GENIE Learn: Human-Centered Generative AI-Enhanced Smart

Learning Environments

Carlos Delgado Kloos

, Juan I. Asensio-Pérez

, Davinia Hernández-Leo

Pedro Manuel Moreno-Marcos

, Miguel L. Bote-Lorenzo

, Patricia Santos

Carlos Alario-Hoyos

, Yannis Dimitriadis

and Bernardo Tabuenca

Universidad Carlos III de Madrid, Av. Universidad 30, 28911 Leganés (Madrid), Spain

Universidad de Valladolid, Paseo de Belén 15, 47011 Valladolid, Spain

Universitat Pompeu Fabra, Roc Boronat 138, 08018 Barcelona, Spain

Universidad Politécnica de Madrid, Calle Alan Turing sn, 28031 Madrid, Spain

Keywords: Smart Learning Environments, Hybrid Learning, Generative Artificial Intelligence, Human Centeredness.

Abstract: This paper presents the basis of the GENIE Learn project, a coordinated three-year research project funded

by the Spanish Research Agency. The main goal of GENIE Learn is to improve Smart Learning Environments

(SLEs) for Hybrid Learning (HL) support by integrating Generative Artificial Intelligence (GenAI) tools in a

way that is aligned with the preferences and values of human stakeholders. This article focuses on analyzing

the problems of this research context, as well as the affordances that GenAI can bring to solve these problems,

but considering also the risks and challenges associated with the use of GenAI in education. The paper also

details the objectives, methodology, and work plan, and expected contributions of the project in this context.

1 INTRODUCTION

Recent advances in the interdisciplinary field of

Technology-Enhanced Learning (TEL) have made

possible novel models of teaching and learning based

on a mixture or fusion of traditional approaches along

different dimensions or dichotomies: learning in

physical/digital spaces, informal/formal learning,

face-to-face/online learning, individual/collaborative

active learning, etc. (Hilli et al., 2019). These novel

TEL models, which showcased their relevance during

the COVID pandemic, are studied under the

theoretical umbrella of the so-called Hybrid

Learning (HL) (Cohen et al., 2020, Gil et al., 2022),

which blurs the boundaries of those dichotomies.

https://orcid.org/0000-0003-4093-3705

https://orcid.org/0000-0002-1114-2819

https://orcid.org/0000-0003-0548-7455

https://orcid.org/0000-0003-0835-1414

https://orcid.org/0000-0002-8825-0412

https://orcid.org/0000-0002-7337-2388

https://orcid.org/0000-0002-3082-0814

https://orcid.org/0000-0001-7275-2242

https://orcid.org/0000-0002-1093-4187

Smart Learning Environments (SLEs) can be

useful as technological support for HL (Delgado

Kloos et al., 2018 and 2022). SLEs are conceived to

provide personalized support to learners, considering

learning needs and context. SLEs “collect data from

the learning context (sense), decode, process the data

collected (analyze), and coherently suggest actions to

ease learning constraints toward improved learning

performance (react)” (Tabuenca et al., 2021). SLEs

expand the work from context-aware, ubiquitous

learning, and adaptive learning systems, actively

supporting students according to their learning

situation, across physical and virtual learning spaces

(Gross, 2016), based on data-driven interventions

(Hernández-Leo, et al., 2023a).

Delgado Kloos, C., Asensio-Pérez, J. I., Hernández-Leo, D., Moreno-Marcos, P. M., Bote-Lorenzo, M. L., Santos, P., Alario-Hoyos, C., Dimitriadis, Y. and Tabuenca, B.

GENIE Learn: Human-Centered Generative AI-Enhanced Smart Learning Environments.

DOI: 10.5220/0013076000003932

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 17th International Conference on Computer Supported Education (CSEDU 2025) - Volume 2, pages 15-26

ISBN: 978-989-758-746-7; ISSN: 2184-5026

SLEs rely on advances in both Learning Design

(LD) (Wasson & Kirschner, 2020) and Learning

Analytics (LA) (Long & Siemens, 2011). Thanks to

LD tools, teachers can make their pedagogical

intentions explicit, and even represent them in

computer-interpretable formats, thus enabling SLEs

to use them as contextual inputs. LA constitutes a key

component of SLEs collecting data from both

physical and virtual spaces. SLEs are aimed at

modelling students in context to provide adequate

personalized support, in many cases making use of

Artificial Intelligence (AI) algorithms (Buckingham

et al., 2019a). The synergetic relationship between

LD and LA is key for supporting data-driven

interventions in SLEs: interventions based on LA

indicators aligned with LDs are typically more

meaningful; and the effectiveness of LDs can be

better understood in the light of LA indicators. This

relationship between LD and LA is also reflected in

the so-called Design Analytics (metrics of design

decisions and related aspects characterizing LDs) and

Community Analytics (metrics and patterns of LD

activity) (Hernández‐Leo, et al., 2019). Finally,

academic analytics, when supporting educational

decision making at institutional level, can be seen as

a variation of LA also in interplay with

complementary data layers (Misiejuk et al., 2023).

Thus, academic analytics can also benefit from the

core functions of SLEs, while differing in the

requirements of the stakeholders (educational

managers, beyond teachers) and the scale of the data

potentially relevant for its analysis (Hernández-Leo et

al., 2019, Ortiz Beltrán et al., 2023). Figure 1

summarizes the research context, which includes

technological, pedagogical and human contexts.

Figure 1: Research context (human, pedagogical,

technological).

This paper builds on this research context and

presents the theoretical foundations and research plan

of the GENIE Learn project, a coordinated research

project funded by the Spanish Research Agency and

with the participation of three Spanish Universities

(Universidad Carlos III de Madrid, Universidad de

Valladolid, and Universitat Pompeu Fabra). GENIE

Learn is aimed at improving SLEs for HL support by

integrating Generative Artificial Intelligence (GenAI)

tools in line with the preferences and values of the key

stakeholders (teachers, learners, and academic

managers). The remaining of this paper is as follows:

Section 2 analyzes potential problems present in the

research context as well as the affordances and

challenges of GenAI tools to address these problems.

Section 3 presents the initial hypothesis and

objectives of the GENIE Learn project. Section 4

discusses the methodology and work plan. Finally,

Section 5 draws the conclusions of this article.

2 PROBLEMS, AFFORDANCES,

AND CHALLENGES

2.1 Problems in the Research Context

There are currently several significant problems (P)

to be addressed when trying to improve the support to

HL provided by state-of-the-art SLEs.

P1. Improving the support for LD in

connection with LA. Educators can use multiple

methods and tools to support the process of LD in a

variety of HL contexts (Wasson & Kirschner, 2020).

