Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Fine-tuning ChatGPT for Automatic Scoring of Written Scientific Explanations in Chinese

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Fine-tuned ChatGPT can automatically score Chinese students' written scientific explanations with accuracy above 0.75 in every scoring category, but its accuracy depends systematically on reasoning complexity and linguistic style…

desk verdict First fine-tuned ChatGPT scoring study for Chinese scientific explanations, but its central complexity-accuracy finding rests on coding with no reliability evidence. read the letter →

arxiv 2501.06704 v1 pith:HTLZC2RC submitted 2025-01-12 cs.AI cs.CL

Jie Yang , Ehsan Latif , Yuze He , Xiaoming Zhai This is my paper · ORCID
classification cs.AIcs.CL
keywords ChatGPTfine-tuningautomaticscoringscientificexplanationsChinesereasoningcomplexitylinguisticfeaturesLLM
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 asks whether ChatGPT, a large language model built mainly for English, can be adapted to grade Chinese students' written scientific explanations—a logographic language with implicit, holistic, and circular reasoning styles. It reports that after fine-tuning on 7,626 responses from 1,593 middle and high school students across seven tasks, the model agrees with human raters on at least 75% of cases in every scoring category, with holistic accuracy ranging from 80% to 94%. The accuracy is not uniform: for students whose answers received a human score of 0, higher reasoning complexity predicted lower machine–human agreement, while for students scored 1, higher reasoning complexity predicted higher agreement. The paper attributes this split to linguistic features, finding that simple, short, clear responses score accurately at lower performance levels, whereas long, comprehensive, information-rich responses score accurately at higher performance levels. If correct, the result supports using fine-tuned LLMs for Chinese automatic scoring while warning that scoring biases depend on reasoning style, not just on overall model quality.

What carries the argument

The load-bearing machinery is the pairing of a domain-specific fine-tuning dataset with a rubric-aligned scoring task and a complexity-stratified analysis. Student responses were transcribed from Chinese handwriting, cleaned, and tokenized, then used to fine-tune ChatGPT on both a holistic rubric (correct or incorrect) and the PTDR (phenomenon–theory–data–reasoning) analytic rubric, with four or five labels per item. The analysis then coded each response into low, medium, or high reasoning complexity using criteria adapted from prior work, split responses into lower- and higher-performing groups based on the human score, and computed Kendall's tau-c between reasoning complexity and machine–human agreement in each subgroup. A qualitative comparison of correctly scored and misscored responses at the word and sentence level (technical terms, sentence length, syntax, prepositional phrases, formulas) completes the mechanism. The machinery that produces the paper's main result is this crosstab comparison of complexity and agreement within performance groups, not the model alone.

What would settle it

Take a random sample of 200 responses from the seven items, have two independent raters who have not seen the paper's labels re-code each response as low, medium, or high reasoning complexity, and compute Cohen's kappa between the raters and with the paper's labels. If agreement is below 0.8, the reported Kendall correlations rest on an unreliable predictor; if agreement is high, a second test is to rewrite a set of high-scoring concise responses into longer, syntactically complex versions with identical science content and check whether the fine-tuned model's accuracy drops, which would confirm the paper's linguistic explanation.

Watch

Extended reading notes

Core claim

The paper's central claim is that a ChatGPT model fine-tuned on 7,626 Chinese middle- and high-school students' responses to seven scientific explanation tasks can score new Chinese responses with practically useful accuracy, with every fine-tuned model exceeding 0.75 accuracy on every rubric category and holistic accuracy ranging from 80% (Items 3 and 4) to 94% (Item 6). The paper further claims that this accuracy is not uniform: after splitting students into lower-level (human score 0) and higher-level (human score 1) performance groups, 29 of 30 Kendall correlations between reasoning complexity and scoring accuracy in the lower-level group were negative (28 statistically significant, mean coefficient -0.31), while 26 of 30 correlations in the higher-level group were positive (10 significant, mean of significant coefficients 0.19). Stated as the authors would state it: fine-tuned ChatGPT can grade Chinese scientific explanations, but it systematically over-credits superficially complex incorrect answers and under-credits concise correct answers, and this bias is tied to linguistic features—simplicity and clarity help lower-level responses score correctly, whereas comprehensiveness and information density help higher-level responses.

Load-bearing premise

The central correlation result assumes that the researchers' manual labels assigning each response to low, medium, or high reasoning complexity are accurate and consistent; the paper reports no check—no inter-rater agreement or validation—for those labels, so if another coder would often disagree with the labels, the negative and positive correlations could be an artifact.

