The DOI matches its Crossref record; 93 of the 93 references it cites were found in a registry or an index; 4 of the 66 sources its claims cite were checked for support.
A Feasibility and Implementation Integrity Study of the Community Builder (CoBi): An AI-based Collaboration Support System in K-12 Classrooms
Open accessAbstract
Recently, RSTs have been developed to provide feedback to students on their collaboration skills. We present the feasibility and implementation integrity study of Community Builder (CoBi), an AI-based collaboration support system in K-12 classrooms. CoBi provides automated feedback to students and teachers during collaborative problem solving.
Overview
CoChat Summary
TL;DR
CoBi is a classroom-wide, speech-based AI system that detects "uplifting" small-group discourse and displays non-evaluative, privacy-preserving visualizations at the class level to support K-12 students' collaboration skills. Across an interview study and two classroom pilots (61 consented students total), students and teachers broadly valued the concept and CoBi was technically feasible, but high-integrity use required substantial teacher professional learning.
Key findings
- Study 1 (3 teachers, 12 students): broad buy-in for the idea; 83.3% of students preferred the qualitative tree over the quantitative radar; 75% of students were comfortable with teachers/researchers seeing their data, but some raised privacy concerns (one said CoBi must not be a "snitch"); all teachers worried real-time displays could disrupt teaching.
- Study 2 (CoBi v1, 3 classrooms, 24 consented students): technically feasible — interfaces loaded, visualizations updated every 10 s, and reached end states after ~15 min; establishing and enacting phases succeeded, but the revisiting phase was cut short by time constraints, and the time slider in both teacher and student views caused confusion about whether visualizations were group- or class-level.
- Study 3 (CoBi v2, 3 classrooms, 37 consented students, 20 days of instruction, ~200 min of video): teachers showed diverse adaptations and high-integrity engagements; some used noticings to prompt meta-analytical debate about classification accuracy, while two teachers steered discussions toward AI literacy and some used visualizations for participation assessment rather than reflection.
- Model performance (Appendix): three fine-tuned RoBERTa classifiers achieved AUROCs of 0.84, 0.77, and 0.82 on human transcripts (0.71, 0.67, 0.71 on Whisper transcripts) for three of the four CA categories; "being equitable" occurred below 1% in training data and was not modeled.
Methods
- Design-based research with three studies conducted in 2023/2024: Study 1 interview study (thematic analysis of teacher transcripts and student notes); Study 2 technical feasibility/initial implementation with CoBi v1 (field notes, classroom video, analysis of student-inputted CAs); Study 3 implementation integrity with CoBi v2 (grounded theory on field notes for 20 days and ~200 min of teacher focal video).
- Participants: Study 1 — 3 middle school STEM/science teachers and 12 seventh graders from two schools; Study 2 — 1 lead teacher (plus a co-teacher for a summer camp) across 3 classrooms, 24 fully consented students; Study 3 — 3 teachers and 37 fully consented students across 3 classrooms (97 including non-consenting), same Western US district.
- CoBi pipeline: omnidirectional tabletop microphones stream 10-s audio chunks to cloud servers, Whisper ASR transcribes, RoBERTa-based discourse classifiers detect CA enactments, and results are aggregated to classroom-level tree, radar, sentence cloud, and rule-based noticings visualizations.
Limitations
- Stated: the study was not designed to test efficacy (future randomized control trials planned); CoBi is unimodal (verbal only, no nonverbal cues such as gaze); testing occurred in a limited number of schools and models were trained on data from one curriculum unit; only three of four CAs are modeled, and the fourth (equitable contribution) is being addressed via work on communicative influence.