However, the complexity of the LD process has been

a recognized challenge, hindering the widespread

adoption of LD methods and tools, despite the field’s

significant relevance (Dagnino, et al., 2018). This

complexity arises from the limitations of the tools in

meeting the theoretical ambitions of the field. The

goal is to enable multiple stakeholders to get involved

in the creative inquiry process of data-driven co-

designing for learning, generating design artifacts that

evolve through time and feed the successive phases

(Michos & Hernández-Leo, 2020). A new generation

of LD tools for SLE is necessary to achieve its goals.

These LD tools should facilitate the collaboration of

various stakeholders towards pedagogically-sound

LDs and enable a thorough exploration of the

educational context’s needs and consideration of LA

from previously implemented designs. These tools

should also support data-driven inspiration drawing

on analytics from related design artifacts (Hernández-

Leo, et al., 2019), predominantly available in

CSEDU 2025 - 17th International Conference on Computer Supported Education

unstructured formats and created by teacher

communities (Gutiérrez-Páez et al., 2023). In

summary, enhanced support is needed for needs

analysis, brainstorming, inspiration, and creative

ideation, informed pedagogical decision-making,

scaffolding and integration of stakeholder

conversations, and improved interpretation and

actionability of advanced (LD community) analytics.

P2. Supporting a wider diversity of learners

(universal design for learning) and ethical design

aspirations. The Universal Design for Learning

(UDL) framework is structured around three main

principles: a) Multiple means of representation; b)

Multiple means of engagement; c) Multiple means of

expression. However, the seamless integration of the

principles into educational practices brings

challenges, demanding a reconsideration of teaching

methodologies and materials. The implementation of

UDL by educators demands specific knowledge as

well as substantial time and effort (Rose and Meyer,

2002; Evmenova, 2018). There is a lack of LD tools

(Gargiulo and Metcalf, 2010) that support the

adaptation of pedagogical and educational material

design considering UDL. Such tools would enable

stakeholders to more efficiently and effectively

embrace UDL principles.

P3. Reducing the burden on teachers

associated with the real-time orchestration

demands of HL environments. An important

challenge in SLEs refers to the “orchestration load”

(Amarasinghe et al., 2022). This concept reflects the

teachers’ attentive processing during the real-time

management of complex learning scenarios.

Designing tools to lower this burden should be a top

priority. The results from a recent international

workshop (Amarasinghe et al., 2023a) highlight the

need to provide orchestration tools that consider

different needs depending on the delivery mode,

pedagogical method (individual/collaborative),

teacher characteristics, and content knowledge being

considered in the learning activity (Hakami et al.,

2022, Raes et al., 2020). This has implications in the

type of data collected from the stakeholders, as well

as in its analysis and presentation to support teachers

in e.g., orchestration dashboards. The main data

source is quantitative data related to student

engagement (LA presented to teachers) and teachers’

actions in the dashboard (for researchers to study

orchestration load) (Amarasinghe, et al., 2021). A

problem is that understanding the orchestration load

involves the use of additional data collection sources

and more advanced approaches for data analysis and

presentation in SLEs. This includes real-time analysis

of students individual and collective self-

explanations in their answers to learning activities

(e.g., to lower the burden when intervening if a

problem is identified, or during debriefing sessions),

as well as in the collection and triangulated analysis

of teachers’ data coming from sensors in SLE that

goes from physiological sensors to reflection diaries

and video recordings.

P4. Large-scale analysis of SLE data to support

decision making in academic management. LA

(from learners) and design analytics (from teachers)

collected in SLE have the potential to support

academic managers in institutional decision making

(e.g., evolution of the educational model,

identification of needs for teacher training)

(Hernández-Leo, et al., 2019). Yet, this potential has

not been exploited. Only limited, but relevant

approaches considering student ratings regarding

their teaching satisfaction, e.g., depending on the HL

modalities (Ortiz Beltrán et al., 2023), and

classifications of course designs approaches have

been developed so far (Toetenel & Rienties, 2018;

Misiejuk et al., 2023). The problem is that the

envisaged potential requires an analysis at large scale

involving multiple courses, data types and formats

(including e.g., course and program descriptions,

student feedback on course design) and an integrated

analysis of the different relevant data sources.

P5. Limited types of automatically generated

personalized context-aware learning tasks and

feedback interventions. SLEs can generate

automatic, data-driven interventions in HL scenarios

with minimum or no teacher involvement, thus

opening the possibility of personalized learning at

different scales (e.g., when many students participate

in the learning scenario). For example, Ruiz-Calleja

et al. (2021) propose the use of Linked Open Data in

the Web for automatically generating large numbers

of contextualized learning tasks in the domain of

cultural heritage. However, that approach relies on

the use of a small and rigid set of teacher-generated

task templates that only allow the generation of some

very specific types of learning tasks (e.g., asking

about the architectural style of a church, or suggesting

taking a photo of a generic architectural element) that

cannot always be automatically assessed. The

generation of more complex learning tasks such as the

explanation of the characteristics of a building in

relation to its historical context cannot be generated

with the current technological solutions. Similarly,

the provision of personalized feedback in large-scaled

educational settings has also been explored (e.g.,

Topali et al., 2024) resulting in the development of

technological support tools (Ortega-Arranz et al.,

2022). However, current technological solutions for

GENIE Learn: Human-Centered Generative AI-Enhanced Smart Learning Environments

personalized feedback in SLEs only support a limited

number of types of feedback interventions (e.g.,

predefined email messages encouraging the students

to carry out certain tasks) based also on a limited set

of LA indicators (e.g., the score in an online quiz).

More elaborated types of feedback interventions (e.g.,

providing an automatic explanation about the reasons

for a low score in a text-based learning task) are also

not possible for current SLEs.

P6. Students’ models automatically generated

at large scales are mostly based on

quantitative/structured data. SLEs can generate

individual students’ models used as inputs for

automatic personalized interventions (e.g., Serrano-

Iglesias et al., 2021). However, state-of-the-art SLEs

base those students’ models mostly on LA indicators

that make use of structured data coming from Virtual

Learning Environment, physical sensors, etc. LA

indicators derived from other, non-structured learning

data such as text-based learning outcomes have also

been proposed but typically based on quantitative

features such as temporal evolution of the length of

the documents, the number of editions, etc. (e.g.,

Suraworachet et al., 2021). The building of students’

models based on unstructured learning data (e.g., the

actual contents of a document produced by a student)

is a desirable feature not found in current SLEs.

