Empirical Research

AI in a danish educational context

Roland Hachmann

Wednesday, April 24, 2024, the recommendations from the expert group appointed by the Danish Government regarding ChatGPT in relation to test and examination formats were released.

They can be found here: https://www.uvm.dk/aktuelt/nyheder/uvm/2024/april/240424-ekspertgruppe-klar-med-anbefalinger-for-brug-af-chatgpt-ved-proever

On Thursday, April 22, 2024, a more nuanced opinion piece arrived. The expert group suggests a paradigm shift and advocates considering fewer and different testing formats, not solely relying on written, reproductive, and individual assessments of students’ knowledge and skills. https://www.altinget.dk/uddannelse/artikel/medlemmer-af-ekspertgruppe-her-er-de-anbefalinger-vi-ikke-blev-bedt-om

This has triggered some thoughts that I would like to share here.

Not much new added

I hardly offend anyone (that’s certainly not my intention) by pointing out that in the recommendations and nuances, there’s not much new added to the table, but rather a reinforcement of something that has been pointed out for years – just with a different rationale than artificial intelligence. And maybe that’s fair enough since it’s not really the task of the expert group. Therefore, it’s particularly pleasing that they subsequently supplement with their other considerations – which were not commissioned by the Ministry of Education.

I also gladly noticed that the article in the Danish online news app Altinget is not just about ChatGPT and digital tools but, more broadly, about generative artificial intelligence. That’s a very important nuance. Artificial intelligence is much more than large language models – as the expert group also emphasises.

With language models in mind, it’s obvious that traditional testing formats no longer make sense. That collaboration, creativity, critical thinking, and communication skills are important is just as obvious. That tests should be based on a practical and student-oriented approach has been discussed since the early 1900s starting with Thorndike and colleagues on how learning transfers.

So, why hasn’t anything happened earlier?

Perhaps because the calculator, computer, internet, Wikipedia, and other technological developments gave us a greater sense of being in control than artificial intelligence does. Perhaps because now, it would be politically foolish not to do something about what has been pointed out for so long in education. Since the consequences of doing nothing would be obvious to everyone, including the public.

It has been said before, but again, it can’t be said enough. We need to rethink the school’s continued logic of industrialization, where instead of taming the world as if it were a wild bull, students are driven through steel gates to slaughter as if they were beef cattle.

At the same time, we might also need to rethink what we understand by life skills in our age. On the one hand, being able to understand and handle the digital layer surrounding us. And on the other hand, being able to emancipate ourselves from being dependent on it.

The school should encompass both.

Recent times with cybercrime, war, and pandemics have clearly shown the helplessness and panic that sneak into a population when technology fails or a minor or major crisis hits us.

One could briefly consider: What do we (as individuals and communities) do if we lose power for 2-3 weeks due to a super solar storm or an attack on critical infrastructure? Neither is as unlikely as we think, and the question is whether we are adaptable enough to handle this?

On a less existential level, smaller challenges such as the Chromebook issues in Danish schools from 2022 (and onwards) can create major concerns and almost paralyze teaching. The Danish Data Protection Agency’s restrictions and decisions regarding the limitation of Google Workspace in Schools led to statements like “We can’t teach without Chromebooks,”. Perhaps an exaggeration to emphasize a point, but also a symptom of how technology can create needs that are difficult to ignore.

Paradoxically, it could prompt the question: Do we want a school that becomes dependent on artificial intelligence and other digital solutions? Or a school that shirks its responsibility to develop versatile and cultured individuals who will navigate a world with these technologies?

So, it’s about balance!

“A teacher that could be replaced by Google – should be!” – A saying well-known in the education landscape in 2016. Could the same sentence be rewritten today with “ChatGPT replacing “Google”?

In any case, reflection is required regarding the balance between teaching and education, which requires human contact and learning that can be accessed through dialogue with artificial intelligence.

That the language models are imprecise, hallucinate, or don’t account for X or Y is only a temporary setback, not a lasting argument for human teachers. It’s just a matter of time before more and larger data and training sets are released and provided – then a large language model such as ChatGPT can provide a more precise and nuanced answer than any teacher or educator.

So, what kind of school/education and teacher/educator is necessary? This is the fundamental question that arises.

In light of the possibilities with artificial intelligence, the most immediate and banal answer is that it will be the school or teacher who focuses not primarily on knowledge and skills, but on relationships, humanity, empathy, adaptability, embodiment, and creativity.

This raises the question of whether our educational systems can handle this kind of school thinking when we also see the need to compare ourselves and live up to international test standards.

