang, and Sophia Ananiadou. 2024. Emollms: A series of emotional large language models and annotation tools for comprehensive affective analysis. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mi...

    2024

  34. [42]

    while trying to help his son with his homework

    Emilia Jiang For Mailonline. 2020. Frustrated Chinese father suffers a heart attack "while trying to help his son with his homework". https: //www.dailymail.co.uk/news/article-8789405/Chinese-father-suffers-heart-attacks-getting-frustrated-helping-son-homework.html

    2020

  35. [43]

    Elaine K McEwan. 2004. How to deal with parents who are angry, troubled, afraid, or just plain crazy . Corwin Press

    2004

  36. [44]

    Albert Mehrabian. 1974. An approach to environmental psychology . The MIT Press

    1974

  37. [45]

    Angelica Moè and Idit Katz. 2018. Brief research report: Parents’ homework emotions favor students’ homework emotions through self-efficacy. The Journal of Experimental Education 86, 4 (2018), 597–609. https://doi.org/10.1080/00220973.2017.1409180

    2018

  38. [46]

    Ogechi Nnamani, Ebere J Idoko, Liziana N Onuigbo, and Uchenna Eze. 2020. Impact of parental involvement on emotional adjustment and school achievement of primary school pupils. Journal Of The Nigerian Academy of Education 16, 1 (2020)

    2020

  39. [47]

    NZ Herald7 Nov and 7 Nov. 2020. Mum almost dies from heart attack when son couldn’t do maths prob- lem. https://www.nzherald.co.nz/lifestyle/chinese-mum-almost-dies-from-heart-attack-when-son-couldnt-do-maths-problem/ 3NJMYW5MDCKRUKG7IOJ6MY3IWE/

    2020

  40. [48]

    José Carlos Núñez, Natalia Suárez, Pedro Rosário, Guillermo Vallejo, Antonio Valle, and Joyce L Epstein. 2015. Relationships between perceived parental involvement in homework, student homework behaviors, and academic achievement: differences among elementary, junior high, and...

    2015 doi

  41. [49]

    Erika A Patall, Harris Cooper, and Jorgianne Civey Robinson. 2008. Parent involvement in homework: A research synthesis. Review of Educational Research 78, 4 (2008), 1039–1101. https://doi.org/10.3102/0034654308325185

    2008 doi

  42. [50]

    Reinhard Pekrun. 2006. The control-value theory of achievement emotions: Assumptions, corollaries, and implications for educational research and practice. Educational psychology review 18 (2006), 315–341

    2006

  43. [51]

    Reinhard Pekrun, Thomas Goetz, Wolfram Titz, and Raymond P Perry. 2002. Academic emotions in students’ self-regulated learning and achievement: A program of qualitative and quantitative research. Educational Psychologist 37, 2 (2002), 91–105. https://doi.org/10. 1207/S15326985EP3702_4

    2002

  44. [52]

    Eva M Pomerantz, Elizabeth A Moorman, and Scott D Litwack. 2007. The how, whom, and why of parents’ involvement in children’s academic lives: more is not always better. Review of Educational Research 77, 3 (2007), 373–410. https://doi.org/10.3102/003465430305567

    2007 doi

  45. [53]

    Richard M Ryan. 1982. Control and information in the intrapersonal sphere: An extension of cognitive evaluation theory. Journal of Personality and Social Psychology 43, 3 (1982), 450. https://doi.org/10.1037/0022-3514.43.3.450

    1982 doi

  46. [54]

    Yvette Solomon, Jo Warin, and Charlie Lewis. 2002. Helping with homework? Homework as a site of tension for parents and teenagers. British Educational Research Journal 28, 4 (2002), 603–622. https://doi.org/10.1080/0141192022000005850a

    2002 doi

  47. [55]

    Andreas Stolcke, Klaus Ries, Noah Coccaro, Elizabeth Shriberg, Rebecca Bates, Daniel Jurafsky, Paul Taylor, Rachel Martin, Carol Van Ess- Dykema, and Marie Meteer. 2000. Dialogue act modeling for automatic tagging and recognition of conversational speech.Computational Proc. AC...