P7. Improving personalized and more effective

learning in scenarios that use conversational

interfaces and (self-selected) artificial intelligence

(AI) tools. SLEs have been incorporating the use of

conversational interfaces to promote active and

authentic learning experiences (e.g., in social media

education, Theophilou, et al., 2023a). Personalization

in these contexts is often achieved through LA,

adapting activity flow to individual needs, and

considering students use their own self-selected tools

(e.g., Ognibene et al., 2023). However, despite the

advantages these scenarios offer for enhanced and

efficient learning, several challenges remain. These

include: a) the restricted free interaction capabilities

of available conversational-oriented SLEs and the

identified effects in the limited writing quality of

students’ submissions to conversational interfaces

(Theophilou et al., 2023b); b) the necessity of

scaffolding functions to bolster student self-

regulation and collaboration quality, considering

socio-emotional aspects (Hadwin, 2017); c) the

effects of individual attitudes (openness) influence

attitudes in the use of supporting tools, especially AI

tools (Sánchez-Reina et al., 2023); d) the significant

concerns related to academic integrity in education

related to the use of AI tools (Kasneci, 2023).

2.2 Affordances of GenAI

GenAI is the branch of AI aimed at creating realistic

content such as text, images, audio, or video, based

on a given input or prompt (Jovanovic & Campbell,

2022). There are many types of GenAI: Generative

Adversarial Networks, Generative Diffusion Models,

Generative Pre-trained Transformers (GPTs), etc.

Although modern GenAI is based on decades-old

models and techniques (e.g., neural networks, back

propagation, etc.), the availability of improved

computational capabilities, the increase in the size of

models, and huge training datasets have made

possible that some recent incarnations of GenAI

systems, such as OpenAI’s GPT-4, show a

performance that is “strikingly close to human-level

performance” (Bubeck at al., 2023). Previous efforts

on AI in Education (AIED) and LA did not anticipate

the public availability of more powerful GenAI tools

(e.g., OpenAI’s ChatGPT or DALL-E, Google

Gemini, etc.) capable of carrying out a wide range of

tasks with “zero-shot” or “few-shot” training (i.e.,

without the need to provide a lot of additional training

data for fine-tuning the model) and by simply using

textual prompts as inputs (Brown et al., 2020). These

new capabilities have enabled students and teachers

to explore with little effort creative ways of

integrating GenAI tools in their learning and teaching

tasks (Kasneci et al., 2023). Despite conflicting

perceptions (both euphoric and worrisome, or even

apocalyptic) of this rushed introduction of GenAI in

education (Rudolph et al., 2023), the TEL community

is making important efforts to understand its impact,

opportunities and challenges. Specifically, in the

context of the research problems previously identified

in HL support using SLEs, new GenAI tools may

bring significant opportunities and affordances (A):

A1 in relation to P1. GenAI may help SLEs

provide better support to communities of teachers

during the process of creating LDs for HL. For

instance, Demetriadis & Dimitriadis (2023) illustrate

how to use GPT-3 for creating a conversational agent

that reuses design knowledge extracted from existing

design conversations. Hernández-Leo (2023b)

proposes speculative functions in which GenAI

integrated with analytics layers (Hernánez-Leo et al.,

2019) may support the LD life cycle. Also, some

preliminary results were published on the enrichment

of LD tools with Design Analytics (Albó et al., 2022)

automatically generated by GenAI. For instance,

Pishtari et al. (2024) study the impact of LLM-

generated feedback on the quality of LDs produced

by teachers.

CSEDU 2025 - 17th International Conference on Computer Supported Education

A2 in relation to P2. UDL principles emphasize

the importance of accommodating diverse learning

styles and needs in educational environments. GenAI

emerges as a potentially highly helpful tool for

educators, facilitating the application of UDL

principles in educational settings in two ways: a) by

reducing the time and effort demanded from teachers

through the provision of specialized knowledge and

the automation of tasks (Lim et al., 2023); and b) by

enhancing the overall quality of teaching and learning

experiences. GenAI can play several roles in the

support of UDL including the recommendation of

effective communication strategies for students with

special needs (Garg and Sharma, 2020), or the

suggestion of diverse, pedagogically sound content

tailored to the students’ needs (Mizumoto, 2023).

A3 in relation to P3. GenAI conversational

approaches offer the potential to support teachers

while implementing learning scenarios (Sharples,

2023), prompting for different analysis and human-

readable explanations (Susnjak, 2023) of student

progress for guiding interventions and feedback on-

the-fly. GenAI may also help to advance the analytics

used in orchestration dashboards so textual

descriptions and visualizations about the students’

actions and unstructured answers are presented to

support teacher-led debriefing in HL (Hernández-

Leo, 2023b). Multimodal analytics based on sensor

data (EEG, heart rate, eye-tracking) and (fine-tuned)

GPTs can also be used to analyze classroom

orchestration data (Crespi et al., 2022; Amarasinghe

et al., 2023b; Tabuenca et al., 2024)

A4 in relation to P4. GenAI may help the large-

scale analysis and integration of relevant data sources

that can potentially support holistic education

decision making to academic managers. The

automatic collection of relevant data (sense),

including qualitative description (e.g., course

descriptions), can take advantage of approaches used

in web analytics (e.g., Calvera-Isabal et al., 2023),

which facilitates the creation of rich datasets for an

integrated analysis of relevant data sources. GenAI

can play pivotal roles in the analysis of unstructured

data and in advancing interactive and explanatory

analytics, which needs to be approached ethically

(Susnjak, 2023; Yan et al, 2023). Examples of the

potential includes from text similarity analysis across

course designs (e.g., extracting information about the

pedagogical model to be triangulated with students’

performance and satisfaction) to the generation of text

explaining insights to the stakeholders about the

alignment of data indicators with institutional

priorities (e.g., development of desired common

competences across study programs).

A5 in relation to P5. Some early research results

(see, e.g., Kalo et al., 2020) suggest that LLMs might

improve the accuracy and flexibility of SLEs that

make use of Linked Open Data in the Web. GenAI

may also help SLEs improve the way personalized

and context-aware reactions are delivered to human

stakeholders (students, teachers, etc.) in HL settings.

For example, Dai et al. (2023) gathered promising

empirical results about the effectiveness of LLM to

automatically generate feedback for learners. In this

way, LLMs could potentially be used to compare a

student’s response with the information available on

the Web of Data and generate a reaction that explains

what aspects of the response were incorrect.

