AI+ED = Artificial Intelligence + Education

The Basics 

As in the article titled "Digital Innovation in Schools," this text, in its original form, dates back to about a year ago (August 2023) and is designed for Italian schools. A practical guide given the new school year and the PNRR funds for school digitalization. Since it remains relevant, I thought it might be useful to share!

In recent years, there has been increasing attention towards Artificial Intelligence (AI) technologies and their applications, which have become increasingly common on the e-commerce sites we use for our purchases, on streaming platforms, as well as in various sectors from healthcare to agriculture, from industry to tourism, and so on. This attention peaks with announcements from major tech companies regarding new products being released on the market and is experiencing what we can call a new hype starting in 2023, particularly after the spread of generative AI tools (Gemini, Claude, ChatGPT, Midjourney, etc.).

As with all technological innovations, the question of how and whether to introduce these tools into teaching is becoming increasingly pressing: What makes sense to teach students about these technologies? Which ones to choose? How to use them? What critical elements should we consider? And, no less important, how do the needs we are called to address in terms of skills and knowledge to equip students during the educational process change so that they can become citizens of the context in which we live? These are just some of the questions that arise when dealing with this topic. In this document, we will try to offer a brief overview of the field and suggest some useful guidelines to follow, while being aware that this is a matter that requires continuous adaptation and is of great complexity.

Premise

A Multiple Approach As is often the case in media education, when we talk about AI, we must distinguish between two approaches to bringing these tools into teaching.

On the one hand, there is a curriculum issue, and on the other, there is the issue of using AI for teaching other subjects. Although theoretically addressable separately, these are two closely related aspects that are worth tackling in an interconnected manner.

To these two, we can add the issue of the impacts that the spread of AI has on our society, including in terms of life and work, and, consequently, how the school should possibly be reconsidered in light of the skills needed today and tomorrow.

We will address all three briefly.

AI Curriculum

As with all media education topics, even in the case of AI, we can imagine two paths: creating an ad hoc curriculum and officially introducing a new school subject in the grades where it is decided to be included (disciplinary integration) or addressing the topic broadly across all (or almost all) taught disciplines (cross-disciplinary integration). In the Italian case at this stage and considering the cross-disciplinary nature of AI technologies, the reasoning can only develop under the second hypothesis. This means that AI competence and basic knowledge should be as common as possible among all teachers, in addition to being the object of a structured reciprocal coordination that allows achieving objectives in the best way possible.
A fundamental aspect if we want to address the AI topic in the classroom will be computational thinking, which is distinct from coding education—an essential element—since it aims to understand the application of tools in real-world scenarios, rather than delving into the internal workings of these tools. A very useful resource, also a partner of the Italian "Programma il Futuro" (Future Program), is Code.org, where materials in Italian are available for both coding and AI topics.

For a complete setup of a possible curriculum to bring into the classroom, it is possible to use, although with few resources translated into Italian, the site of the initiative https://ai4k12.org and the many resources proposed within the framework, which is based on five foundations that can be explored in various ways.

The poster illustrating them is available in Italian as well as some of the listed resources that can be found in Italian, for example, on the http://www.code.org website.

AI Tools for Teaching

As mentioned in the introduction, the topic of AI in schools also opens the issue of educational tools that can be experimented with to support the teacher's activities as well as to support students.

AIEd Developments and Applications

For a few decades now, a separate discipline has emerged that deals with this topic and is called AIEd (Artificial Intelligence in Education). From the early developments of AI, there has been a strong interest in the possible applications in the educational field, particularly in exploring the idea of increasingly personalized pathways for each student: AI was seen as a possible adaptive tool that could respond to this ambitious goal. Moreover, the discipline has been of particular interest to various fields of study that have contributed to its development since the 1960s.

It was an interesting field for AI researchers because AI and education share three fundamental pillars:

• learning,
• forgetting,
• and changing.

For cognitive scientists who saw education as the ideal field in which to conduct their research on psychological phenomena, and for educational researchers who aspired to better understand the mechanisms of human learning through computational ones.

Since then, this new discipline has taken shape and has addressed various issues over time, generating multiple experiments on possible applications. It is important to note, therefore, that there is a rich history of studies that can be useful in understanding what the promises of this field are, what to possibly revisit and update with the new possibilities offered by AI advancements, and how to avoid the mistakes of the past.

The key point that emerges from retracing the results of this discipline is that only technological products rooted in and permeated by a pedagogical context can offer solutions that are not only effective but, under certain conditions, more effective than traditional methods. In the absence of a structured pedagogical project, we are faced with interesting tools that do not meet the real objectives of the educational context in which they are used.