- Evident from the text: consented samples were small (24 and 37 students) and drawn from a single district with limited demographic diversity (district-wide: 20% Hispanic or Latino, 73% White, 1% Black, 3% Asian, 3% Two or more races); Study 1 student interviews were very short (2–5 min), unrecorded, and captured only as notes; Study 3 collected only teacher focal video, not student video.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
4 Valid · 5 Not valid · 31 Not checked · 129 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“The computational models for detecting evidence of CAs in students' collaborative discourse consist of three fine-tuned utterance-level RoBERTa transformer models (Liu et al., 2019), which outperformed few-shot prompting of LLMs (Ganesh et al., 2024).”
What the sources say
- Supportedcites Liu et al., 2019 source
The citation to Liu et al. identifies the RoBERTa model used by the citing authors. Section 5 (p. 6) introduces RoBERTa, and the paper describes fine-tuning it for downstream tasks. The comparison with few-shot prompting is cited separately to Ganesh et al. (2024), not attributed to Liu et al. (LLM, conf=0.91)
- Not matchedcites Ganesh et al., 2024
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The interviews revealed a strong preference for the tree over the radar from students (83.3%).”
boosted- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In addition, teachers regarded CoBi as a meaningful incentive for students to practice collaboration skills.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“When asked about having their conversations recorded, most students (75%) were content with letting teachers and researchers see their data.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Two students from separate pairs favored the more quantitative radar.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
69 more claims
Create a free account to see all 169 claims and to verify the 4 references still open.
- Probably not a substantial claim. 63
- It may state two separate claims as one. 53
- The quote only partly states this claim. 2
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ Design Principles and Recommendations · p. 25
“they were not comfortable sharing their data with the teacher even though this is what teachers preferred.”
§ Results · p. 13
“When asked about having their conversations recorded, most students (75%) were content with letting teachers and researchers see their data.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Appendix A: Technical Details · p. 28
“Data Availability No datasets were generated or analysed during the current study.”
§ Methods · p. 20
“The dataset included notes for 20 days of instruction and around 200 min of CoBi-related video.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Study 1 Review & Synthesis · p. 14
“Both groups expressed a preference for the tree as a metaphor for organic classroom dialogue.”
§ Results · p. 13
“Teachers' responses were more mixed. One favored the tree while another favored the radar; the third teacher saw benefits in both.”
Citations
Cites
RoBERTa: A Robustly Optimized BERT Pretraining Approach
Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., Levy, O., Lewis, M., Zettlemoyer, L., & Stoyanov, V. (2019) · arXiv · doi:10.48550/arxiv.1907.11692 · Cited by 16,515
Grounded theory research: Procedures, canons, and evaluative criteria
Corbin, J. M., & Strauss, A. (1990) · Qualitative Sociology · doi:10.1007/bf00988593 · Cited by 11,133
Barredo Arrieta, A., Díaz-Rodríguez, N., Del Ser, J., Bennetot, A., Tabik, S., Barbado, A., Garcia, S., Gil-Lopez, S., Molina, D., Benjamins, R., et al. (2020) · Information Fusion · doi:10.1016/j.inffus.2019.12.012 · Cited by 10,423 · cited 1×
“…educational technologies, which highlights that offering pathways for the critical reflection about a system’s inner workings can alleviate their concerns and contribute to building the appropriate level of trust (Arrieta et al., 2020; Hyesun Choung and Ross 2023; Ter-zopoulos & Satratzemi, 2020).”
Gesture Visible Action as Utterance
Kendon, A. (2004) · Cambridge University Press eBooks · doi:10.1017/cbo9780511807572 · Cited by 3,891
What is AI Literacy? Competencies and Design Considerations
Long, D., & Magerko, B. (2020) · Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3313831.3376727 · Cited by 3,241 · cited 1×
“High-quality professional teacher learning can help increase teachers’ AI literacy (Long & Magerko, 2020) and visualization literacy (Firat et al., 2022) to better integrate AI visualizations into their pedagogy.”
The Construction of Shared Knowledge in Collaborative Problem Solving.