A6 in relation to P6. GenAI may help SLEs

collect (sense) and analyze learning data sources

based on natural language, thus widening the range of

HL situations that might be supported by SLEs. For

example, Amarasinghe et al. (2023b) fine-tuned

GPT-3 to automatically code text data from a learning

setting and provided evidence of its performance with

respect to alternative approaches. This type of GenAI

applications might be used by SLEs to automatically

assess which concepts are not adequately covered by

students in, e.g., a written essay, and thus trigger

personalized feedback interventions and

recommendations of additional learning tasks to

reinforce those concepts (Pereira et al., 2023).

A7 in relation to P7. Advances in language

learning models with zero-shot learning capabilities

suggest a new possibility for developing educational

chatbots for personalized learning using a prompt-

based approach. Preliminary tests were already

conducted in a case study on effective educational

chatbots with ChatGPT prompts (Koyuturk et al.,

2023). The results are encouraging, although more

research is needed as challenges are posed by the

limited history maintained for the conversation and

the highly structured form of responses by ChatGPT,

as well as their variability. It would be possible to use

automatized prompting engineering methodologies

(Pryzant et al., 2023) to advance this line of research.

On the other hand, providing features embedded in

learning platforms that offers adaptive (depending on

students’ needs) teachable moments (Ognibene et al.,

2023; Hernández-Leo, 2022) related AI literacy (e.g.

learning to prompt) is expected to improve attitudes

and quality of interactions with AI-driven systems

(Theophilou et al., 2023c). Finally, the definition of

new constructs for LA considering the new

requirements imposed using self-selected AI tools in

learning processes would enable the development of

systems facilitating solutions to address concerns

related to academic integrity in education.

GENIE Learn: Human-Centered Generative AI-Enhanced Smart Learning Environments

2.3 Challenges of GenAI in Education

Despite the promising advantages of using GenAI,

the SLE-based support to HL cannot be oblivious to

the significant risks and challenges (C) that GenAI

pose to education (see, e.g., European Commission,

2023; UNESCO, 2023) and that need to be addressed

by the research initiatives, educational institutions

and policy makers. Some of these challenges include:

C1. Lack of alignment of GenAI with human

values for learning. Although the ethical

implications of AI in education have been for years a

strong concern for the research community (e.g.,

Akgun & Greenhow, 2021) and for policy makers

(e.g., HLEG-AI, 2019), many of the recently released

GenAI tools may be contributing to the worsening of

the situation. UNESCO (2023) has already identified

ethical risks of GenAI in education and research: a)

concentration of GenAI usage in technologically

advanced countries, b) lack of democratic control of

GenAI companies, c) data protection and copyright

issues, d) use of unexplainable models, e) biases in

model training and generated content, f) reduction of

diversity, and g) manipulation of content. The lack of

transparency, privacy and the diminishing of equality

have also been identified by Yan et al. (2023) in

recent research about GenAI in education. However,

the same authors also agree on the need for “adopting

a human-centered approach throughout the

developmental process” of GenAI tools so as “to

protect human agency and genuinely benefit students,

teachers and researchers” (UNESCO, 2023), thus

trying to overcome the posed ethical challenges.

Interestingly, the GenAI research community itself is

also paying attention to the challenge of making

GenAI tools (with special emphasis on LLMs)

“aligned” with human preferences and values, i.e.,

making them more helpful, honest, and harmless

(HHH framework) (Askell et al., 2021). To address

the alignment problem, recent research suggests

going beyond the mere scaling up of GenAI models

and incorporating model fine-tuning based on human

feedback (Ouyang et al., 2022). In the case of SLEs

for the support of HL, incorporating GenAI solutions

may contribute, especially in the “react” function, to

less explainable and more biased automatically

generated interventions, which can increase the

barrier for adoption of this technology among human

stakeholders (Serrano-Iglesias et al., 2023).

C2. Challenges to current forms of Human-AI

collaboration. The low effort required to use GenAI

tools may have a negative effect on the creativity and

critical thinking skills of the human actors in

education, eventually causing a heavy reliance on

GenAI (Kasneci et al., 2023). Beyond the ongoing

debate on whether GenAI tools should be either

banned or fostered in education (see, e.g., Rudolph et

al., 2023), there seems to be a consensus on the need

for teachers to develop new skills to incorporate

GenAI tools into their practice (Baidoo-Anu &

Ansah, 2023). Kasneci et al. (2023) suggest several

ways to address this goal: develop new education

theory, provide adequate guidance and teacher

training, create resources and guidelines for educators

and institutions, nurture communities of educators to

share and reuse knowledge in applying these new AI

tools, among others. All these changes in education

should be accompanied by new models of the so-

called Human-AI collaboration (also known as hybrid

AI-human approaches): finding the proper balance

when sharing tasks among humans and AI at different

moments of the teaching-learning processes. Recent

proposals (see, e.g., Molenaar, 2022, and Järvelä et

al., 2023) explore AI-human approaches in TEL

settings, although their proposals do not consider the

specific case of GenAI. In the case of SLEs for the

support of HL finding a right balance between

automation and human autonomy and decision-

making (also known as “agency”) is particularly

challenging. This balance has been explored through

educators’ agency (e.g., Alonso-Prieto, 2023) mainly

from the perspective of “orchestration”. Also,

learners’ agency has been explored (see, e.g., Villa-

Torrano et al., 2023) from the perspective of self,

socially shared and co-regulation of learning (Hadwin

et al., 2017) in which learners take metacognitive

control of their individual and/or collective cognitive,

behavioral, motivational and emotional processes.

Although the use of previous generations of AI for

detecting and supporting regulation of learning has

been widely explored in the LA field (see, e.g.,

Järvelä, et al., 2023), capabilities of GenAI in terms

of natural language processing (e.g., for detecting

regulatory episodes in group conversations) and text

generation (e.g., for automatically generating

personalized feedback regarding learning regulation

issues) are still under explored (Gamieldien, 2023). In

any case, the impact of novel GenAI solutions in the

agency of human stakeholders involved in HL

settings supported by SLEs needs to be re-assessed

(Hernández-Leo, 2022): will GenAI tools increase

teachers’ orchestration load (in the already

challenging environment of HL)?; how does GenAI-

enhanced support for LD affect efficiency but also

perspective taking and pedagogical creativity of

educators?; how may GenAI tools affect learners’

metacognitive processes now that certain learning

tasks can be easily solved by those tools?.

CSEDU 2025 - 17th International Conference on Computer Supported Education

C3. There is scarce research evidence about the

benefits of GenAI for education in authentic

educational practice. For instance, Yan et al. (2023)

did not identify any research work showcasing the use

of LLMs in “successful operations” (TRL-6 level in

the Technology Readiness Level scale, ISO, 2013).