After this important premise, however, we can certainly highlight the fact that there are many promising applications in this area, and it is worth following the path of prudent and responsible experimentation if we want to respond to the challenges posed by contemporary times in educational terms.

The areas of application of this discipline can be divided according to the following scheme (Holmes and Tuomi, 2022):

FOCUS ON THE STUDENT (further divided between study support and actual teaching)

This area includes all those solutions designed for use directly by students independently. In particular:

• Intelligent Tutoring Systems (ITS).
• AI-assisted apps for language learning or scientific subjects.
• Simulations that combine AI with applications of virtual or augmented reality and with a game-based approach.
• Tools for supporting individuals with special needs.
• Tools for automatic essay and text writing.
  Chatbots.
• Tools for automatic assessment.
• Open exploratory environments.
• Automatic agents to whom students can teach.
• Automatic assistants and tutors that support students in planning their educational projects.

FOCUS ON THE TEACHER

In this area, we find tools that support teachers in improving classroom management, planning and managing educational activities, as well as personalizing learning. In this area, we find:

Curating educational materials.
Classroom monitoring.
Automatic assessments.
Automatic assistant to support the teacher's activities.
Classroom management.
Plagiarism detection.

FOCUS ON THE INSTITUTION

In this area, there are tools that aim to support educational institutions or universities in activities such as:

• Planning educational activities.
• Security.
• Identifying students at risk of dropping out.
• Selection and admission.
• Electronic supervision of tests and exams.
  
• Quickly scrolling through this taxonomy, one can immediately notice how some areas of application present numerous ethical and privacy challenges. Moreover, many of these are still in a research phase that requires waiting to demonstrate the actual effectiveness of the proposed solution.

The most promising opportunities in our opinion at the time of writing this guide lie in a few specific areas:

• Support for teaching planning for teachers.
• The possibility, respecting privacy as this solution requires the collection of numerous data, to identify and support students at risk of dropping out.
• Support for students with special needs and,
• more generally, the possibility of personalizing learning according to individual needs.

These last two points are particularly interesting due to the renewed and improved conversational interaction possibilities offered by generative AI.

A Proposed Approach

We can now try to outline a proposed approach to using existing tools and the many new ones we will have at our disposal in the future. First of all, we can recall, as Stefania Giannini, Assistant Director-General for Education, did in a document published in July 2023, that any technology to be evaluated as a teaching tool must meet certain quality standards that can serve as a sort of checklist for us:

• Content accuracy
• Age appropriateness
• Relevance of pedagogical methods
• Cultural and social appropriateness, including checks to protect against bias
  

UNESCO has also recently published a study on digital teaching, highlighting that many of the technologies experimented with so far do not have their roots in pedagogy but in technical aspects and that they are not always applied with the right rules, resulting in neutral or potentially negative impacts, while when the digital integration project starts with verifying these aspects, the impact is positive.

The same will apply to AI.

Another aspect to consider is the ethical one: what do we give up if we decide to use AI for certain activities? And if we decide not to use it? What impacts does it have on the system and on students?

To navigate this, we can follow a framework proposed by philosopher Luciano Floridi, who invites us to reflect on four parameters that we apply to this context:

• Feasibility: Is what the system under evaluation proposes humanly feasible? For example, if we imagine providing all Italian students with a specific AI system to support their learning activities, we are in the realm of feasibility and not fantasy, whereas if we imagined having a technology that could fully address the learning needs of all humans, we would be imagining an unrealizable product.
• Sustainability: In environmental and social terms, how can we assess the sustainability of the choice? For example, a system might consume too much with a negative impact on the environment that is not worth facing in light of the benefits it offers.
• Acceptability: If we apply a certain standard, is there a risk of having negative social impacts? Do we risk perhaps losing cultural diversity or excluding certain categories of people with special needs or difficulties with the Italian language?
  Preferability: But above all, is what we are deciding to adopt preferable? Is it something we would choose?
  

If we can pass all these barriers, the product or service we are evaluating can be adopted, and it makes a lot of sense to do so.

It is important to note that the way we implement our project is also relevant: one thing will be to propose to students a project with generative AI (for example, using ChatGPT, Gemini, Claude, or others), closely monitoring the use made of it in terms of privacy settings and inputs provided to the tool to critically evaluate its functioning and learn to navigate these technologies; another will be to completely replace, for example, the assessment methods in a discipline with an AI system without human intervention.

A Practical Framework

There are many ethical frameworks for AI and many specifically for AIEd. A fundamental document is the Beijing Consensus on Artificial Intelligence and Education of UNESCO (2019), which establishes some fundamental principles later developed in a set of recommendations in 2021.

Here, we find it useful to suggest the use of the framework produced by the work of the Institute for Ethical AI in Education, which allows identifying a series of application principles for the assessment of risks and opportunities.