Editorial extensions

If this is right

  • Chinese science classrooms can use a fine-tuned ChatGPT as a working automatic scorer, with holistic agreement between 80% and 94% on the seven tasks tested here.
  • Scoring reports should be element-specific rather than a single global number, since the Reasoning element showed the widest accuracy range across items (75% to 96%) while Phenomenon scoring was generally highest.
  • Low-performing students who write long, technically dense but incorrect explanations are at risk of receiving inflated scores, while high-performing students who write concise correct explanations risk deflated scores.
  • Automatic scoring claims validated on English text should not be assumed to transfer to Chinese; domain-specific fine-tuning on Chinese data is required to reach usable accuracy.
  • Fine-tuning for Chinese scientific explanations involves a design trade-off: favoring simple, clear responses helps lower-level accuracy, while favoring comprehensive, information-rich responses helps higher-level accuracy, so a single scoring policy will be biased at one end or the other.

Reading between the lines

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

  • An extension the paper leaves implicit: the same complexity-dependent bias pattern may appear for other logographic languages (e.g., Japanese, Korean) and for other LLMs, but that remains untested.
  • The qualitative linguistic findings could be mechanized: response length, clause count, technical-term density, and formula use are measurable features that could feed an error-prediction model to flag responses the scorer is likely to misgrade.
  • A practical design not explored here would be a two-stage scorer that first estimates a response's likely performance level or reasoning complexity, then applies an accuracy-optimized scoring threshold for that group; the paper's data suggest this could reduce both over- and under-scoring.
  • The paper's exploratory qualitative analysis implies a testable claim: rewriting identical scientific content in more complex sentence forms should change machine scores even when human scores do not; that experiment would directly test whether the reported linguistic features are causal or merely correlated.
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 fine-tunes ChatGPT on 7,626 Chinese middle- and high-school students' written scientific explanations across seven tasks, scores them with holistic and PTDR analytic rubrics, and reports 75–98% agreement with human scores. It then codes each response's reasoning complexity as low/medium/high, computes Kendall correlations between complexity and scoring accuracy separately for lower- and higher-performing groups, and performs a qualitative linguistic-feature analysis. The authors conclude that fine-tuned ChatGPT can score Chinese scientific explanations accurately, that low-level responses tend to be overrated when they are linguistically complex while high-level responses tend to be underrated when concise, and that these patterns reflect a tension between comprehensiveness and simplicity/clarity.

Significance. If the reported results hold, this is a useful empirical extension of LLM-based automatic scoring to a logographic language, with an unusually large corpus (N = 7,626) and seven tasks. The human scoring procedure, with a Cohen's kappa threshold above 0.8, is a strength. The paper's distinctive contributions—the opposite-signed complexity-accuracy correlations and the linguistic-feature account—are, however, conditional on reliability evidence for the reasoning-complexity coding and on the baseline comparison promised in Section 4.4.1, neither of which is currently provided. The accuracy figures are plausible, but they are point estimates without uncertainty quantification. The agreement metric is computed on held-out responses from the same seven items, which is standard supervised evaluation and not circular, but it does not by itself establish generalization to new tasks. No code or data are provided, limiting reproducibility.