Moreover, evidence is needed as it is imperative for

ethical GenAI solutions supporting education to be

aware of their limitations (Sharples, 2023). Those

limitations need to be continuously considered in the

responsible conception of AI for education so they are

transparently presented in the designed functions for

users for their mindful use (Hernández-Leo, 2022).

Future research on GenAI and education also needs

to pay careful attention to the way empirical results

are reported (Yan et al., 2023), trying to provide as

many details as possible (e.g., employed prompts,

models, source code, data for fine-tuning, etc.) with

the ultimate goal of fostering replicability as much as

possible in a context in which most training datasets

and algorithms are not disclosed. This scarcity of

empirical evidence about the impact of GenAI also

affects the support of HL with SLEs. However, HL

and the associated technological support in the form

of SLEs provide a relevant and wide pedagogical and

technological context for researching on the impact

and risks of GenAI in TEL settings.

3 INITIAL HYPOTHESIS AND

OBJETIVES

The GENIE Learn project focuses on the

technological support for HL under the initial

hypothesis that the integration of GenAI tools into

SLEs may help to overcome important limitations in

the current state-of-the-art. Therefore, the main goal

of this project can be formulated as: “to improve

SLEs for HL support by integrating GenAI tools

in a way that is aligned with the preferences and

values of human stakeholders”. Alignment is here

understood using the HHH framework (helpful,

honest, and harmless) (Askell et al., 2021) as a

starting point and addressed following human-

centered principles (Buckingham Shum et al., 2019b;

HLEG-AI, 2019) and hybrid AI-human approaches

(Järvelä et al., 2023; Siemens et al., 2022). More

specifically, GENIE Learn applies Value Sensitive

Design (VSD) principles and techniques to consider

human values when designing such GenAI-enhanced

SLEs. Considering the value alignment perspective,

human-centered principles, AI-human collaboration

approaches and VSD principles as transversal

requirements for the project, the main objectives (O)

of this project can be formulated as follows:

• O1: To define a research framework, consisting

of: 1) a systematic analysis on the use of GenAI

tools for supporting SLE functions, as well as on

novel Human-AI collaboration models in

education; 2) a pedagogical model for HL that

considers the affordances and impact of GenAI in

the different stakeholders; 3) the definition of a set

of HL scenarios in the context of SLEs, co-

designed with collaborating teachers and

educational institutions, that illustrate the

affordances of GenAI tools and their eventual lack

of alignment; 4) the definition of human-AI

collaboration models applicable to the project

scenarios that take into account educational goals

as well as the agency of teachers (focused on

orchestration) and learners (focused on regulated

learning); and 5) the definition of a set of research

instruments and a methodology for reporting

GenAI-related research results.

• O2: To design and develop GenAI-enhanced

solutions for teachers and academic managers

improving learning design and academic

decision making in SLEs for HL support by

addressing the problems: 1) improving the

support for LD in connection with LA; 2)

supporting a wider diversity of learners (UDL)

and ethical design aspirations; 3) reducing the

burden on teachers associated with real-time

orchestration demands of HL environments; and,

4) large-scale analysis of SLE data to support

decision making in academic management.

• O3: To design and develop GenAI-enhanced

solutions to improve support for learners in

SLEs for HL support by addressing the

problems: 1) limited types of automatically

generated personalized context-aware learning

tasks and feedback intervention; 2) students’

models automatically generated at large scales are

mostly based on quantitative/structured data and

may include several types of biases (e.g., gender

bias); and, 3) improving personalized and more

effective learning in scenarios that use

conversational interfaces and AI tools.

• O4: To define a technology framework as an

integrated infrastructure, consisting of: 1) a

selection of GenAI platforms and tools (according

to O1); 2) the architecture and a development of

the integrated infrastructure of an advanced SLE

that makes use of the affordances of GenAI tools

in ways that are aligned with the stakeholder

values and Human-AI collaboration models (O1),

and that provides the infrastructure for the outputs

GENIE Learn: Human-Centered Generative AI-Enhanced Smart Learning Environments

from the tasks derived from O2 and O3; and; 3)

technical guidelines for the integration of existing

educational datasets with the proposed GenAI-

enhanced architecture, so as to improve current

approaches to data-driven interventions in SLEs.

• O5: To design, implement and evaluate pilot

experiences in real settings, of the outcomes of

the project following a human-centered approach.

The pilots demonstrate the potential of the project

contributions considering several educational

topics and levels e.g., primary/secondary

education, higher education, and lifelong

learning. The HL scenarios (O1) are used as a

starting point for the design of the pilot

experiences, and the infrastructure (O4) is the

basis for their technological implementation.

4 METHODOLOGY AND WORK

PLAN

4.1 Methodology

The main goal of the GENIE Learn project involves

enhancing SLEs for HL by incorporating GenAI in a

manner consistent with the values of educational

stakeholders. The project aims to develop solutions

that address multiple challenges in particular

educational settings (see objectives O1 and O5),

while simultaneously expanding the understanding of

effective technology design (O2, O3, and O4). Given

these aims, the Design Science Research

Methodology (DSRM) is a suitable methodological

framework for guiding the project. DSRM (Peffers et

al., 2007) is a widely used iterative methodology in

information systems research. DSRM consists of six

phases: 1) problem identification and motivation, 2)

definition of the objectives for a solution, 3) design

and development of the artifact, 4) demonstration of

the artifact in a relevant context, 5) evaluation of the

artifact and its outcomes, and 6) communication of

the research results.

DSRM is also appropriate for the objectives of the

GENIE Learn project due to its humanistic dimension

(i.e., the issue of aligning GenAI improvements with

human stakeholder values), since DSRM is designed

to “create things that serve human purposes” with a

research-oriented perspective (Peffers et al., 2007).

The human dimension of the present project also calls

for the use of human-centered research and

educational technology design methods (see, e.g.,

Buckingham Shum et al., 2019), and a human-

centered AI perspective (HLEG-AI, 2019).

Particularly, the project follows a hybrid human-AI

approach (Järvelä et al., 2023; Siemens et al., 2023)

in which the AI elements are not meant to fully

automate the activities of educational stakeholders,

but rather complement their abilities in a way that

preserves human agency.

The GENIE Learn project also relies on value-

sensitive design (VSD) to align human and AI. VSD

(Friedman et al., 2017) is a theoretically grounded

approach to the design of technology that explicitly

inquires and models human values, their trade-offs,

and tensions. In terms of human stakeholder values

and their impact on the design of the project’s

conceptual and technological proposals, the project

uses Askell’s (2021) HHH framework (helpful,

honest, and harmless) as the starting point, which is

expanded by the state-of-the-art activities needed to

define the project’s research framework (O1), and is

further aligned with the more specific human

stakeholder values to be elicited during early

engagement activities with stakeholders from the

educational settings and domains of application.