    2000

  48. [56]

    Anselm Strauss and Juliet Corbin. 1994. Grounded theory methodology: An overview. (1994)

    1994

  49. [57]

    Natalia Suárez Fernández, María Estrella Fernández Alba, Bibiana Regueiro, Pedro Rosário, Jianzhong Xu, José Carlos Núñez Pérez, et al

  50. [58]

    Benjamin Tag, Zhanna Sarsenbayeva, Anna L Cox, Greg Wadley, Jorge Goncalves, and Vassilis Kostakos. 2022. Emotion trajectories in smartphone use: Towards recognizing emotion regulation in-the-wild. International Journal of Human-Computer Studies 166 (2022), 102872. https://doi...

    2022

  51. [59]

    Petter Törnberg. 2024. Large language models outperform expert coders and supervised classifiers at annotating political social media messages. Social Science Computer Review (2024), 08944393241286471. https://doi.org/10.1177/0894439324128647

    2024 doi

  52. [60]

    Sara F Waters, Helena Rose Karnilowicz, Tessa V West, and Wendy Berry Mendes. 2020. Keep it to yourself? Parent emotion suppression influences physiological linkage and interaction behavior. Journal of Family Psychology 34, 7 (2020), 784. https://doi.org/10.1037/ fam0000664

    2020

  53. [61]

    Vera Liao, Rania Abdelghani, and Pierre-Yves Oudeyer

    Ziang Xiao, Xingdi Yuan, Q. Vera Liao, Rania Abdelghani, and Pierre-Yves Oudeyer. 2023. Supporting qualitative analysis with large language models: combining codebook with GPT-3 for deductive doding. In Companion Proceedings of the 28th International Conference on Intelligent ...

    2023

  54. [62]

    Jianzhong Xu. 2005. Homework emotion management reported by high school students. School Community Journal 15, 2 (2005), 21–36. Appendix A More Information about Collected Data Figure 14 shows the number of responses (daily surveys paired with audio recordings) for each partic...

    2005

  55. [66]

    **Parse the Dictionary List**: Read through each transcription entry

  56. [67]

    **Identify Transcription Errors**: Carefully examine and identify homophones or typographical errors in the `content` section. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 9, No. 3, Article 80. Publication date: September 2025. 80:28 • Gao et al

    2025

  57. [68]

    **Correct Errors**: Correct the errors without altering the original sentence style to improve the clarity and accuracy of the text

  58. [69]

    **Maintain Original Style**: When updating the `content` field, ensure that the original expression style is preserved, striving to maintain the natural flow of the original text

  59. [70]

    id": The conversation ID. -

    **Structure Validation**: After correction, ensure that the structure of each dictionary remains identical to the corresponding original dictionary, including but not limited to field order and data types. ## Output Format Please return the output in the following strict forma...

    2025

  60. [71]

    Role inference should be based on the content of the conversation, not just the sequence of speakers

  61. [72]

    others" when it is very clear; otherwise, do not infer

    For three or more speakers, carefully distinguish between roles. Only label as "others" when it is very clear; otherwise, do not infer "others"

  62. [73]

    others". Consider marking

    For each speaker, analyze their speech content in detail to avoid mistakenly labeling an actual participant in the conversation as "others". Consider marking "others" only in very rare cases. E Prompt for Understanding Behavioural Dynamics During Homework Involvement E.1 Promp...

    2025

  63. [74]

    How can you be so useless? You can 't even do this right!

    High 2. Medium 3. Low ### Judgment Criteria: #### High - **Tone**: The tone of the conversation is extremely intense, often involving shouting, criticism, or scolding. - **Severity of Language**: The language includes insulting, belittling, or blaming remarks. - **Length**: Th...

    2025

  64. [334]

    Due to the fact that you knew the homework involvement behaviour today was recorded, did this affect your true performance?

    https://doi.org/10.1037/0021-9010.69.2.334 4The question is “Due to the fact that you knew the homework involvement behaviour today was recorded, did this affect your true performance?” Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., Vol. 9, No. 3, Article 80. Publicati...

    doi

  65. [1048]

    https://doi.org/10.3389/fpsyg.2019.01048

    2019

  66. [2022]

    Psicothema (2022)

    Parental involvement in homework during Covid-19 confinement. Psicothema (2022). https://doi.org/10.7334/psicothema2021.532

    2022 doi

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.