major comments (4)
  1. [§4.4.3, Table 5, Tables 8–9] The second and third research questions depend entirely on the manual coding of responses into low/medium/high reasoning complexity, but the manuscript reports no inter-rater agreement, no coding training or blinding, and no validity check for this coding. Table 6 presents the resulting distribution as ground truth, and Tables 8 and 9 use those labels to compute Kendall correlations. If the complexity coding is noisy or is systematically correlated with human scores or with the linguistic features analyzed in Section 5.3, the central correlation and feature-attribution claims become uninterpretable. Please report at least a subsample of double coding with kappa/percent agreement, describe who coded and under what instructions, and state whether coders were blind to human scores and to ChatGPT outputs.
  2. [§4.4.1, §5] Section 4.4.1 explicitly promises a 'Comparison with Baseline' against the original GPT-3.5-turbo model, but no baseline results appear in Section 5 or in any table. Since the abstract and conclusions attribute the reported success to 'domain-specific adaptation,' the omission undermines the central framing. The authors should either add the untuned baseline accuracy on the same held-out splits or revise the claims so that they do not imply a comparative advantage over the base model.
  3. [§5.1, Tables 8–9] None of the accuracy or correlation results are accompanied by measures of uncertainty. The test sets in Table 2 contain only 162–330 responses each, so the 75–98% point estimates in Figure 3 and the corresponding item comparisons could easily be within sampling noise. Likewise, Tables 8 and 9 report 60 significance tests without any multiple-comparison correction or stated pre-analysis plan. Please report standard errors or confidence intervals for accuracies and correlations and address multiplicity, or explicitly label the cross-item comparisons as descriptive.
  4. [§4.4.4, §5.3, §7] The qualitative analysis is presented as explaining the correlations ('These correlations stem from linguistic features'), but it is a manual matching procedure for which no blinding or inter-rater reliability is described, and it provides no quantitative support for claims such as 'averaged only 20 Chinese characters' versus '40 Chinese characters' or 'typically ranging from 5 to 10 sentences.' The authors acknowledge in Section 7 that the analysis is exploratory, yet the abstract and conclusions state the linguistic-feature relationships as established findings. Please either reframe the conclusions as hypotheses supported by exploratory evidence or add a quantitative analysis (e.g., a regression of scoring errors on measured linguistic features) with coder agreement reported.
minor comments (6)
  1. [§4.4.3] The equation for Kendall's tau-c is typeset in a way that obscures the denominator ('τc = 2(nc − nd) n2 · m−1 m'); please present the standard formula explicitly as τc = 2(nc − nd) / [n^2 · (m − 1)/m].
  2. [§4.4.3] The illustrative result τ(97) = -0.32 is described as a 'moderate negative correlation,' whereas Table 4 categorizes coefficients of 0.10–0.39 as 'weak'; please make the interpretation consistent.
  3. [§4.3, Table 2] The text describes each scoring element as binary, but Table 2 labels the datasets as 'Multi-Label/Multi-Class' with 4 or 5 labels; please clarify how the label count relates to the binary elements and whether the model predicts multiple labels jointly.
  4. [§4.4.1] The fine-tuning description lists loss function, learning rate, epochs, batch size, and data augmentation only in qualitative terms; exact hyperparameters, the number of augmented samples, and the generation procedure should be provided for reproducibility.
  5. [§7, Abstract] The abstract and conclusions present the linguistic-feature relationships as established, contradicting the paper's own statement in Section 7 that these analyses are exploratory; please align the abstract with the stated evidentiary status.
  6. [Global] There are numerous language and typographical issues, including 'languish' for 'language' (e.g., Section 5.3 introduction and Section 6.3 heading), a placeholder '?' after 'feedback for student responses' in Section 2, and the ungrammatical phrase 'justify their claims build' in Section 2; a full editorial pass is needed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: held-out supervised evaluation and independently coded complexity analysis are not derivable from their inputs.

full rationale

The paper's central claims are empirical rather than derivational: fine-tuned ChatGPT is scored against human ratings on held-out test splits of the same seven items (Table 2, Section 5.1), and the accuracy is defined as agreement with human scoring (Section 4.4.2). This is standard supervised evaluation; the held-out split means the reported accuracies are not forced by the training labels. The reasoning-complexity analysis (Section 4.4.3, Tables 8-9) uses an independently manually coded complexity variable crossed with human-machine agreement, and the correlations could have been null or opposite in sign, so the finding is falsifiable rather than constructed. The qualitative linguistic-feature analysis in Section 5.3 is explicitly exploratory and post-hoc, not presented as a derivation from the model. The only same-author citation that appears in a load-bearing position is Latif and Zhai (2024), used as background for the fine-tuning approach and for comparison with English-language results; the Chinese-language accuracy and correlation results come from this paper's own data, not from that citation. The promised GPT-3.5 baseline comparison in Section 4.4.1 is not reported in Section 5, but an omitted comparison is a reporting gap, not circularity. No equation or fitted parameter is shown to reduce to its own input, and no self-citation chain is used to justify the central empirical claims. Therefore the paper shows no significant circularity.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claims rest on human scoring validity, reliable reasoning-complexity coding, and the qualitative attribution of accuracy differences to linguistic features. The paper supplies no inter-rater reliability for the complexity coding, no external benchmark, and no exact fine-tuning configuration. The fine-tuning hyperparameters and augmentation procedure are free parameters chosen through validation but never reported.

free parameters (2)
  • Fine-tuning hyperparameters (learning rate, epochs, batch size) = unspecified
    Chosen per item through validation, but exact values are not reported, so the central accuracy results depend on undocumented training choices.
  • Data augmentation sample size and generation method = unspecified
    Section 4.4.1 says artificially generated samples imitating complex reasoning were incorporated, but the count and generation procedure are absent, and they can change model behavior.
assumptions (4)
  • domain assumption Human expert scores under the PTDR rubric are a valid ground truth for student explanation quality.
    The entire accuracy metric is agreement with these human scores; no external validity check of the rubric is provided.
  • domain assumption The reasoning complexity categories (low, medium, high) can be applied reliably to student responses.
    Section 4.4.3 adapts criteria from prior work, but no inter-rater reliability for this coding is reported; the correlation results for RQ2 and RQ3 depend on it.
  • domain assumption Held-out test responses from the same seven items represent the scoring distribution well enough to estimate accuracy.
    Train and test splits are per item in Table 2 and appear to be random; no cross-validation or external dataset is used.
  • ad hoc to paper The linguistic features named in the qualitative analysis (simplicity, clarity, comprehensiveness) are the causal drivers of scoring accuracy.
    Section 5.3 compares subgroups qualitatively, and the paper itself labels the analysis exploratory in Section 7, yet the abstract presents the causal link as a finding.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fine-tuning ChatGPT for Automatic Scoring of Written Scientific Explanations in Chinese." pith.science (2026). https://pith.science/paper/HTLZC2RC