Roschelle, J., & Teasley, S. D. (1995) · Computer Supported Collaborative Learning · doi:10.1007/978-3-642-85098-1_5 · Cited by 2,274 · cited 3×
“…coordinated talk (Barron, 2000), co-construction (Azmitia & Montgomery, 1993; Forman & Kraker, 1985; Hatano, 1993; Hogan et al., 1999; Schwartz, 1995), co-regulation (Volet et al., 2009) and shared regulation (Iiskala et al., 2011; Roschelle & Teasley, 1995; Vau-ras et al., 2003), among others.”
Grounded Theory Research: Procedures, Canons, and Evaluative Criteria
Corbin, J., & Strauss, A. (1990) · Zeitschrift fur Soziologie · doi:10.1515/zfsoz-1990-0602 · Cited by 1,710
“Benefits of Collaborative Learning.”
Laal, M., & Ghodsi, S. M. (2012) · Procedia - Social and Behavioral Sciences · doi:10.1016/j.sbspro.2011.12.091 · Cited by 1,107 · cited 1×
“In theory, classrooms that incorporate aspects of collaborative learning (Laal & Ghodsi, 2012; Smith & MacGregor, 1992; Vygotsky & Cole, 1978) should help students develop collaboration skills, for example, by helping them actively navigate social dynamics in collaborative contexts.”
Trust in AI and Its Role in the Acceptance of AI Technologies
Choung, H., David, P., & Ross, A. (2023) · International Journal of Human–Computer Interaction · doi:10.1080/10447318.2022.2050543 · Cited by 963
Behavioral Development and Construct Validity: The Principle of Aggregation
Rushton, J. P., Brainerd, C. J., & Pressley, M. (1983) · Psychological Bulletin · doi:10.1037/0033-2909.94.1.18 · Cited by 940 · cited 1×
“This was deemed adequate for the present purpose because the predictions were aggregated from the utterance to the classroom level, which increases reliability due to the principle of aggregation (see Rushton et al., 1983).”
Research works printed but not linked: 12
- Adamson, 2014, “Intensification of Group Knowledge Exchange with Academically Productive Talk Agents.”, Proceedings of the 10th International Conference on Computer Supported Collaborative Learning., 10
- Biddy, 2021, A Professional Development Model to Integrate Computational Thinking Into, Contemporary Issues in Technology and Teacher Education, 21, 53
- Bush, 2024, Curriculum routines to support collaborative AI partner deployment in classrooms, The workshop on child-centered AI design (CCAI) at CHI 2024.
- Cooperrider, 1987, A Contemporary Commentary on Appreciative Inquiry in Organizational Life, Adv Appreciative Inq, 1, 129
- Dede, 2010, “Comparing Frameworks for 21st Century Skills.”, 21st century skills: Rethinking how students learn.
- Hyesun, 2023, “Trust in AI and Its Role in the Acceptance of AI Technologies.”, International Journal of Human–Computer Interaction, 39, 1727, 10.1080/10447318.2022.2050543 · DOI · Paper page
- 2017, “PISA 2015 Collaborative Problem Solving Framework.”, PISA 2015 Assessment and Analytical Framework: Science, Reading, Mathematic, Financial Literacy and Collaborative Problem Solving, 131
- Radford, Alec, Jong Wook Kim, Tao Xu, Greg Brockman, Christine Mcleavey, and Ilya Sutskever. 2023. “Robust Speech Recognition via Large-Scale Weak Supervision.” Pp. 28492–518 in Proceedings of the 40th International Conference on Machine Learning. vol. 202, Proceedings of Machine Learning Research, edited by A. Krause, E. Brunskill, K. Cho, B. Engelhardt, S. Sabato, and J. Scarlett. PMLR.
- Resnick, 2010, “How (Well Structured) Talk Builds the Mind.”, Innovations in Educational Psychology: Perspectives on Learning, Teaching and Human Development, 163
- Richardson, Daniel C., and Rick Dale. 2006. “UC Merced Proceedings of the Annual Meeting of the Cognitive Science Society Title Grounding Dialogue: Eye Movements Reveal the Coordination of Attention During Conversation and the Effects of Common Ground Grounding Dialogue: Eye Movements Reveal the Coordination of Attention During Conversation and the Effects of Common Ground.” (28):28.
