Tag: GenAI for Teachers Series

Attention and Its Implications in Teaching

The ‘new science of education’ made a significant impact on teacher education after Dylan Wiliam’s tweet in 2017. Since then, it has gained immense momentum, sparking debates on social media and within academia. Attention has become the golden currency of this new science, rapidly achieving the status of a gold standard. You may recall the viral video of Pritesh Raichura, which ignited widespread discussion in the sector and seemed to provoke a Marmite response—you either love it or hate it. So why has attention become such a hot topic and led to such extreme reactions? To answer this, we need to examine its concerning connection to neurofeedback training—also known as brain training—and explore how GenAI could play a pivotal role in the system.

Until the last ten years or so, cognitive science was largely kept outside of the educational domain. But in 2023, in response to growing interest, the Core Content Framework (CCF) was introduced for initial teacher education. The introduction of the Core Content Framework (CCF) and the soon-to-be ITTECF brought with it a requirement for all providers to educate trainees on the theory and application of cognitive science. The more I work with trainees and explore the fundamentals of cognitive load theory (Sweller, 1988), the more I recognise that attention is the key hinge point. The connection between what the teacher feels they ‘do’ and what the learner ‘does’.

Attention is a complex field of study, one that the CCF and ITTECF heavily simplify. However, a clear aim emerges: attention is treated as a commodity to be maximised in order to achieve long-term information retention. I propose that attention serves as the connecting mechanism for two otherwise predominantly closed systems: the external world (the classroom) and the internal world (the learner). In this light, attention could be likened to the role of gravity in brane theory: it permeates and connects otherwise distinct systems. Its effects are felt in and across each system, but the connection is not always tangible or observable.

Generative AI offers new possibilities for supporting this dynamic. By gathering real-time data from classrooms and operationalising advancements in neurofeedback technology, GenAI could provide personalised insights into students’ attentional patterns. AI-driven adjustments could ensure that key moments of instruction align with students’ peak cognitive receptivity, thereby enhancing memory retention and engagement. But is this ethical? How does this relate to the free will of the pupil? Is this the responsibility of the teacher?

Navigating Challenges and Ethical Considerations

While the integration of GenAI and neurofeedback tools offers exciting possibilities, educators must address the accompanying ethical concerns. Monitoring and measuring attention could shift the focus from cultivating student autonomy to prioritising compliance and control. If attention becomes a commodity to be technologically quantified, are we at risk of reducing students to data points and Fourier transforms?

A major concern is the commercialisation of attention-focused devices. As highlighted in a recent paper by Kotouza, D., Pickersgill, M., Jessica Pykett, & Williamson, B. (2025), neurofeedback technologies and EEG devices have already been marketed for use in classrooms, with some targeting “students with high numbers of disciplinary ‘office referrals’” to improve their concentration. How should educators respond when faced with the growing pressure to adopt such tools? What safeguards are needed to prevent data exploitation, bias in AI algorithms, and breaches of privacy?

Attention should be cultivated as both a skill and a choice. Using the marriage of neurofeedback and GenAI systems has the potential to remove or heavily reduce this choice. The UN Convention on the Rights of the Child (UNCRC) states that every child has the right to:

• Relax and play
• Freedom of expression
• Be safe from violence
• An education
• Protection of identity
• Sufficient standard of living
• Know their rights
• Health and health services

There are multiple aspects of this list which are at risk through the use of systems that aim to maximise attention.

So, where’s the line between control exhibited by a teacher, and control exhibited by a machine? Pritesh’s video maximises attention skilfully – how does this compare to a neurofeedback-GenAI system which effectively does the same?

We are currently in a pivotal period of change and ethical debate – what does it mean to be a teacher in this context of neurofeedback and GenAI and how can these tools enhance education without compromising the agency, privacy, and humanity of learners?

In conclusion, the convergence of neurofeedback, GenAI, and education offers both promise and peril. As educators, we must critically evaluate whether these technologies truly serve the learner or merely reduce them to data-driven outputs. The potential to enhance attention and engagement is undeniable, but so too is the risk of undermining student agency and the foundational principles of education. 

The question remains: how can we harness these advancements responsibly, ensuring they enhance teaching rather than challenging humanity at its core? As we grapple with this pivotal moment in educational evolution, the choices we make today will shape not only the future of teaching but the very essence of what it means to learn.