@misc{pith2026250106704,
  author       = {Pith},
  title        = {Pith review of: Fine-tuning ChatGPT for Automatic Scoring of Written Scientific Explanations in Chinese},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HTLZC2RC}},
  note         = {Machine review of arXiv:2501.06704}
}
read the original abstract

The development of explanations for scientific phenomena is essential in science assessment, but scoring student-written explanations remains challenging and resource-intensive. Large language models (LLMs) have shown promise in addressing this issue, particularly in alphabetic languages like English. However, their applicability to logographic languages is less explored. This study investigates the potential of fine-tuning ChatGPT, a leading LLM, to automatically score scientific explanations written in Chinese. Student responses to seven scientific explanation tasks were collected and automatically scored, with scoring accuracy examined in relation to reasoning complexity using the Kendall correlation. A qualitative analysis explored how linguistic features influenced scoring accuracy. The results show that domain-specific adaptation enables ChatGPT to score Chinese scientific explanations with accuracy. However, scoring accuracy correlates with reasoning complexity: a negative correlation for lower-level responses and a positive one for higher-level responses. The model overrates complex reasoning in low-level responses with intricate sentence structures and underrates high-level responses using concise causal reasoning. These correlations stem from linguistic features--simplicity and clarity enhance accuracy for lower-level responses, while comprehensiveness improves accuracy for higher-level ones. Simpler, shorter responses tend to score more accurately at lower levels, whereas longer, information-rich responses yield better accuracy at higher levels. These findings demonstrate the effectiveness of LLMs in automatic scoring within a Chinese context and emphasize the importance of linguistic features and reasoning complexity in fine-tuning scoring models for educational assessments.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Using Large Language Models to Assess Teachers' Pedagogical Content Knowledge

    cs.AI 2025-05 conditional novelty 5.0 of 10

    In video-based teacher knowledge assessments, GPT-4 scoring was more lenient than both human raters and a supervised ML model, while rater-related factors dominated construct-irrelevant variance.

Reference graph

Works this paper leans on

66 extracted references · 50 canonical work pages · cited by 1 Pith paper

  1. [1]

    write newline

    " write newline " cite write " FUNCTION editor.postfix editor num.names #1 > "( )" "( )" if FUNCTION editor.trans.postfix editor num.names #1 > "( )" "( )" if FUNCTION trans.postfix translator num.names #1 > "( )" "( )" if FUNCTION authors.editors.reflist.apa5 'field := 'dot := field num.names 'numnames := numnames 'format.num.names := format.num.names na...

  2. [2]

    , " * write output.state after.block = add.period write newline

    ENTRY address author booktitle chapter doi edition editor eid howpublished institution journal key keywords month note number organization pages publisher school series title type url volume year eprint archive archivePrefix primaryClass adsurl adsnote version label INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.sta...

  3. [3]

    write newline

    " write newline "" before.all 'output.state := FUNCTION if.digit duplicate "0" = swap duplicate "1" = swap duplicate "2" = swap duplicate "3" = swap duplicate "4" = swap duplicate "5" = swap duplicate "6" = swap duplicate "7" = swap duplicate "8" = swap "9" = or or or or or or or or or FUNCTION n.separate 't := "" #0 'numnames := t empty not t #-1 #1 subs...

  4. [4]

    APACrefauthors \ 2007

    abdi2007kendall APACrefauthors Abdi, H. APACrefauthors \ 2007 . The Kendall rank correlation coefficient The kendall rank correlation coefficient . Encyclopedia of Measurement and Statistics. Sage, Thousand Oaks, CA 508--510,

    work page 2007

  5. [5]

    , Kaya, M.H

    adiguzel2023revolutionizing APACrefauthors Adiguzel, T. , Kaya, M.H. Cansu, F.K. APACrefauthors \ 2023 . Revolutionizing education with AI: Exploring the transformative potential of ChatGPT Revolutionizing education with ai: Exploring the transformative potential of chatgpt . Contemporary Educational Technology 15 3 ep429,

    work page 2023

  6. [6]

    , Nazaretsky, T

    ariely2023machine APACrefauthors Ariely, M. , Nazaretsky, T. Alexandron, G. APACrefauthors \ 2023 . Machine learning and Hebrew NLP for automated assessment of open-ended questions in biology Machine learning and hebrew nlp for automated assessment of open-ended questions in biology . International journal of artificial intelligence in education 33 1 1--34,

    work page 2023

  7. [7]

