Designing for human–AI complementarity in K‐12 education
Open accessAbstract
Recent work has explored how complementary strengths of humans and artificial intelligence (AI) systems might be productively combined. However, successful forms of human–AI partnership have rarely been demonstrated in real‐world settings. We present the iterative design and evaluation of Lumilo, smart glasses that help teachers help their students in AI‐supported classrooms by presenting real‐time analytics about students’ learning, metacognition, and behavior. Results from a field study conducted in K‐12 classrooms indicate that students learn more when teachers and AI tutors work together during class. We discuss implications of this research for the design of human–AI partnerships. We argue for more participatory approaches to research and design in this area, in which practitioners and other stakeholders are deeply, meaningfully involved throughout the process. Furthermore, we advocate for theory‐building and for principled approaches to the study of human–AI decision‐making in real‐world contexts.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
AI tutors in CA form can outperform human tutors.
“Thus, many of them have been recognized both in academia or practice [19, 23] for proving that AI tutors in CA form can outperform human tutors through meta-analysis [28].”
From The Promise and Peril of ChatGPT in Higher Education: Opportunities, Challenges, and Design Implications (2024) · cites this paper as [28]
The CASE STUDY Background reports that meta-analyses found AI tutors can help students learn more effectively than other forms of instruction. It does not specify conversational-agent (CA) tutors or show that they outperform human tutors. This paper reports the meta-analytic finding; it does not conduct a meta-analysis. (LLM, conf=0.85)
Experimental — accuracy not yet measuredIn another field study of an AI-enhanced classroom, the authors observed a human teacher approach a student who struggled learning with the AI tutor and found out that the student's struggle did not result from interaction with AI but a break up with his significant other.
“in another fi eld study of an AI-enhanced classroom, the authors observed a human teacher approach a student who struggled learning with the AI tutor and found out that the student's struggle did not result from interaction with AI but a break up with his significant other [28].”
From The Promise and Peril of ChatGPT in Higher Education: Opportunities, Challenges, and Design Implications (2024) · cites this paper as [28]
Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measuredHolstein et al. discovered that the least preferred feature of AI tutors among both teachers and students was providing automated emotional support to frustrated students.
“Holstein et al. [28, 30], which discovered the least preferred feature of AI tutors among both teachers and students was providing automated emotional support to frustrated students.”
From The Promise and Peril of ChatGPT in Higher Education: Opportunities, Challenges, and Design Implications (2024) · cites this paper as [28]
The Case Study Background describes exploring teachers’ information needs, and the Iterative Piloting section describes teachers offering encouragement that the AI tutor could not provide. Neither reports a preference ranking from both teachers and students or identifies automated emotional support for frustrated students as the least preferred feature (searched: least preferred, frustration, emotional support, and teacher and student preferences). (LLM, conf=0.72)
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
The future of employment: How susceptible are jobs to computerisation?
Frey, C. B., & Osborne, M. A. (2017) · Technological Forecasting and Social Change · doi:10.1016/j.techfore.2016.08.019 · Cited by 8,727 · see citing works
Manipulating and Measuring Model Interpretability
Poursabzi-Sangdeh, F., Goldstein, D. G., Hofman, J. M., Wortman Vaughan, J. W., & Wallach, H. (2021) · Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3411764.3445315 · Cited by 728 · see citing works
Re-examining Whether, Why, and How Human-AI Interaction Is Uniquely Difficult to Design
Yang, Q., Steinfeld, A., Rosé, C., & Zimmerman, J. (2020) · Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3313831.3376301 · Cited by 622 · cited 3× · see citing works
“However, further research is urgently needed to develop new technical and design methods that can meaningfully engage non-technical stakeholders in understanding and working with AI as a design material, for example, see Holstein et al. (2018), Yang et al., 2020.”
BackgroundYang, Q., Steinfeld, A., & Zimmerman, J. (2019) · Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3290605.3300468 · Cited by 247 · see citing works
A Case for Humans-in-the-Loop: Decisions in the Presence of Erroneous Algorithmic Scores
De-Arteaga, M., Fogliato, R., & Chouldechova, A. (2020) · Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3313831.3376638 · Cited by 221 · see citing works
Artificial Intelligence as an Effective Classroom Assistant
du Boulay, B. (2016) · IEEE Intelligent Systems · doi:10.1109/mis.2016.93 · Cited by 213 · see citing works
Holstein, K., McLaren, B. M., & Aleven, V. (2019) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-23204-7_14 · Cited by 189 · see citing works
Self-citationStreet-Level Algorithms: A Theory at the Gaps Between Policy and Decisions
Alkhatib, A., & Bernstein, M. (2019) · Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3290605.3300760 · Cited by 188 · see citing works
A Conceptual Framework for Human–AI Hybrid Adaptivity in Education
Holstein, K., Aleven, V., & Rummel, N. (2020) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-52237-7_20 · Cited by 168 · see citing works
Self-citationThe TA Framework: Designing Real-time Teaching Augmentation for K-12 Classrooms
An, P., Holstein, K., d'Anjou, B., Eggen, B., & Bakker, S. (2020) · Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3313831.3376277 · Cited by 69 · cited 8× · see citing works
“…to be very important to teachers, who would often use the examples to second guess the system’s judgments and try to infer deeper underlying causes of student difficulties—an example of one form of complementarity between the AI and the teacher (An et al. 2020; Holstein, McLaren, and Aleven 2019a).”