Another important issue in the project’s

methodology is the complexity of both the concerned

technologies (SLEs, and their GenAI enhancements)

and their contexts of application (HL). The

understanding of this complexity demands a mixed

methods strategy (Johnson et al., 2007) in the data

collection and analysis, going beyond the use of

purely quantitative/qualitative approaches to obtain a

more holistic picture, e.g., of the pilot experiences of

use of the technology.

4.2 Work Plan

The work plan for GENIE Learn is organized in 6

work packages (WPs) following the DSRM (see

Figure 2 with more details below); each of the first 5

WPs pursues one of the 5 objectives (O1-O5) while

the last WP deals with the coordination of the project

(Figure 3).

• WP1: Research Framework. This WP includes

tasks related to the state of the art on GenAI tools

for supporting SLE functions and on Human-

GenAI collaboration models, the scenarios that

are used to illustrate and evaluate the outcomes of

the project, the pedagogical model that considers

how GenAI might affect teaching-learning

processes that need to be aligned with the values

of the different human stakeholders, and the

definition of research methods and instruments

for HL supported by GenAI-enhanced SLEs.

• WP2: GenAI-enhanced support for management

and design of learning. This WP includes tasks

CSEDU 2025 - 17th International Conference on Computer Supported Education

Figure 2: Evolution of the project following the DSRM phases in three cycles. Blue labels represented project tasks.

related to the proposal, development, and testing

of approaches aimed at sensing, analyzing, and

reacting for GenAI-enhanced support for teachers

and academic managers.

• WP3: GenAI-enhanced support for learning. This

WP includes tasks related to the proposal,

development, and testing of approaches aimed at

supporting the sensing, analyzing, and reacting

core functions of SLEs in HL that are relevant to

support learners considering the affordances of

GenAI.

• WP4: Technological Framework: integrated

infrastructure. This WP includes tasks related to

the selection of GenAI platforms and tools, and

platforms, tools, and devices in HL where GenAI

is integrated, the design of an architecture of a

GenAI-enhanced SLE for HL, the integration of

the proposed technologies in a prototype of a

GenAI-enhanced SLE for HL, and the proposal of

technical guidelines for the integration of existing

educational datasets in the proposed architecture.

• WP5: Pilot experiences. This WP includes tasks

related to the design of the evaluation plan, the co-

design, implementation, and evaluation of pilot

experiences, and the sharing of datasets.

• WP6: Coordination, dissemination, and data

management. This WP is transversal to the

GENIE Learn project and includes specific tasks

for coordination, dissemination, and data

management.

Figure 3: Structure of WPs of the project.

Figure 2 shows the adaptation of the DSRM

methodology for the specific case of the GENIE

Learn project considering the WPs and tasks to be

addressed. The project is structured into three main

iterations, one for each year. Each iteration follows

the three DSRM phases, outlining the tasks relevant

to each phase. Two phases are particularly crucial: the

initial phase, where the objectives are set at the start

of the project and for each cycle; and the evaluation

phase, especially important in the second and third

cycles when proposals are assessed.

5 CONCLUSIONS

The GENIE Learn project aims to enhance SLEs for

HL by incorporating GenAI tools in a manner that

aligns with the values and preferences of human

GENIE Learn: Human-Centered Generative AI-Enhanced Smart Learning Environments

stakeholders. The current problems presented by the

research context have been analyzed, considering the

human, technological and pedagogical contexts, the

affordances of GenAI, but also the risks and challenges

of this technology. The expected contributions of

GENIE Learn include: 1) a Research Framework with

the pedagogical model, human values, and scenarios

for HL supported by GenAI-enhanced SLEs; 2)

GenAI-enhanced approaches for management and

design of learning; 3) GenAI-enhanced approaches for

learning; 4) a Technological framework as an

integrated infrastructure; and 5) Pilot experiences co-

designed with educational stakeholders. The

complexity of the research context can only be

effectively addressed through projects of this nature,

supported by an interdisciplinary approach.

ACKNOWLEDGEMENTS

This work was supported by grants PID2023-

146692OB-C31, PID2023-146692OB-C32 and

PID2023-146692OB-C33 funded by MICIU/AEI/10.

13039/501100011033 and by ERDF/EU, project

GENIELearn. DHL (Serra Húnter) acknowledges

support by ICREA Academia.

REFERENCES

Akgun, S. & Greenhow, C. (2021). Artificial intelligence in

education: Addressing ethical challenges in K-12

settings. AI and Ethics, 1-10.

Albó, L., et al. (2022). Knowledge-based design analytics

for authoring courses with smart learning content.

International Journal of Artificial Intelligence in

Education, 32(1), 4-27.

Alonso-Prieto, V., et al. (2023). Studying the Impact of

Orchestrating Intelligent Technologies in Hybrid

Learning Spaces for Teacher Agency. In European

Conference on Technology Enhanced Learning (EC-

TEL) (pp. 636-641).

Amarasinghe, I., et al. (2021). Deconstructing orchestration

load: comparing teacher support through mirroring and

guiding. International Journal of Computer-Supported

Collaborative Learning, 16(3), 307-338.

Amarasinghe, I., et al. (2022). Teacher orchestration load:

What is it and how can we lower the burden? Digital

Promise and the International Society of the Learning

Sciences (pp. 1–13).

Amarasinghe, I., et al. (2023a). Orchestrating Hybrid

Learning Scenarios: Challenges and Opportunities.

In ISLS Annual Meeting 2023 (p. 72).

Amarasinghe, I., et al. (2023b). Generative Pre-trained

Transformers for Coding Text Data? An Analysis with

Classroom Orchestration Data. In European

Conference on Technology Enhanced Learning (EC-

TEL) (pp. 32-43).

Askell, A., et al. (2021). A general language assistant as a

laboratory for alignment. arXiv preprint

arXiv:2112.00861.

Baidoo-Anu, D. & Ansah, L. O. (2023). Education in the

era of generative artificial intelligence (AI):

Understanding the potential benefits of ChatGPT in

promoting teaching and learning. Journal of AI, 7(1),

52-62.

Brown, T., et al. (2020). Language models are few-shot

learners. Advances in neural information processing

systems, 33, 1877-1901.

Bubeck, S., et al. (2023). Sparks of artificial general

intelligence: Early experiments with gpt-4. arXiv

preprint arXiv:2303.12712.