    \ Johnson, D.R

    beaty2021automating APACrefauthors Beaty, R.E. \ Johnson, D.R. APACrefauthors \ 2021 . Automating creativity assessment with SemDis: An open platform for computing semantic distance Automating creativity assessment with semdis: An open platform for computing semantic distance . Behavior research methods 53 2 757--780,

    work page 2021

  8. [8]

    beggrow2014assessing APACrefauthors Beggrow, E.P. , Ha, M. , Nehm, R.H. , Pearl, D. Boone, W.J. APACrefauthors \ 2014 . Assessing scientific practices using machine-learning methods: How closely do they match clinical interview performance? Assessing scientific practices using machine-learning methods: How closely do they match clinical interview performa...

    work page 2014

Show all 66 references
  1. [9]

    , Wilhelmsen, T

    Bjerknes2024curiosity APACrefauthors Bjerknes, A L. , Wilhelmsen, T. Foyn-Bruun, E. APACrefauthors \ 2024 . A Systematic Review of Curiosity and Wonder in Natural Science and Early Childhood Education Research A systematic review of curiosity and wonder in natural science and ...

    2024

  2. [10]

    , Moreira, P

    Cabello2021elementary APACrefauthors Cabello, V. , Moreira, P. Morales, P.G. APACrefauthors \ 2021 . Elementary students’ reasoning in drawn explanations based on a scientific theory Elementary students’ reasoning in drawn explanations based on a scientific theory . Education ...

    2021

  3. [11]

    \ Zhang, Y

    chen2021influence APACrefauthors Chen, H. \ Zhang, Y. APACrefauthors \ 2021 . The Influence of Cultural Differences between Chinese and English on Translation [J] The influence of cultural differences between chinese and english on translation [j] . Cross Current Int J Peer Re...

    2021

  4. [12]

    , Freire, S

    de2019constructing APACrefauthors De Andrade, V. , Freire, S. Baptista, M. APACrefauthors \ 2019 . Constructing scientific explanations: A system of analysis for students’ explanations Constructing scientific explanations: A system of analysis for students’ explanations . Rese...

    2019

  5. [13]

    \ Bose, I

    dhar2023we APACrefauthors Dhar, S. \ Bose, I. APACrefauthors \ 2023 . Are We Nearing Singularity? A Study of Language Capabilities of ChatGPT Are we nearing singularity? a study of language capabilities of chatgpt . Analytics Global Conference Analytics global conference \ ( \...

    2023

  6. [14]

    , Akhter, M.E

    John2024GPTreasoning APACrefauthors Dougrez-Lewis, J. , Akhter, M.E. , He, Y. Liakata, M. APACrefauthors \ 2024 . Assessing the Reasoning Abilities of ChatGPT in the Context of Claim Verification Assessing the reasoning abilities of chatgpt in the context of claim verification...

    2024 arXiv

  7. [15]

    , Newton, P

    driver2000establishing APACrefauthors Driver, R. , Newton, P. Osborne, J. APACrefauthors \ 2000 . Establishing the norms of scientific argumentation in classrooms Establishing the norms of scientific argumentation in classrooms . Science education 84 3 287--312,

    2000

  8. [16]

    , Jiang, S

    fiacco-etal-2022-toward APACrefauthors Fiacco, J. , Jiang, S. , Adamson, D. Ros \'e , C. APACrefauthors \ 2022 07 . Toward Automatic Discourse Parsing of Student Writing Motivated by Neural Interpretation Toward automatic discourse parsing of student writing motivated by neura...

    2022

  9. [17]

    , Di Mitri, D

    gombert2023coding APACrefauthors Gombert, S. , Di Mitri, D. , Karademir, O. , Kubsch, M. , Kolbe, H. , Tautz, S. Drachsler, H. APACrefauthors \ 2023 . Coding energy knowledge in constructed responses with explainable NLP models Coding energy knowledge in constructed responses ...

    2023

  10. [18]

    , H Nehm, R

    Ha2011computerized APACrefauthors Ha, M. , H Nehm, R. , Urban-Lurain, M. E., M.J. APACrefauthors \ 2011 . Applying computerized-scoring models of written biological explanations across courses and colleges: prospects and limitations Applying computerized-scoring models of writ...

    2011

  11. [19]

    \ Zhai, X

    haudek2023examining APACrefauthors Haudek, K.C. \ Zhai, X. APACrefauthors \ 2023 . Examining the Effect of Assessment Construct Characteristics on Machine Learning Scoring of Scientific Argumentation Examining the effect of assessment construct characteristics on machine learn...