In particular:

1) Educational objectives at the centre

Any application must be based on educational objectives and evaluated and monitored accordingly.

2) Augmented assessment

The use of AI should aim to expand the parameters verified.

3) Administration and workload

AI should serve to reduce workloads and support people while respecting human relationships.

4) Equity

The application must promote equity among different groups of students and must never discriminate.

5) Autonomy

It is important through these systems to increase the level of control over their objectives and learning activities and the autonomy of students.

6) Privacy

The right level of privacy must be sought for all users.

7) Transparency and responsibility

The system must be at least scrutable to allow its users (teachers and students) to understand at least a general level of the system's functioning mechanisms and their implications.

8) Informed participation

All users must be adequately informed (and trained).

9) Ethical design

The entire design of the system and the way it will be used must be based on ethical principles.

In general, when using AI tools, it will be important to implement continuous and repeated monitoring of these aspects: algorithms evolve and can undergo changes over time, and their impacts could vary.

Teacher and Student Training

As highlighted by point 8 of the proposed frameworks, a fundamental aspect is that of (in)formation of users. UNESCO presented in September 2023 a draft for two frameworks on the skills to be developed in the field of AI for teachers and students, available at this link, in addition to extremely useful and essential work with the guidelines on the use of GenAI in the education and research sector. 

In general, we believe that some key aspects can be listed to be included in teacher training to promote responsible and informed experimentation with these tools:

• Innovative teaching and digital teaching
• Specific AI topics:
  • Fundamentals and functioning.
  • Ethics.
  • Security and privacy.
  • AI culture.

New Skills?

The Center for Curriculum Redesign has addressed the question of how to rethink the school in a world where AI applications have been emerging in recent decades and has well-schematized the type of change we are experiencing with this image, where the time dedicated to transmitting knowledge and skills is overturned due to the emergence of search engines and AI:

Thus, without entering into a discussion concerning specific subjects, we are invited to focus mainly on the issue of methodology: to concentrate our efforts on developing an understanding that serves as a springboard for the development of expertise and transferability skills. The way we engage students cannot be simply transmissive and must focus primarily on a process rather than a product to be rewarded with a grade.

In the words of Harvard's Next Level Lab, "An education strongly centred on products reduces learning to a transaction, exchanging a product for a grade, rather than providing a transformative human experience. Learning is much more than generating a product; indeed, the essence of learning is in the process: the journey rather than the destination. Learning to write does not primarily mean producing a well-structured piece of text, but developing the ability to organize one's ideas, connect them to those of others, analyze statements, synthesize insights, and fulfill our fundamental need to communicate and learn from others (Chiang, 2023).

Similarly, learning a new language is much more than being able to speak in that language; bilingualism is about learning another style of thinking, adopting a different cultural identity, and embodying an alternative way of being and living in the world (Mills & Moulton, 2017)."

LINKS Curriculum Area

• https://programmailfuturo.it/ [coding and computational thinking]
• For lower secondary school: https://ammagamma.com/lucy/
• https://machinelearningforkids.co.uk/
• https://AI4k12.org
• https://www.curiositymachine.org/get-started/

AIEd Applications Area

• Beijing Consensus On Artificial Intelligence And Education: https://unesdoc.unesco.org/ark:/48223/pf0000368303
• Ethical AI In Education Framework: https://www.buckingham.ac.uk/wp-content/uploads/2021/03/The-Institute-for-Ethical-AI-in-Education-The-Ethical-Framework-for-AI-in-Education.pdf
• UNESCO AI Competence Framework: https://www.unesco.org/en/digital-education/ai-future-learning/competency-frameworks
• UNESCO Artificial Intelligence In Education. Compendium Of Promising Initiatives: https://iite.unesco.org/publications/ai-in-ed-compendium-of-promising-initiatives-mlw-2019/

SKILLS AREA

• https://nextlevellab.gse.harvard.edu/2023/07/28/navigating-a-world-of-generative-ai-suggestions-for-educators/

OTHER SOURCES CONSULTED

Handbook of Artificial Intelligence in Education
https://curriculumredesign.org/

NOTES

[1] Stefania Giannini, 2023, Reflection on generative AI and the future of education. UNESCO 2023

[2] https://www.unesco.org/gem-report/en

[3] https://unesdoc.unesco.org/ark:/48223/pf0000368303

[4]https://www.buckingham.ac.uk/wp-content/uploads/2021/03/The-Institute-for-Ethical-AI-in-Education-The-Ethical-Framework-for-AI-in-Education.pdf

[5] https://curriculumredesign.org/

[6] https://nextlevellab.gse.harvard.edu/2023/07/28/navigating-a-world-of-generative-ai-suggestions-for-educators/