Buckingham Shum, S., et al. (2019a). Learning analytics

and AI: Politics, pedagogy and practices. British

journal of educational technology, 50(6), 2785-2793.

Buckingham Shum, S., et al. (2019b). Human-centred

learning analytics. Journal of Learning Analytics, 6(2),

1-9.

Calvera-Isabal, M., et al. (2023). How to automate the

extraction and analysis of information for educational

purposes. Comunicar, 31(74), 23-35.

Cohen, A., et al. (2020). Hybrid learning spaces––Design,

data, didactics. British journal of educational

technology, 51(4), 1039-1044.

Crespi, F., et al. (2022). Estimating orchestration load in

CSCL situations using EDA. In 2022 International

Conference on Advanced Learning Technologies

(ICALT), (pp. 128-132).

Dagnino, F. M., et al. (2018). Exploring teachers’ needs and

the existing barriers to the adoption of Learning Design

methods and tools: A literature survey. British journal

of educational technology, 49(6), 998-1013.

Dai, W., et al. (2023). Can large language models provide

feedback to students? A case study on ChatGPT. In

2023 International Conference on Advanced Learning

Technologies (ICALT 2023) (pp. 323-325).

Delgado Kloos, C., et al. (2018). SmartLET: Learning

analytics to enhance the design and orchestration in

scalable, IoT-enriched, and ubiquitous SLEs. In Sixth

International Conference on Technological Ecosystems

for Enhancing Multiculturality (TEEM 2018) (pp. 648-

653).

Delgado Kloos, C., et al. (2022). H2O Learn-Hybrid and

Human-Oriented Learning: Trustworthy and Human-

Centered Learning Analytics for Hybrid Education. In

2022 IEEE global engineering education conference

(EDUCON) (pp. 94-101).

Demetriadis, S. & Dimitriadis, Y. (2023). Conversational

Agents and Language Models that Learn from Human

Dialogues to Support Design Thinking. In International

Conference on Intelligent Tutoring Systems (ITS) (pp.

691-700).

European Commission (2023), A European

approach to artificial intelligence. https://digital-

strategy.ec.europa.eu/en/policies/european-approach-

artificial-intelligence

CSEDU 2025 - 17th International Conference on Computer Supported Education

Evmenova, A. (2018). Preparing teachers to use universal

design for learning to support diverse learners. Journal

of Online Learning Research, 4(2), 147-171.

Friedman, B., et al. (2017). A survey of value sensitive

design methods. Foundations and Trends in Human–

Computer Interaction, 11(2), 63-125.

Gamieldien, Y. (2023). Innovating the Study of Self-

Regulated Learning: An Exploration through NLP,

Generative AI, and LLMs (Doctoral dissertation,

Virginia Tech).

Gargiulo, R. M. & Metcalf, D. (2010). Teaching in today's

inclusive classrooms: A universal design for learning

approach. Cengage Learning.

Gil, E., et al. (Eds.). (2022). Hybrid learning spaces.

Springer.

Gros, B. (2016). The design of smart educational

environments. Smart Learning Environments, 3, 1-11.

Garg, S., & Sharma, S. (2020). Impact of artificial

intelligence in special need education to promote

inclusive pedagogy. International Journal of

Information and Education Technology, 10(7), 523-

527.

Gutiérrez-Páez, N. F., et al. (2023). A study of motivations,

behavior, and contributions quality in online

communities of teachers: A data analytics approach.

Computers & Education, 201, 104829.

Hadwin, A., et al. (2017). Self-regulation, co-regulation,

and shared regulation in collaborative learning

environments. In Handbook of self-regulation of

learning and performance (pp. 83-106).

Hakami, L., et al. (2022). Exploring Teacher’s

Orchestration Actions in Online and In-Class

Computer-Supported Collaborative Learning. In

European Conference on Technology Enhanced

Learning (EC-TEL) (pp. 521-527).

Hernández‐Leo, D., et al. (2019). Analytics for learning

design: A layered framework and tools. British journal

of educational technology, 50(1), 139-152.

Hernández-Leo, D. (2022). Directions for the responsible

design and use of AI by children and their communities:

Examples in the field of Education, In Artificial

Intelligence and the Rights of the Child: Towards an

Integrated Agenda for Research and Policy, EUR

31048 EN, pp. 73-74.

Hernández-Leo, D., et al. (2023a). Editorial: Technologies

for Data-Driven Interventions in Smart Learning

Environments. IEEE Transactions on Learning

Technologies, 16(3), 378-381.

Hernández-Leo, D. (2023b). ChatGPT and Generative AI

in Higher Education: user-centered perspectives and

implications for learning analytics, In LASI Spain,

Madrid.

Hilli, C., et al. (2019). Designing hybrid learning spaces in

higher education. Dansk Universitetspædagogisk

Tidsskrift, 15(27), 66-82

HLEG-AI (High-Level Expert Group on Artificial

Intelligence) (2019), Ethics Guidelines for

Trustworthy AI. https://ec.europa.eu/futurium/en/ai-

alliance-consultation/guidelines/1

ISO 2013. International Standardization Organization

(2013), 16290:2013 Space systems — Definition of the

Technology Readiness Levels (TRLs) and their

criteria of assessment. https://www.iso.org/standard/

56064.html

Järvelä, S., et al. (2023). Advancing SRL Research with

Artificial Intelligence. Computers in Human Behavior,

147, 107847.

Johnson, R. B., et al. (2007). Toward a definition of mixed

methods research. Journal of mixed methods research,

1(2), 112-133.

Jovanovic, M. & Campbell, M. (2022). Generative artificial

intelligence: Trends and prospects. Computer, 55(10),

107-112.

Kalo, J. C., et al. (2020). KnowlyBERT-Hybrid query

answering over language models and knowledge

graphs. In 19th International Semantic Web Conference

(ISWC) (pp. 294-310).

Kasneci, E., et al. (2023). ChatGPT for good? On

opportunities and challenges of large language models

for education. Learning and individual differences, 103,

102274.

Koyuturk, C., et al. (2023). Developing Effective

Educational Chatbots with ChatGPT: Insights from

Preliminary Tests in a Case Study on Social Media

Literacy, In 31st International Conference on

Computers in Education (ICCE) (pp. 1-58).

Lim, W. M., et al. (2023). Generative AI and the future of

education: Ragnarök or reformation? A paradoxical

perspective from management educators. International

journal of management education, 21(2), 100790.

Long, P. & Siemens, G. (2011), What is Learning

Analytics? In 1st International Conference Learning

Analytics and Knowledge (LAK). ACM.