    2023

  12. [20]

    , Chuhao, W

    zhang2023assessing APACrefauthors He, Z. , Chuhao, W. , Jingyi, X. , Lyu, Y. , Jie, C. Carroll, J.M. APACrefauthors \ 2023 . Redefining Qualitative Analysis in the AI Era: Utilizing ChatGPT for Efficient Thematic Analysis Redefining qualitative analysis in the ai era: Utilizin...

    2023 arXiv

  13. [21]

    , Vorholzer, A

    krell2022scientific APACrefauthors Krell, M. , Vorholzer, A. Nehring, A. APACrefauthors \ 2022 . Scientific Reasoning in Science Education: From Global Measures to Fine-Grained Descriptions of Students’ Competencies. Educ. Sci. 2022, 12, 97 Scientific reasoning in science educ...

    2022

  14. [22]

    , Czinczel, B

    kubsch2022toward APACrefauthors Kubsch, M. , Czinczel, B. , Lossjew, J. , Wyrwich, T. , Bednorz, D. , Bernholt, S. others APACrefauthors \ 2022 . Toward learning progression analytics—Developing learning environments for the automated analysis of learning using evidence center...

    2022

  15. [23]

    \ Boulanger, D

    kumar2021automated APACrefauthors Kumar, V.S. \ Boulanger, D. APACrefauthors \ 2021 . Automated essay scoring and the deep learning black box: How are rubric scores determined? Automated essay scoring and the deep learning black box: How are rubric scores determined? Internati...

    2021

  16. [24]

    \ Lawson, A.E

    kwon2000linking APACrefauthors Kwon, Y J. \ Lawson, A.E. APACrefauthors \ 2000 . Linking brain growth with the development of scientific reasoning ability and conceptual change during adolescence Linking brain growth with the development of scientific reasoning ability and con...

    2000

  17. [25]

    , Hand, B

    lamb2021computational APACrefauthors Lamb, R. , Hand, B. Kavner, A. APACrefauthors \ 2021 . Computational modeling of the effects of the science writing heuristic on student critical thinking in science using machine learning Computational modeling of the effects of the scienc...

    2021

  18. [26]

    \ Zhai, X

    latif2024fine APACrefauthors Latif, E. \ Zhai, X. APACrefauthors \ 2024 . Fine-tuning chatgpt for automatic scoring Fine-tuning chatgpt for automatic scoring . Computers and Education: Artificial Intelligence 100210,

    2024

  19. [27]

    , Gobert, J

    Liqutomated APACrefauthors Li, H. , Gobert, J. Dickler, R. APACrefauthors \ 2017 01 . Automated Assessment for Scientific Explanations in On-line Science Inquiry Automated assessment for scientific explanations in on-line science inquiry

    2017

  20. [28]

    , Meng, Y

    li2019word APACrefauthors Li, X. , Meng, Y. , Sun, X. , Han, Q. , Yuan, A. Li, J. APACrefauthors \ 2019 . Is word segmentation necessary for deep learning of Chinese representations? Is word segmentation necessary for deep learning of chinese representations? arXiv preprint ar...

    2019 arXiv

  21. [29]

    Pai, K C

    liao2012effectivenesss APACrefauthors Liao , Chen-Huei , Kuo, B C. Pai, K C. APACrefauthors \ 2012 . Effectiveness of Automated Chinese Sentence Scoring with Latent Semantic Analysis Effectiveness of automated chinese sentence scoring with latent semantic analysis . Turkish On...

    2012

  22. [30]

    , Jakwatanatham, S

    limna2022review APACrefauthors Limna, P. , Jakwatanatham, S. , Siripipattanakul, S. , Kaewpuang, P. Sriboonruang, P. APACrefauthors \ 2022 . A review of artificial intelligence (AI) in education during the digital era A review of artificial intelligence (ai) in education durin...

    2022

  23. [31]

    liu2023context APACrefauthors Liu, Z. , He, X. , Liu, L. , Liu, T. Zhai, X. APACrefauthors \ 2023 . Context matters: A strategy to pre-train language model for science education Context matters: A strategy to pre-train language model for science education . arXiv preprint arXi...

    2023 arXiv

  24. [32]

    APACrefauthors \ 2015

    lockwood2015handbook APACrefauthors Lockwood, J. APACrefauthors \ 2015 . Handbook of automated essay evaluation: Current applications and new directions Handbook of automated essay evaluation: Current applications and new directions . Writing & Pedagogy 6 437--441,

    2015

  25. [33]

    \ Krajcik, J

    mcneill2008inquiry APACrefauthors McNeill, K.L. \ Krajcik, J. APACrefauthors \ 2008 . Inquiry and scientific explanations: Helping students use evidence and reasoning Inquiry and scientific explanations: Helping students use evidence and reasoning . Science as inquiry in the s...

