Source-linked AI summary
Human-Centred Learning Analytics and AI in Education: a Systematic Literature Review
Riordan Alfredo, Vanessa Echeverria, Yueqiao Jin, Lixiang Yan, Zachari Swiecki, Dragan Gašević, Roberto Martinez-Maldonado
TL;DR
Privacy, agency, and limited stakeholder involvement create gaps in understanding human-centred LA/AIED design, particularly regarding control, automation, and trustworthy operation. This systematic literature review analysed 108 papers using a human-centred AI framework and found growing but incomplete consideration of these concerns, with limited end-user involvement in design. It recommends sustained stakeholder participation, especially by students, alongside stronger attention to safety, reliability, and trustworthiness.
Problem
The review addresses limited evidence about stakeholder involvement, human control, automation, and safety, reliability, and trustworthiness in human-centred LA/AIED systems.
Method
The authors conducted a PRISMA-guided systematic literature review of 108 empirical LA/AIED system studies involving educational stakeholders.
Results
The review found some consideration of human control and safety, reliability, and trustworthiness, but limited end-user involvement in actual system design.
Takeaways & Limitations
Future human-centred LA/AIED work should balance stakeholder involvement across design and deployment, involve target end-users, and address safety, reliability, and trustworthiness.
Takeaways & Limitations
The review may have missed relevant studies because of search terms, database coverage, exclusion of grey literature, and selection-judgment effects.
Abstract
from arXiv · showhide
The rapid expansion of Learning Analytics (LA) and Artificial Intelligence in Education (AIED) offers new scalable, data-intensive systems but also raises concerns about data privacy and agency. Excluding stakeholders -- like students and teachers -- from the design process can potentially lead to mistrust and inadequately aligned tools. Despite a shift towards human-centred design in recent LA and AIED research, there remain gaps in our understanding of the importance of human control, safety, reliability, and trustworthiness in the design and implementation of these systems. We conducted a systematic literature review to explore these concerns and gaps. We analysed 108 papers to provide insights about i) the current state of human-centred LA/AIED research; ii) the extent to which educational stakeholders have contributed to the design process of human-centred LA/AIED systems; iii) the current balance between human control and computer automation of such systems; and iv) the extent to which safety, reliability and trustworthiness have been considered in the literature. Results indicate some consideration of human control in LA/AIED system design, but limited end-user involvement in actual design. Based on these findings, we recommend: 1) carefully balancing stakeholders' involvement in designing and deploying LA/AIED systems throughout all design phases, 2) actively involving target end-users, especially students, to delineate the balance between human control and automation, and 3) exploring safety, reliability, and trustworthiness as principles in future human-centred LA/AIED systems.
1 Introduction
LA and AIED systems promise scalable, adaptive educational experiences but raise concerns about privacy, bias, accountability, and stakeholder trust. This review applies a human-centred AI framework to examine stakeholder involvement, human control, automation, and safety, reliability, and trustworthiness.
- LA/AIED systems offer data-intensive approaches intended to make teaching and learning more adaptive, personalised, accessible, scalable, and effective.
- Privacy and educational-data security concerns accompany the proliferation of LA/AIED systems.
- Limited student and teacher involvement can potentially undermine understanding and trust while raising accountability and transparency questions.
- Human-centred design prioritises human needs, values, and perspectives in designing and deploying LA/AIED systems.
- The review uses the human-centred AI framework to examine design, stakeholder involvement, human control versus automation, and safety, reliability, and trustworthiness.
2 Background
Human-centred AI frames LA/AIED as complementary human–machine systems requiring attention to agency, stakeholder participation, and safety, reliability, and trustworthiness. The review identifies unanswered questions about how these principles are represented across educational research and design.
- Human-centred AI emphasises complementarity between humans and machines, aiming to amplify, augment, and empower individuals through human-value principles.
- The HCAI framework distinguishes varying levels of human control and computer automation, which are not mutually exclusive.
- A single LA/AIED system can combine different control and automation levels across tasks and intended users, producing four framework quadrants.
- The four quadrants range from low-control, low-automation systems to high-control, high-automation systems that provide automated assistance while enabling manual operation.
- Systems combining substantial human control and automation should consider safety, reliability, and trustworthiness to support ethical practice and reduce data-misuse risk.
- The review asks how education levels, methods, design phases, HCD techniques, stakeholder contributions, control, automation, and HCAI principles have been addressed.