Michos, K. & Hernández-Leo, D. (2020). CIDA: A

collective inquiry framework to study and support

teachers as designers in technological environments.

Computers & Education, 143, 103679.

Misiejuk, K., et al. (2023). Changes in online course

designs: Before, during, and after the pandemic.

In Frontiers in Education (FIE) (Vol. 7, p. 996006).

Mizumoto, A. (2023). Data-driven Learning Meets

Generative AI: Introducing the Framework of

Metacognitive Resource Use. Applied Corpus

Linguistics, 3(3), 100074.

Molenaar, I. (2022). The concept of hybrid human-AI

regulation: Exemplifying how to support young

learners’ self-regulated learning. Computers and

Education: Artificial Intelligence, 3, 100070.

Ognibene, D., et al. (2023). Challenging social media

threats using collective well-being-aware

recommendation algorithms and an educational virtual

companion. Frontiers in Artificial Intelligence, 5,

654930.

Ortega-Arranz, A., et al. (2022). e-FeeD4Mi: Automating

Tailored LA-Informed Feedback in Virtual Learning

Environments. In European Conference on Technology

Enhanced Learning (EC-TEL) (pp. 477-484).

Ortiz Beltrán, A., et al. (2023). Surviving and thriving: How

changes in teaching modalities influenced student

GENIE Learn: Human-Centered Generative AI-Enhanced Smart Learning Environments

satisfaction before, during and after COVID-19.

Australasian Journal of Educational Technology,

39(6), 72–88.

Ouyang, L., et al. (2022). Training language models to

follow instructions with human feedback. Advances in

Neural Information Processing Systems, 35, 27730-

27744.

Peffers, K., et al. (2007). A Design Science Research

Methodology for Information Systems Research.

Journal of management information systems, 24(3),

45–77.

Pereira, F. D., et al. (2023). Towards human-AI

collaboration: a recommender system to support CS1

instructors to select problems for assignments and

exams. IEEE Transactions on Learning Technologies,

16(3), 457:472.

Pishtari, G., et al. (2024). Mirror on the wall, what is

missing in my pedagogical goals? The impact of an AI-

driven feedback system on the quality of teacher

learning designs. In 14th Learning Analytics and

Knowledge Conference (LAK) (pp. 145-156).

Pryzant, R., et al. (2023). Automatic prompt optimization

with “gradient descent” and beam search. arXiv

preprint arXiv:2305.03495.

Raes, A. (2022). Exploring student and teacher experiences

in hybrid learning environments: Does presence

matter? Postdigital Science and Education, 4(1), 138–

159.

Rose, D. H. & Meyer, A. (2002). Teaching every student in

the digital age: Universal design for learning.

Association for Supervision and Curriculum

Development.

Rudolph, J., et al. (2023). ChatGPT: Bullshit spewer or the

end of traditional assessments in higher education?.

Journal of applied learning and teaching, 6(1), 342-

363.

Ruiz-Calleja, A., et al. (2021). Supporting contextualized

learning with linked open data. Journal of Web

Semantics, 70, 100657.

Sánchez-Reina, R., et al. (2023). “Shall we rely on bots?”

Students’ adherence to the integration of ChatGPT in

the classroom. In International Conference on Higher

Education Learning Methodologies and Technologies

Online (HELMeT) (pp. 128-130).

Serrano-Iglesias, S., et al. (2021). Demonstration of

SCARLETT: A Smart Learning Environment to

Support Learners Across Formal and Informal

Contexts. In EC-TEL 2021 (pp. 404-408).

Serrano-Iglesias, S., et al. (2023). Exploring Barriers for the

Acceptance of Smart Learning Environment

Automated Support. In European Conference on

Technology Enhanced Learning (EC-TEL) (pp. 636-

641).

Siemens, G., et al. (2022). Human and artificial cognition.

Computers and Education: Artificial Intelligence, 3,

100107.

Sharples, M. (2023). Towards social generative AI for

education: theory, practices and ethics, Learning:

Research and Practice, 9(2), 159-167.

Susnjak, T. (2023). Beyond Predictive Learning Analytics

Modelling and onto Explainable Artificial Intelligence

with Prescriptive Analytics and ChatGPT,

International Journal of Artificial Intelligence in

Education, 34(2), 452-482.

Suraworachet, W., et al. (2021). Examining the relationship

between reflective writing behaviour and self-regulated

learning competence: A time-series analysis. In

European Conference on Technology Enhanced

Learning (EC-TEL) (pp. 163-177).

Tabuenca, B., et al. (2021). Affordances and core functions

of smart learning environments: A systematic literature

review. IEEE Transactions on Learning Technologies,

14(2), 129-145.

Tabuenca, B., et al. (2024). Greening smart learning

environments with Artificial Intelligence of Things.

Internet of Things, 25, 101051.

Theophilou, E., et al. (2023a). Gender‐based learning and

behavioural differences in an educational social media

platform. Journal of Computer Assisted Learning.

Early view.

Theophilou, E., et al. (2023b). Students’ generated text

quality in learning environment: effects of pre-

collaboration, individual, and chat-interface

submissions. In International Conference on

Collaboration Technologies and Social Computing

(CollabTech 2023) (pp. 101-114).

Theophilou, E., (2023c). Learning to Prompt in the

Classroom to Understand AI Limits: A pilot study, In

International Conference of the Italian Association for

Artificial Intelligence, (pp. 481-496).

Toetenel, L., & Rienties, B. (2016). Analysing 157 learning

designs using learning analytic approaches to evaluate

the impact of pedagogical decision making. British

journal of educational technology, 47(5), 981-992.

Topali, P., et al. (2024). Unlock the Feedback Potential:

Scaling Effective Teacher-Led Interventions in

Massive Educational Contexts. In Innovating

Assessment and Feedback Design in Teacher

Education (pp. 1-19). Routledge.

UNESCO (2023). Guidance for generative AI in education

and research. https://www.unesco.org/en/articles/

guidance-generative-ai-education-and-research

Villa-Torrano, C. et al., (2023). Towards Visible Socially-

Shared Regulation of Learning: Exploring the Role of

Learning Design. In 16th International Conference on

Computer-Supported Collaborative Learning (CSCL

2023) (pp. 289-292).

Wasson, B. & Kirschner, P. A. (2020). Learning design: Eu

approaches. TechTrends, 64, 815-827.

Yan, L., et al. (2023). Practical and ethical challenges of

large language models in education: A systematic

literature review. British Journal of Educational

Technology, 55(1), 90-112.

CSEDU 2025 - 17th International Conference on Computer Supported Education