    2008

  26. [34]

    , Yan, D

    mislevy2020automated APACrefauthors Mislevy, R.J. , Yan, D. , Gobert, J. Sao Pedro, M. APACrefauthors \ 2020 . Automated scoring in intelligent tutoring systems Automated scoring in intelligent tutoring systems . Handbook of automated scoring Handbook of automated scoring \ ( ...

    2020

  27. [35]

    Moharreri2014EvoGrader APACrefauthors Moharreri, K. , Ha, M. Nehm, R.H. APACrefauthors \ 2014 . EvoGrader: an online formative assessment tool for automatically evaluating written evolutionary explanations Evograder: an online formative assessment tool for automatically evalua...

    2014

  28. [36]

    Nehm2012Transforming APACrefauthors Nehm, R. , Ha, M. Mayfield, E. APACrefauthors \ 2012 . Transforming biology assessment with machine learning: automated scoring of written evolutionary explanations Transforming biology assessment with machine learning: automated scoring of ...

    2012

  29. [37]

    \ Retelsdorf, J

    neri2022role APACrefauthors Neri, N.C. \ Retelsdorf, J. APACrefauthors \ 2022 . The role of linguistic features in science and math comprehension and performance: A systematic review and desiderata for future research The role of linguistic features in science and math compreh...

    2022

  30. [38]

    PISA2018 Well-being Framework Pisa2018 well-being framework \

    OECD2019 APACrefauthors OECD APACrefauthors \ 2019 . PISA2018 Well-being Framework Pisa2018 well-being framework \ . OECD

    2019

  31. [39]

    , Lottridge, S

    ormerod2023automated APACrefauthors Ormerod, C. , Lottridge, S. , Harris, A.E. , Patel, M. , van Wamelen, P. , Kodeswaran, B. Young, M. APACrefauthors \ 2023 . Automated short answer scoring using an ensemble of neural networks and latent semantic analysis classifiers Automate...

    2023

  32. [40]

    , Berland, L.K

    reiser2012engaging APACrefauthors Reiser, B.J. , Berland, L.K. Kenyon, L. APACrefauthors \ 2012 . Engaging students in the scientific practices of explanation and argumentation Engaging students in the scientific practices of explanation and argumentation . The Science Teacher...

    2012

  33. [41]

    , Bichler, S

    riordan2020empirical APACrefauthors Riordan, B. , Bichler, S. , Bradford, A. , Chen, J.K. , Wiley, K. , Gerard, L. Linn, M.C. APACrefauthors \ 2020 . An empirical investigation of neural methods for content scoring of science explanations An empirical investigation of neural m...

    2020

  34. [42]

    , Boer, C

    schober2018correlation APACrefauthors Schober, P. , Boer, C. Schwarte, L.A. APACrefauthors \ 2018 . Correlation coefficients: appropriate use and interpretation Correlation coefficients: appropriate use and interpretation . Anesthesia & analgesia 126 5 1763--1768,

    2018

  35. [43]

    \ TIAN, Z x

    sun2017cultural APACrefauthors SUN, M. \ TIAN, Z x. APACrefauthors \ 2017 . The cultural differences between English and Chinese courtesy languages The cultural differences between english and chinese courtesy languages . Journal of Literature and Art Studies 7 3 340--344,

    2017

  36. [44]

    APACrefauthors \ 2012

    Vosniadou2019understanding APACrefauthors Vosniadou, S. APACrefauthors \ 2012 . The development of students' understanding of science The development of students' understanding of science . In Frontiers in Education 4 32,

    2012

  37. [45]

    , Liu, X

    Wang2021automated APACrefauthors Wang, C. , Liu, X. , Wang, L. , Sun, Y. zhang, H. APACrefauthors \ 2021 . Automated scoring of Chinese grades 7–9 students’ competence in interpreting and arguing from evidence Automated scoring of chinese grades 7–9 students’ competence in int...

    2021

  38. [46]

    \ Chen, J

    wang2013differences APACrefauthors Wang, Y. \ Chen, J. APACrefauthors \ 2013 . Differences of English and Chinese as Written Languages and Strategies in English Writing Teaching. Differences of english and chinese as written languages and strategies in english writing teaching...

    2013

  39. [47]

    \ Bever, T

    Williamsf2010Chinese APACrefauthors Williams, C. \ Bever, T. APACrefauthors \ 2010 . Chinese character decoding: a semantic bias? Chinese character decoding: a semantic bias? Reading and Writing 589-605,

    2010

  40. [48]

    Wu2023GPTOverview APACrefauthors Wu, T. , He, S. , Liu, J. , Sun, S. , Liu, K. , Han, Q L. Tang, Y. APACrefauthors \ 2023 . A Brief Overview of ChatGPT: The History, Status Quo and Potential Future Development A brief overview of chatgpt: The history, status quo and potential ...