Is GenAI socially constructed? As it is trained on a large subset of the internet, I’d argue that it is. I’d go so far as to say the concept of GenAI itself has parallels with a giant socially constructed machine such as Viki in I, Robot. Teachers have user profiles for these GenAI tools and accounts, so by using them, they are, by definition, subscribing to the beginnings of a Teacher GenAI Identity. What could the diversity of these identities look like? How well do they map to the classroom? How representative or relational are they to the identity observed in classroom practice? These are all questions that we are just beginning to explore. This aligns well with the idea of a teacher identity being ‘dynamic’. Would a teacher identity survive without a GenAI aspect? At what point will there be any teachers left who don’t use GenAI in some way or another? This human-machine way of interfacing could be seen as the ultimate ‘hybrid’ component of teacher identity.

It seems likely that a GenAI Identity could indeed fulfil the criteria of being socially constructed, dynamic, and hybrid, but what about the other aspects? How can GenAI be used with real-life teacher narratives? A classroom is full of narratives. A teacher has their own narratives. Arguably, this is such a nuanced and personally-loaded concept that it is difficult to see how it could be mirrored by the digital domain. I’ve spent some time trying to deconstruct this concept of narratives, especially for teachers early in their training or careers. A narrative is context-dependent and value-laden, and awareness of them requires advanced reflective capacity. I’ve discovered something, though, with prior knowledge scenarios. Our prior knowledge generator is perhaps the best way to illustrate this. While we are using the ‘new science’ within the CCF/ITTECF, in our profession we will be focusing on establishing the prior knowledge of children for some time to come.

Prior knowledge is more than just prior learning, though. It is contextual and dependent on lived experience and personal significance. These are difficult aspects for beginning teachers to master, as their lived experience and database of what prior knowledge could look like for children is in its infancy. So our open-source prior knowledge scenario prompt generator aims to provide a scaffold to promote the consideration of the narratives of the individuals in a class. In this capacity, although the scenarios may not be exactly representative of the specific needs of the class, they encourage a teacher to consider the diversity of the possible prior knowledge of the children in their class. A scaffold for one size does NOT fit all approach.

Continuing with the idea of individual differences, how can we view a Teacher GenAI Identity as being affected by emotion or relationships with others? I’ve observed that the mere fact that there is the capability of GenAI being used to do ‘teacher’s work’ evokes an emotional response in individuals, such as those who are actively resistant to AI in education (Shea, 2024). How can a team of colleagues function, in the future age of GenAI adoption, if there are huge differences in the response to its use? How do departments and institutions address this, and is there even a need or requirement to? Is an early adopter of GenAI Teacher tools at a social advantage? Are they perceived as holding a position of responsibility to influence the GenAI identities of others in a team? In this regard, I could argue that a Teacher GenAI Identity is not just a personal attribute but one that is socially impacted and relevant to teams of employees and their functionality.

Even before GenAI tools burst into our pedagogically-informed lives, we pondered on how to judge the ‘goodness’ of a lesson plan. There is no definition for a ‘good’ lesson plan, as essentially this concept could be broadly thought of as dependent upon:

• Context (the needs of your specific learners)
• Teacher identity (your epistemological beliefs about knowledge etc.)
• School ethos (what values does your school promote?)
• Policy (what frameworks do you have to work within?)

When asked to define a ‘good’ lesson plan, most teachers struggle to condense their thoughts into a formula or set of beliefs. This is something that educationalists intuitively understand but may find challenging to operationalise. Yet in my experience, it is something student teachers and ECTs commonly yearn for.

Enter the Lesson Inspector; a tool we’ve developed to evaluate lesson plans based on the themes present in the Core Content Framework (CCF), Initial Teacher Training Core Content Framework (ITTECF), and the Teachers’ Standards. Our Lesson Inspector offers two metrics: a quantitative score and qualitative evaluation/analysis by criteria. This qualitative analysis provided by Lesson Inspector offers potential for content alignment of the LLM with the aforementioned Statutory frameworks.

In this discussion, we use two lesson plan generators (Aila and Teachmate) to produce lesson plans on the same topic of KS4 waves, with the same objective (below). We then upload these to Lesson Inspector and create Inspection reports for them both. We will discuss:

• User Process: The steps and methodology each tool uses and how the user inputs their lesson information.
• Report content : Such as prior knowledge and other relevant themes included in the Lesson Inspector report.