3 Method
The authors conducted a PRISMA-guided systematic literature review of recent, empirical, stakeholder-involving LA/AIED system studies. They searched four databases, screened studies using defined criteria, and analysed design phases, stakeholder roles, control–automation levels, and safety, reliability, and trustworthiness discussions.
- The review followed the PRISMA protocol and searched Scopus, ACM Digital Library, IEEE Xplore, and Web of Science.
- The search covered publications from January 1, 2012 to November 1, 2023 and initially yielded 1,678 articles, with 511 duplicates removed.
- Eligibility required empirical studies developing data-intensive educational systems for end-users and involving stakeholders, while excluding purely technical studies without learning implications.
- 108 articles were selected for data extraction after full-text screening, with Cohen’s kappa values of 0.71 and 0.67 indicating substantial agreement.
- The review mapped study characteristics, stakeholder involvement, design phases, HCD techniques, control–automation quadrants, and safety, reliability, and trustworthiness themes.
4 Results
The review found that human-centred LA/AIED research concentrates on middle design phases, commonly involves students and teachers, and more often considers substantial human control than high automation. However, stakeholder participation is frequently passive, especially among students, and safety, reliability, and trustworthiness receive uneven attention.
- Current research state: 53% of studies addressed Phase 2, while only 9% addressed Phase 5 of the HCD process.Phase 2 covered exploration, synthesis, and design implications; Phase 5 covered launch and monitoring.
- Current research state: Interviews (52%), questionnaires (47%), co-design (28%), and prototype validation (26%) were the most common HCD techniques.Focus groups, observations, and surveys were used less often, at 10%, 10%, and 8%, respectively.
- Stakeholder involvement: Stakeholder involvement was lowest in Phases 1 and 5, at 18% and 10%, while Phase 4 was the only phase with more passive than active involvement.Phase 4 had 31% passive and 25% active involvement, suggesting evaluation outcomes were not consistently used to improve designs.
- Stakeholder involvement: Students appeared in 71% and teachers in 59% of studies, but students had 52% passive involvement and comparatively low active involvement.Teachers had higher active involvement at 39%, while their passive involvement was 20%.
5 Discussion
The review finds that human-centred LA/AIED systems often prioritise human control, but students and teachers remain insufficiently involved across design phases. It recommends broader stakeholder participation, clearer end-user roles, and stronger attention to system safety, reliability, and trustworthiness.
- Design phases: Phase 1 and Phase 5 contained only 18% and 9% of systems, indicating insufficient attention to inception and monitoring.The authors argue that stakeholder engagement at these stages can improve alignment with learning needs and support sustainable solutions.
- Recommendations: The review recommends expanding research beyond universities and K-12, increasing active co-creation, and involving end-users throughout all design phases.Recommended approaches include co-design, participatory design, value-sensitive design, clearer teacher and student roles, and explicit attention to student involvement.
- Stakeholder involvement: Stakeholder involvement was limited during evaluation and refinement, potentially leaving design flaws unresolved in authentic educational settings.The review recommends post-hoc reflection interviews and other activities that allow end-users to realign system features with their experiences.
- Stakeholder involvement: Students represented 71% of participants, but their active involvement in design was only 19%.Students were the most frequently engaged stakeholders overall, yet their participation in design activities remained relatively low.
- Human control and automation: Human-centred systems predominantly prioritised human control, with 32% in Q2 and 47% in Q4.The review used the two-dimensional HCAI framework to classify systems by human control and computer automation.
6 Concluding remarks
The review reports growing interest in human-centred LA and AIED research while highlighting the importance of safety, reliability, trust, communication, and user control. It also identifies research gaps and methodological challenges that future work must address.
- Concluding remarks: Human-centred LA and AIED research is growing, but gaps and methodological challenges remain.The conclusion highlights safety, reliability, trust, transparent communication, and user control as important areas for design and implementation.
A HCAI two-dimensional quadrants in LA/AIED references
The appendix lists included LA/AIED references according to the four quadrants of the two-dimensional HCAI framework. The framework distinguishes combinations of human control and computer automation.
- Coverage: The reference list includes studies assigned to Q1, Q2, Q3, and Q4 across the reviewed LA/AIED literature.The appendix entries span multiple studies and venues, including the references continued in the table listing.
- Quadrant reference list: Table 1 organises included articles by quadrant, using Q1–Q4 classifications in the two-dimensional HCAI framework.The listed entries provide reference information alongside their quadrant assignments.