More skilled or lazier

As the expert group points out, large language models in education increase the need for students to be good at asking questions rather than providing answers. At the same time, one might add that students should also become adept at modelling questions about the world computationally and properly validating answers so that artificial intelligence is a help and enrichment for students’ activities in school.

The fear of cheating is real enough, but perhaps we should fear laziness even more in the long term. Not that we become lazy in the sense that we just lean back; on the contrary, one can imagine that we now must accomplish even more in less time since AI can assist us in solving different tasks more efficiently. No, lazy in the sense that we no longer need to think, ponder, remember, and concentrate – because artificial intelligence entices us with quick answers and solutions. Neuroscience and brain science researchers have long pointed out that digital technologies have consequences for the aforementioned brain functions. That our ability to remember is closely related to bodily experiences and memories thereof. So, prompting AI to do our thinking tasks poses some risks of making our brains lazy.

Therefore, the developing tests and evaluation formats should embed AI as a tool and be based on the students’ situational contexts. AI can be a powerful tool in idea creation and in helping to aggregate, organise and summarise some forms of knowledge. However, solving real-world human problems in contexts dependent on action-based solutions requires humans.

What the future holds is uncertain. As the researcher and futurist Roy Amara once said, “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.” I do think that we need to think carefully about AI in education. With thought on the works of Joseph Weizenbaum, there are things AI can do that we do not want it to do.

Thoughts

Five Didactical Perspectives on Computational Thinking

Roland Hachmann

Introduction

This post contributes with five perspectives on how to didactically adress progression in the work with Computational Thinking (CT) in teaching.
I take on here a perspective of CT, that was revitalized and popularized in 2010 by Jeanette Wing, who defined the term as:
“The thought processes involved in formulating problems and their solutions so that the solutions are represented in a form that can be effectively carried out by an information-processing agent” (Wing, 2010).

Wings take can be seen as problematic in different ways. However, this will not be treated in-depth here.
Although Wing’s formulation is somewhat abstract, special emphasis is placed on the fact that these are cognitive problem-solving strategies, where students model the world through specific forms of representation that are based on informatics and computer science as disciplines. Wing emphasizes that it is about thinking like a computer scientist and that the special modelling competence inherent in the disciplines can be generalized across all subject disciplines in the same way as reading, writing and arithmetic.

With this point of departure, it is the aim here, to provide teachers with perspectives for didactic reflections on CT as a specific way of working in the subjects. It is, therefore, not an attempt to provide instructions, but more to point to points of attention on planning.

Five perspectives:

What follows are five didactic perspectives that point out important aspects in the student’s development of CT as a problem-solving strategy. The five perspectives are:

1. The pre-computational perspective
2. The bodily perspective
3. The abstract perspective
4. Perspectives on problems and their solutions
5. Perspectives on the creative and imaginative

I do not claim this to be a model of progression, but the five points reflect a movement from the concrete towards the abstract. Although students, to a certain degree, possess the ability to think abstractly in different ways, they often come with little or no prerequisites for computational modelling. Therefore, an important and overall point of attention here is to create opportunities for the students to be gradually introduced to computational strategies and methods in interaction between the tangible/concrete and the symbolic/abstract.

The pre-computational perspective

Working computationally involves a basic understanding of symbol manipulation and modelling. This requires the students to become aware that e.g. coding and programming are based on certain kinds of logic and terminologies. The students must learn a special of computational grammar (certain syntaxes) before they can begin concrete computational work. For example, abilities within mathematical logic, decomposition and pattern recognition are part of the prerequisites students must possess in order to work computationally.
This can be taught in many ways, but a well-known and well-tested method for older children is to have them solve mathematical puzzles. One of many examples is Cut Hive Puzzles (http://inabapuzzle.com/),  which I explain in a previous post

Although the computational historically has its origins in mathematics, CT is not only limited to mathematical problem-solving. However, by working with logic tasks, students learn basic principles that can later be transformed into situated computational methods. For example, writing programs consist of certain logical rules and patterns.

The pre-computational perspective must be seen as an indication that students do not necessarily have the necessary prerequisites to think computationally in the sense of a computer scientist, but that this, along with reading and arithmetic, must be trained and developed. In the CT literature, coding and programming are often compared to professional skills that are similar to complex mathematical problem-solving in the natural sciences, and high-level literature analyses in the language subjects. Therefore, here too, the students need a basic understanding that precedes the computational.