    2023

  41. [49]

    , Zhang, Y

    yang2018subword APACrefauthors Yang, J. , Zhang, Y. Liang, S. APACrefauthors \ 2018 . Subword encoding in lattice LSTM for Chinese word segmentation Subword encoding in lattice lstm for chinese word segmentation . arXiv preprint arXiv:1810.12594 ,

    2018 arXiv

  42. [50]

    \ Guo, Y

    Yao2018explanation APACrefauthors Yao, J. \ Guo, Y. APACrefauthors \ 2018 . Validity evidence for a learning progression of scientific explanation Validity evidence for a learning progression of scientific explanation . Journal of Research in Science Teaching 55 2 299-317,

    2018

  43. [51]

    APACrefauthors \ 2022

    zhai2022assessing APACrefauthors Zhai, X. APACrefauthors \ 2022 . Assessing high-school students' modeling performance on Newtonian mechanics Assessing high-school students' modeling performance on newtonian mechanics . Journal of Research in Science Teaching 59 8 1313--1353,

    2022

  44. [52]

    APACrefauthors \ 2023

    zhai2023chatgpt APACrefauthors Zhai, X. APACrefauthors \ 2023 . Chatgpt for next generation science learning Chatgpt for next generation science learning . XRDS: Crossroads, The ACM Magazine for Students 29 3 42--46,

    2023

  45. [53]

    APACrefauthors \ 2024

    zhai2024ai APACrefauthors Zhai, X. APACrefauthors \ 2024 . AI and Machine Learning for Next Generation Sci-ence Assessments Ai and machine learning for next generation sci-ence assessments . Machine Learning, Natural Language Processing, and Psychometrics 201,

    2024

  46. [54]

    \ Nehm, R.H

    zhai2023ai APACrefauthors Zhai, X. \ Nehm, R.H. APACrefauthors \ 2023 . AI and formative assessment: The train has left the station Ai and formative assessment: The train has left the station . Journal of Research in Science Teaching ,

    2023

  47. [55]

    \ Pellegrino, J.W

    zhai2023large APACrefauthors Zhai, X. \ Pellegrino, J.W. APACrefauthors \ 2023 . Large-scale assessment in science education Large-scale assessment in science education . Handbook of research on science education Handbook of research on science education \ ( \ 1045--1097). Routledge

    2023

  48. [56]

    , Shi, L

    zhai2021meta APACrefauthors Zhai, X. , Shi, L. Nehm, R.H. APACrefauthors \ 2021 . A meta-analysis of machine learning-based science assessments: Factors impacting machine-human score agreements A meta-analysis of machine learning-based science assessments: Factors impacting ma...

    2021

  49. [57]

    , Yin, Y

    zhai2020applying APACrefauthors Zhai, X. , Yin, Y. , Pellegrino, J.W. , Haudek, K.C. Shi, L. APACrefauthors \ 2020 . Applying machine learning in science assessment: a systematic review Applying machine learning in science assessment: a systematic review . Studies in Science E...

    2020

  50. [58]

    , " * write output.state after.block = add.period write newline

    ENTRY address archive author booktitle chapter doi edition editor eid eprint howpublished institution journal key keywords month note number organization pages publisher school series title type url volume year archivePrefix primaryClass adsurl adsnote version label extra.labe...

  51. [59]

    write newline

    " write newline "" before.all 'output.state := FUNCTION add.period duplicate empty 'skip "." * add.blank if FUNCTION if.digit duplicate "0" = swap duplicate "1" = swap duplicate "2" = swap duplicate "3" = swap duplicate "4" = swap duplicate "5" = swap duplicate "6" = swap dupl...

  52. [60]

    write newline

    " write newline "" before.all 'output.state := FUNCTION output.doi doi empty skip "doi:" doi * "" * output if FUNCTION format.archive archivePrefix empty "" archivePrefix ":" * if FUNCTION format.primaryClass primaryClass empty "" " [" primaryClass * "] " * if FUNCTION format....

  53. [61]

    write newline

    " write newline "" before.all 'output.state := FUNCTION string.to.integer 't := t text.length 'k := #1 'char.num := t char.num #1 substring 's := s is.num s "." = or char.num k = not and char.num #1 + 'char.num := while char.num #1 - 'char.num := t #1 char.num substring FUNCTI...

  54. [62]

    , " * write output.state after.block = add.period write newline

    ENTRY address archive author booktitle chapter edition editor eprint howpublished institution journal key keywords month note number organization pages publisher school series title type url doi volume year archivePrefix primaryClass eid adsurl adsnote version label INTEGERS o...

  55. [63]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

  56. [64]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

  57. [65]

    Available from:

    ENTRY address assignee author booktitle chapter cartographer day edition editor howpublished institution inventor journal key keywords month note number organization pages part publisher school series title type volume word year eprint doi url lastchecked updated archive archi...

  58. [66]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.