- Southwell, 2022, Challenges and feasibility of automatic speech recognition for modeling student collaborative discourse in classrooms, Proceedings of the 15th international conference on educational data mining (EDM 2022).
- Tegos, 2017, “Conversational Agents Improve Peer Learning through Building on Prior Knowledge.”, Journal of Educational Technology & Society, 20, 99
Cited by
Penuel, W. R., Bakker, A., & Akkerman, S. F. (2025) · Mathematical Thinking and Learning · doi:10.1080/10986065.2025.2595715 · Cited by 2
Self-citationA KNN-Based Bilingual Book Recommendation System with Gamification and Learning Analytics
Kassenkhan, A. M. (2026) · Information · doi:10.3390/info17020120 · Cited by 1
Ko, M. M., Chandler, C., Rose, S., Cline, B., Watts, E., Reitman, J. G., Foltz, P. W., & D’Mello, S. K. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-29760-0_59
Self-citationResponsible AI in Education: Multilingual, Multimodal, and Relational Futures
Lee, H. H. (2026) · Teachers College Record The Voice of Scholarship in Education · doi:10.1177/01614681261493389
Sawaya, S., Brown, C. S., McKoy, A., Breideband, T., & D’Mello, S. K. (2026) · International Conference on Affective Computing and Intelligent Interaction Workshops and Demos (ACIIW) · doi:10.1109/aciiw71585.2026.11712119
“We selected the tree representation from [36] because it balanced three priorities: tolerance to ASR errors, protection of student privacy, and asset-based framing.”
Self-citation
Metadata
- Authors
- Breideband, T., Bush, J. B., Reitman, J. G., Rose, S., Weatherley, J., Penuel, W. R., & D’Mello, S. K.
- Affiliations
- University of Colorado Boulder (US)
- Corresponding author
- Breideband, Thomas
- ORCID
- Venue
- International Journal of Artificial Intelligence in Education
- Publisher
- Elsevier BV
- Published
- 2025-12
- Volume
- 35
- Issue
- 6
- Pages
- 3579-3613
- ISSN
- 1560-4292 (ISSN-L)
- OpenAlex
- W4416141883
- Cited by
- 5
- References
- 85
- Citations by year
- 2026: 4 · 2025: 1
- FWCI
- 2.02
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- www.elsevier.com/tdm/userlicense/1.0/ (tdm)
- www.elsevier.com/legal/tdmrep-license (tdm)
- creativecommons.org/licenses/by-nc-nd/4.0/ (vor, 142 d)
- Locations
- International Journal of Artificial Intelligence in Education (publishedVersion, cc-by)
- Funders
- National Science Foundation (2019805)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Intelligent Tutoring Systems and Adaptive Learning
- Social Sciences › Education › Artificial Intelligence in Education
- Social Sciences › Psychology › Developmental and Educational Psychology › Innovative Teaching and Learning Methods
- Keywords
- classroom collaboration, speech recognition, junior high school students, teacher learning, implementation fidelity, K-12 education
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: A Feasibility and Implementation Integrity Study of the Community Builder (CoBi): An AI-based Collaboration Support System in K-12 Classrooms. https://app.cochat.ai/papers/10.1007/s40593-025-00509-2
- The paper
- Breideband, T., Bush, J. B., Reitman, J. G., Rose, S., Weatherley, J., Penuel, W. R., & D’Mello, S. K. (2025). A Feasibility and Implementation Integrity Study of the Community Builder (CoBi): An AI-based Collaboration Support System in K-12 Classrooms. International Journal of Artificial Intelligence in Education. https://doi.org/10.1007/s40593-025-00509-2
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.