The bodily perspective

A more concrete and less abstract approach to understanding and implementing CT in relation to a professional task can be done through physical activities. Activities such as “program your friend” or “bodygramming” could be examples of such approaches. Programming a friend to execute an algorithm can help show how precisely one needs to formulate rules if the computational agent (the friend) is to execute it correctly and the same way every time. The bodily perspective makes it concrete and visible how the algorithm is performed and whether this is done correctly. In such a process, students can isolate and correct the places where things go wrong and discuss how, for example, IF/THEN can be formulated more precisely.

The Finnish researcher Jussi Mikkonen proposes “bodygramming” as a bodily method to teach students programming. Bodygramming means that students physically behave like a computer program, through step-by-step prescribed actions. In this way, the students get an experience of the synchronous processes that are connected to a code at a low, human pace. In this way, an alternative option is offered to understand basic programming concepts and abstractions.
The bodily perspective points out the need for CT to be concrete for the students. These two examples (Program a friend and Bodygrammin) can make programming visible and concrete for students in a way that makes it tangible and debatable. In this way, it is ensured that CT does not just become something that takes place as part of abstract thought processes, but also something that physically unfolds in the world. At the same time, this perspective can shed light on how we as humans (as opposed to computers and AI) partly build our understanding of phenomena on contextual interpretations and intuition.

The abstract perspective

CT embeds a perspective that is based on students’ ability to think abstractly about phenomena in the world and translate these into ways that can be processed computationally and rule-based.

With the Cut Hive example above in mind, this means a move from being able to solve this puzzle to writing down rules for its solution.

A simple example is:

IF <a given input>
THEN <a specific action>

IF <a hexagon with area 2 contains the number 1 or 2>
THEN <the second hexagon contains the missing number>

Writing down rules for programs includes abstractions, generalizations and pattern recognition. Students must be able to exclude other parts of the puzzle in order to simplify and generalize, while at the same time comparing the generalized rules with the rest of the game’s rules (pattern recognition).

An example of the rewriting of a concrete event into a general rule:

The example shows how the abstract perspective in Wing’s definition of CT can play out. What distinguishes this approach to problem-solving in school from others is that it involves thought processes that seek to reduce the complexity and interpretation possibilities of phenomena in such a way that a machine (which cannot make decisions based on intuitive interpretations and emotions) can solve tasks, that would be too difficult or take too long for humans to do. The abstract perspective thus points to the fact that there are specific ways of thinking with special purposes that students must learn.

The problem-based perspective

In most contexts, CT is related to problem-solving of complex problems that cannot be handled by humans alone. In Wing’s formulation, it is also explicitly mentioned that it is both about formulations of problems and solutions to them. Not all problems are relevant nor solvable with the implicit methods alone.
Often, the problems that are sought to be solved are closely connected with and take their place in the subject. However, there are some fundamental characteristics of computational problems in non-computer science subjects:

1. That data is collected and processed (analysed).
2. That an algorithm is created that can help solve the problem
3. That the algorithm can be executed by a computer or human
4. That the problems have multiple or often open solutions

As hinted at academic problems are not tied to the field of informatics or data science. Poems are, for example, full of patterns (e.g. rhyme, metric, syllables) that can be transformed into rules and classifications in relation to recognizing and categorizing other poems. Another approach in language subjects could be to write four sentences and let the students put them in a logical order based on the data the sentences contain.
The problem-based perspective points to a need to reflect on which problem types are suitable for CT and what level and complexity the problem has in relation to the students’ prerequisites. Working with analyzing poems will be difficult for beginners because it requires skills that are based on complex knowledge of genre and language. Putting four sentences together so they make logical sense based on a content analysis would be easy for a high-school student.

The creative and imaginative perspective

Although Wing, in her definition, is particularly concerned with the problem-solving perspective, there are examples in the literature, that CT is involved in more than that. There are also arts and creative perspectives associated with designing solutions to problems or expressing oneself and imaginaries through coding. In such a perspective, creativity is part of shaping something from a set of conditions. Conditions here could mean the fact that algorithms or algorithmic processes are included in the design itself and/or in the product in one way or another.
The German professor Yasmin Kafai, who is, among other things, is one of the developers behind the coding platform Scratch, has in recent years worked with young people’s design of electronic textiles as a special way of expressing themselves. Kafai emphasizes that this way of working with the computational offers special opportunities for young people to have a critical-constructive voice and participate democratically through freedom of expression. The students’ thoughts, feelings and attitudes are expressed, when they design and create different products from textiles that are combined with microcomputers such as Lillypads (see e.g. https://www.exploringcs.org/e-textiles). The creative perspective thus points out that CT in teaching is not always only associated with problems but can also support the creative and soft forms of interpretation.