BackgroundMethodInfluentialSelf-citation
Flagged first, then by citations
Research works printed but not linked: 24
- Buçinca, Z., P. Lin, K. Z. Gajos, and E. L. Glassman. 2020. "Proxy Tasks and Subjective Measures Can Be Misleading in Evaluat- ing Explainable AI Systems." In Proceedings of the ACM IUI Conference, 454-64.
- Green, B., and Y. Chen. 2019. "Disparate Interactions: An Algorithm- in-the-loop Analysis of Fairness in Risk Assessments." In Proceed- ings of the ACM FAccT Conference, 90-9.
- Holstein, K. 2019. "Designing Real-time Teacher Augmentation to Combine Strengths of Human and AI Instruction," Thesis, Carnegie Mellon University, Pittsburgh, PA.
- Holstein, K., E. Harpstead, R. Gulotta, and J. Forlizzi. 2020. "Replay Enactments: Exploring Possible Futures through Historical Data." In Proceedings of the ACM DIS Conference.
- Holstein, K., G. Hong, M. Tegene, B. M. McLaren, and V. Aleven. 2018. "The Classroom As a Dashboard: Co-designing Wearable Cogni- tive Augmentation for K-12 Teachers." In Proceedings of the LAK Conference, 79-88.
- Holstein, K., B. M. McLaren, and V. Aleven. 2017a. "SPACLE: Investigating Learning across Virtual and Physical Spaces Using Spatial Replays." In Proceedings of the LAK Conference, 358-67.
- Holstein, K., B. M. McLaren, and V. Aleven. 2017b. "Intelligent Tutors as Teachers' Aides: Exploring Teacher Needs for Real-time Analyt- ics in Blended Classrooms." In Proceedings of the LAK Conference, 257-66.
- Holstein, K., B. M. McLaren, and V. Aleven. 2018a. "Informing the Design of Teacher Awareness Tools Through Causal Alignment Analysis." In Proceedings of the ICLS Conference, 104-11.
- Holstein, K., B. M. McLaren, and V. Aleven. 2018b. "Student Learn- ing Benefits of a Mixed-reality Teacher Awareness Tool in AI- enhanced Classrooms." In Proceedings of the AIED Conference, 154-68.
- Holstein, K., B. M. McLaren, and V. Aleven. 2019a. "Co-designing a Real-time Classroom Orchestration Tool to Support Teacher-AI Complementarity." Journal of Learning Analytics 6: 27-52.
- Horvitz, E., and T. Paek. 2007. "Complementary Computing: Poli- cies for Transferring Callers from Dialog Systems to Human Receptionists." In Proceedings of the UMUAI, 159-82.
- Kamar, E. 2016. "Directions in Hybrid Intelligence: Complementing AI Systems with Human Intelligence." In Proceedings of the IJCAI, 4070-3.
- Licklider, J. C. 1960. "Man-computer Symbiosis." IRE Transactions on Human Factors in Electronics HFE-1: 4-11.
- Lubars, B., and C. Tan. 2019. "Ask Not What AI Can Do, But What AI Should Do: Towards a Framework of Task Delegability." In Proceedings of the NeurIPS, 57-67.
- Martinez-Maldonado, R., J. Kay, K. Yacef, M. T. Edbauer, and Y. Dimitriadis. 2013. "MTClassroom and MTDashboard: Supporting Analysis of Teacher Attention in an Orchestrated Multi-tabletop Classroom." In Proceedings of the CSCL Conference, 320-7.
- Martinez-Maldonado, R., A. Pardo, N. Mirriahi, K. Yacef, J. Kay, and A. Clayphan. 2015. "The LATUX Workflow: Designing and Deploying Awareness Tools in Technology-enabled Learning Set- tings." In Proceedings of the LAK Conference, 1-10.
- Miller, W. L., R. S. Baker, M. J. Labrum, K. Petsche, Y. H. Liu, and A. Z. Wagner. 2015. "Automated Detection of Proactive Remediation by Teachers in Reasoning Mind classrooms." In Proceedings of the LAK Conference, 290-4.