For a fair comparison, both tools were given the same objective from the AQA (2015) GCSE Physics Specification: “Students should be able to describe wave motion in terms of their amplitude, wavelength, frequency, and period.”

Erm, great but…what is ‘content alignment’?

You could be forgiven for having missed the introduction of this term  in the GenAI context. We are using the definition from AI for Education, as can be seen in the screenshot adjacent. Essentially the terms ‘correctly calibrated’ and ‘curriculum standard’ found within this definition could (and should) be debated. Here, we will introduce a ‘curriculum standard’ (the CCF/ITTECF and Teachers’ Standards – although the latter is an assessment framework and not a curriculum) and discuss the notion of being ‘correctly calibrated’.

But first, let’s look at how these two interfaces are used to produce the lesson plans, what is produced, and what Lesson Inspector provides as reports.

Lesson Inspector Analysis (Reports)

Both the lesson plans were put through our Lesson Inspector. See below for the .pdf versions of the reports (we are working on making these look a lot prettier!).There are several pages to each.

Discussion on the Lesson Plans and Reports

Starting from the same objective, it’s interesting to see the paths that each tool has taken. This echoes what anyone in teacher ed will have noticed: there can be a huge diversity in planning, given the same initial boundary conditions. In some sense, it could be argued that lesson plan generation tools should mimic this. Certainly the probablistic nature of GenAI should mean that by definition a vast array of lesson plans should occur.

That brings us on to considering how useful, or correct this diversity of plans could be.

Many will wonder if the Inspector could score a perfect zero. We’d invite you to consider the (very brief) lesson plan below, and amuse yourself with the Inspector’s report…

Fire Dragon Report PDF

Activity Ideas Generator

One of the options to ‘refine my answer’ was to ‘differentiate’. Once we had recovered from shock of the use of the ‘dirty word’ of ITT, we couldn’t help but click on it and see what was produced. Compare the two below. Perhaps the educational use of the word ‘differentiation’ (which has varied somewhat over the last 20 years) hasn’t been used by the model, and a more generalised definition of the word has been applied instead?

Concept Explainer

It produced a three page explainer. Being scientists, we are sad to see it hasn’t actually pulled on any of the wealth of information that is out there, for example the Institute of Physics ‘IOP Spark‘ Platform (which we REALLY hope gets integrated with GenAI very soon, hint-hint nudge-nudge!) or anything from the Association for Science Education and their database. Another criticism is that it’s written with a severe lack of technical vocab, or showing any link with prior and sequential ideas on forces. It’s no match for the great plans that Plan Assess have created without the use of GenAI. On the plus side, it’s a start if a teacher has no experience or existing perception?

When prompted for refining the answer, we just HAD to click on the ‘add in some common misconceptions’ option. The second document below contains the amendments. Again, it’s a start for teacher to remember to consider misconceptions, but it’s missed an opportunity to signpost readers to the sources of information, where they can deepen their subject knowledge.

Jingle Maker

We are all for the arts and science combined. But perhaps a somewhat underwhelming jingle on friction….from TeachMateAI here?!

So, we asked Copilot to make a jingle on friction, for comparative purposes. Indulge yourself with its response!

Summary

Lets face it, this is good for free. It’s an introduction. Somewhat missing the opportunity to link to subject specific services and platforms for development of subject knowledge. The functionality is somewhat clunky – it ignores many requests typed into the ‘refine my answer’ box without indicating that it can’t help you. Definitely not hitting the adaptive teaching mark or ITTECF criteria.

Let’s pull out the pros and the cons:

Pro’s:

• These tools were free
• Gets teachers thinking about misconceptions and scaffolding their materials
• Fun little jingles

Cons:

• Depth of subject knowledge and misconceptions is missing.
• No links to subject-specific information sources or platforms for teachers to improve the depth of their subject knowledge or subject pedagogy
• The concept of differentiation is contestable and not aligned to the ITTECF.
• No integration of SEND awareness, and not useful for adaptive teaching.
• Ignores ‘refine my answer’ requests.

So, in summary let’s say: Free, basic, shallow, introductory.

If anyone has paid for the full version and would like to add to this, please contact us!