Concluding remarks

In the recent literature in Denmark, CT is seen as part of a movement focussing on empowering children to be able to take a critical stand towards digitization and the use of the media. In an educational context, one could ask, what possibilities and limitations does the computer as a tool entail when it is involved in the solution of problems and what do students need in order to use them? In this light, a CT competence makes good sense as a modelling competence that, through certain methods, enables the student to transform concrete specific problems into something a computer can help to solve. However, it is also important that teachers consider the following:

  • What CT is not. Is it, as Wing imagines, a transversal and general competence that reaches into, but also beyond, the subjects? Or is it just one of many methodological tools that students must have in their toolbox?
  • How is CT reconciled with the subject’s already built-in logics?
  • To what extent does CT contribute to the students’ general education?

Teaching is complex and the questions here show that the five perspectives highlighted here are not exhaustive, but just specific points of attention that can be included and discussed in relation to the teacher’s other general and subject-didactic reflections.

Thoughts

Few thoughts on math puzzles and computational problem-solving

Roland Hachmann

Recently I read the book The Power of Computational Thinking: Games, Magic and Puzzles to Help You Become a Computational Thinker by Paul Curzon and Peter W. McOwan. I got me thinking about math, problem-solving, and computation. Not least what it takes for young children to grasp some of the concepts, that is involved in computational thinking.

Working computationally involves a basic understanding of symbol manipulation and modeling. This requires that the pupils become aware that e.g., coding and programming are subject to certain kinds of logic and terminologies. The pupils must learn special grammar (certain syntaxes) and think in specific ways before they can begin concrete computational work. For example, abilities within mathematical logic, decomposition, and pattern recognition are part of the prerequisites pupils must possess to work computationally.

These can be taught in many ways, but a well-known and well-tested method is to have pupils solve mathematical puzzles. One of many examples is Cut Hive Puzzles (http://inabapuzzle.com/), which in short, are about having a pattern (cubes) where some walls are marked with thicker lines. There are two rules:

  1. Each marked area must contain numbers from 1 to the maximum number of cubes in the marked area (4 in the example below).
  2. The same number may not appear within the same marking, and the same number may not appear in two cubes that are in contact with each other.

https://teachinglondoncomputing.org/cut-hive-puzzles/ 

By working with this type of task, the pupils learn basic principles that can later be transferred to computational methods, such as the fact that writing programs consist of certain logical rules and patterns.

Although computation historically has its origins in mathematics, CT is not limited to mathematical problem-solving. The pre-computational perspective considers that pupils do not necessarily have the necessary prerequisites to think computationally but that this, along with reading and arithmetic, must be trained and developed. In the CT literature, the abilities to code and program are often compared to literacy skills that are needed for complex mathematical problem-solving in the natural sciences and high-level literature analysis in the language subjects. Therefore, pupils need a basic understanding that precedes the computational.

CT embeds a perspective that is based on the pupils’ ability to think abstractly about phenomena in the world and translate these in ways that can be processed computationally and rule-based. With the Cut Hive example above in mind, this means going from being able to solve a puzzle to writing down rules for solving it.

A completely banal example is:

IF <a given input>
THEN <a specific action>.IF <a hexagon with area 2 contains the number 1 or 2>
THEN <the second hexagon contains the second number>

Writing down rules for programs includes abstractions, generalizations, and pattern recognition. Pupils must be able to exclude other parts of the puzzle to simplify and generalize while simultaneously comparing the generalized rules with the rest of the game’s rules (pattern recognition).

The example here shows how the abstract perspective is fundamental in CT.

What distinguishes this approach to problem-solving in school from others is that, in most cases, it is a question of thought processes that seek to reduce the complexity and interpretation possibilities of phenomena in such a way that a machine (that cannot make decisions based on intuitive interpretations and emotions) can solve tasks that would be too difficult, or take too long for humans to complete.

See more here: https://teachinglondoncomputing.org/cut-hive-puzzles/

Book Chapter, Publication

New publication on Multimodality

Roland Hachmann

Today I received my copy of the new handbook “Erhvervsdidaktisk opslagsbog” [Roughly translated: “Handbook for Vocational School” Didactics].
Together with my good colleague Dorthe Carlsen, I wrote a piece on “Multimodality” and multimodal learning design.
The book contains 58 chapters from 44 authors covering the fundamental concepts and issues concerning the vocational education.
I hope you will enjoy reading it.

Book, Poster Presentation

CT for All?

Roland Hachmann

Together with a good colleague – Stig Børsen Hansen – I presented a poster on “Computational Thinking in Language Arts”. The poster session was held online at the ISLS (International Society of the Learning Sciences) Annual Conference 2022.