- Ritter, S., M. Yudelson, S. E. Fancsali, and S. R. Berman. 2016. "How Mastery Learning Works at Scale." In Proceedings of the ACM Learning@ Scale Conference, 71-9.
- Rodriguez Triana, M. J., L. P. Prieto Santos, A. Vozniuk, M. Shirvani Boroujeni, B. A. Schwendimann, A. C. Holzer, and D. Gillet. 2017. "Monitoring, Awareness and Reflection in Blended Technology Enhanced Learning: A Systematic Review." International Journal of Technology Enhanced Learning 9: 126-50.
- Schofield, J. W., R. Eurich-Fulcer, and C. L. Britt. 1994. "Teach- ers, Computer Tutors, and Teaching: The Artificially Intelligent Tutor As an Agent for Classroom Change." American Educational Research Journal. 31(3): 579-607.
- Stang, J. B., and I. Roll. 2014. "Interactions between Teaching Assistants and Students Boost Engagement in Physics Labs." Phys- ical Review Special Topics -Physics Education Research 10(2): 020117.
- Tan, S., J. Adebayo, K. Inkpen, and E. Kamar. 2018. "Investigating Human + Machine Complementarity for Recidivism Predictions." https://arxiv.org/abs/1808.09123. · arXiv
- Wilder, B., E. Horvitz, and E. Kamar. 2020. "Learning to Complement Humans." In Proceedings of the IJCAI, 1526-33.
- Yacef, K. 2002. "Intelligent Teaching Assistant Systems." In Proceed- ings of the Conference on Computers in Education, 136-40.
Cited by
Sorted by citations
Jeon, J., & Lee, S. (2023) · Education and Information Technologies · doi:10.1007/s10639-023-11834-1 · Cited by 696 · see citing works
Leading teachers' perspective on teacher-AI collaboration in education
Kim, J. (2024) · Education and Information Technologies · doi:10.1007/s10639-023-12109-5 · Cited by 376 · see citing works
Exploring How Machine Learning Practitioners (Try To) Use Fairness Toolkits
Deng, W. H., Nagireddy, M., Lee, M. S. A., Singh, J., Wu, Z. S., Holstein, K., & Zhu, H. (2022) · 2022 ACM Conference on Fairness Accountability and Transparency · doi:10.1145/3531146.3533113 · Cited by 99 · see citing works
Self-citationPark, H., & Ahn, D. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642785 · Cited by 84 · see citing works
Evaluating Trust in AI, Human, and Co-produced Feedback Among Undergraduate Students
Zhang, A., Gao, Y., Suraworachet, W., Nazaretsky, T., & Cukurova, M. (2025) · arXiv · doi:10.48550/arxiv.2504.10961 · Cited by 7 · see citing works
Being Brave in a New World: Leveraging ChatGPT in Signal Processing Classes
Banavar, M. K., Shri, L., Sparks, N., & Cohen, A. (2023) · 2023 IEEE Frontiers in Education Conference (FIE) · doi:10.1109/fie58773.2023.10342917 · Cited by 4 · see citing works
- Authors
- Holstein, K., & Aleven, V.
- Affiliations
- Carnegie Mellon University (US)
- Corresponding author
- Holstein, Kenneth
- Venue
- AI Magazine
- Publisher
- Wiley
- Published
- 2022-06
- Volume
- 43
- Issue
- 2
- Pages
- 239-248
- ISSN
- 0738-4602 (ISSN-L)
- 2371-9621
- OpenAlex
- W3144300208
- Cited by
- 156
- References
- 10
- Citations by year
- 2026: 58 · 2025: 56 · 2024: 19 · 2023: 18 · 2022: 5
- FWCI
- 8.06
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- Locations
- AI Magazine (publishedVersion, cc-by-nc)
- arXiv (Cornell University) (submittedVersion)
- AAAI Publications (The Association for the Advancement of Artificial Intelligence (AAAI)) (submittedVersion, cc-by)
- arXiv (Cornell University) (cc-by)
- Funders
- Division of Information and Intelligent Systems (1530726, 1822861)
- Institute of Education Sciences (R305A180301, R305B150008)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Intelligent Tutoring Systems and Adaptive Learning
- Physical Sciences › Computer Science › Information Systems › Artificial Intelligence in Education
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Augmented Reality Applications
- Keywords
- human-AI complementarity, human-AI collaboration, K-12 education, smart glasses, AI tutors, teacher-AI collaboration, metacognition, student learning, participatory design, stakeholder engagement, human-AI decision-making, theory development
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.
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