Click on image to get the PDF

Stig Børsen Hansen also wrote a really nice piece on Philosophers of Technology unfolding the understandings of technology by philosophers Langdon Winner, Albert Borgmann, Karl Marx, Herbert Marcuse, Martin Heidegger, John Dewey, and Bruno Latour.

Book Chapter

THE CYBER WEAPON

Roland Hachmann

Computational Literacy in Scenario-Based Teaching Practices

Today I contributed with a chapter for a new Danish handbook on Scenario-based Teaching.
The purpose of the chapter is to contribute to the field of scenario-based teaching and computational literacy as an emerging field in primary school. I do this by exemplifying a concrete and tested scenario fragment where the students solve an educational escape puzzle through analysis and manipulations of technologies. The activity forms the starting point for a close investigation of students’ computational problem-solving strategies, as a focal point in computational literacy. The analysis is based on a socio-cultural view seeing tangible artifacts as mediating tools for actions unfolding in the given practice. From here the chapter deals with computational issues and the significance of artifacts as mediators for scenario-based activities that have computer literacy as part of the subject domain. A specific emphasis here is on the opportunities for socio-epistemic negotiations, the tangible artifacts creates. On a more practical level, the article seeks to deliver a well-described example that can inspire practitioners and support pedagogical decisions when designing for computational literacy.

The book is in danish, and can be found here:
https://www.saxo.com/dk/haandbog-i-scenariedidaktik_bog_9788772196442

Empirical Research

Poster Session – Fablearn Europe 2022

Roland Hachmann

This year I had the pleasure of attending Fablearn Europe with a poster presentation of my work. The presentation highlighted the potential of Educational Puzzles in Computer Science education. Drawing on some of my empirical work, I highlighted how tangible computational things, hold the potential for learning the basics of problem-solving, decomposition, and algorithmic thinking.

Click the image to get the PDF

Empirical Research

Design activity #3: Final prototypes

Roland Hachmann

In today’s design activity, we finished the first prototype and readied it for the first testing in school. We will present the prototype to the teachers who have volunteered to test the game in their classrooms.

Decisions and discussions on materiality
Lene and I have been discussing the materialities of different solutions, and whether the first prototypes should be made in e.g., cardboard or wood, the consistency in the graphical expression of the single pieces of the game, the correlation between the analog and digital part of the game and how to balance the game to both constrain the students into a specific fictional genre and afford openness to let the student’s imagination and creativity flourish.

Our considerations on materialities are also based on which kind of feedback we want to get from the teachers. For us, this concerns the balance between presenting a prototype, that looks done but not as a completely finished design. We don’t want the teacher to be concerned with the time and work put into the prototypes and thereby eventually limiting their evaluation. our hypothesis here is, that if the teachers think we have too much ownership or put to much work into it, then that will limit their evaluation and honesty towards concerns or critique of the design.

Lene checking the quality of the pieces and preparing them for the final finish
Design Experiments, Empirical Research

Design activity #2: First prototyping

Roland Hachmann

After sketching the raw ideas of the game, my colleague (Lene Illum Skov) and I went through the first initiating design workshop. The aim of the workshop was to get closer to which elements we wanted to include in the game and how these would serve a specific purpose in regard to supporting the students’ creation of interactive stories.

The approach chosen in the project borrows its methodology from the field of participatory design research as described by Spinnuzi (2005):

As the name implies, the approach is just as much about design—producing artifacts, systems, work organizations, and practical or tacit knowledge—as it is about research. In this methodology, design is research. That is, although participatory design draws on various research methods (such as ethnographic observations, interviews, analysis of artifacts, and sometimes protocol analysis), these methods are always used to iteratively construct the emerging design, which itself simultaneously constitutes and elicits the research results as co-interpreted by the designer-researchers and the participants who will use the design.

Lene cutting and glueing pieces of the story to fit with specific tokens of the game

With that in mind, the design workshop is part of the research and the conversation unfolding during that process of great value for the choice made.
During the workshop, some important choices were negotiated, e.g.:

1) How did we want the tokens and pieces of the game to look?

2) Which text parts would be the best to include, keeping the storyline as open as possible?

3) Which materials would suffice and be suited to the job?

4) Simplicity in the expression of the artwork

Playfulness and game activities provide what Eva Brandt calls “dream material” (Halse, Brandt, Clark & Binder, 2010) that supports participants in playing out different versions of futures and outcomes. The process of designing the game could be seen as such, providing both Lene and me with opportunities to see meaningfulness and purpose while being engaged in developing ideas and artifacts.

The final outcome of the day. Not finished – but one big step closer