AI technology and teaching Archives

Can we preserve human agency in a world of AI?

That’s a question we can all ask ourselves as we interrogate the UN International Day of Education. This year’s theme is AI and education. What does the teaching profession gain and lose with Gen AI?

Two hundred million people use ChatGPT each month, with growth doubling in one year. A recent article from Harvard Business Review correctly identifies that generative AI (Gen AI) is a prediction machine that can summarise, synthesise, code, and draw based on its training with the corpus of knowledge from the internet and custom data sets. The article points out that: “The efficacy of predictions is contingent on the underlying data. The quality and quantity of data significantly impact the accuracy of AI predictions…. (T)he successful implementation of AI requires good judgment… It involves knowing which predictions to make, the costs associated with different types of mistakes, and what to do with the AI’s outcome… Judgment over what matters in a particular situation is fundamental to the successful use of generative AI.”

Time to ask those questions about AI again

Anecdote and research suggest that students in schools and universities increasingly use Gen AI tools in various ways to undertake learning and assessment. There has been a flurry of activity by government, state departments and regulators in providing policy, guidelines and resources for educators and students on the technology. Discourse has seemed to turn a corner from using Gen AI as “cheating” to either adjusting assessment by having students apply or adapt AI text to the real world or embracing AI outputs by critiquing or improving on them.

Two years after the widespread uptake of ChatGPT, the most popular Gen AI tool, I think it is time to pause and re-ask ourselves as educators what exactly do my students need to know and be able to do to demonstrate competency in learning. This question is the true core of curriculum. And it goes directly to thinking creatively or innovatively in education.

There is a lot of talk about Gen AI augmenting human intelligence with its efficient summary outputs. As the Harvard Business review article points out, such outputs require “good judgment” in order to assess the quality for a “particular situation”.

Still a place for old-fashioned exams

Educators can certainly set tasks where students generate such outputs and develop skills to assess output quality. Of course this means explicitly teaching those quality assessment skills (research, information literacy, critical thinking) and then having some way of knowing if students are using/developing the skills without using Gen AI to produce a fake trail of skill development. This might involve seeing drafts of work with commentary on how the skills were used with real time presentations on this coupled with teacher and peer questions. There is still a place for old fashioned exams too as one way to assess knowledge acquisition and transfer, as unpopular as this may be in some AI evangelical circles.

If we want students to be more than adept prompt jockeys, then we have to really think about how we want them to demonstrate learning.

Software that purports to provide AI generator matches are pretty hopeless and give warning about this, so teachers shouldn’t rely on these but on carefully developed dialogue and iterative processes with students. In other words, carefully crafted learning and assessment activities and knowing their students well. This is easier in schools than in universities where large cohorts, online learning, intensified academic workloads and a highly casualised workforce act as barriers to developing genuine, long term educator-student connections.

On standards

Now, let me unpack the issue from my context as a teacher educator. Australian teachers have a set of standards they need to meet at various career stages. The curriculum of teacher education needs to directly respond to these standards and teacher education is commonly structured according to : (1) content (discipline) knowledge, (2) method which is curriculum, pedagogy and assessment, (3) understanding learning and the learning context of students (educational psychology and sociology), and (4) how 1-3 translate into practice through professional experience known as practicums and internships. Summarising and synthesising from the corpus of the internet, Gen AI can easily produce outputs for assessment related to areas 1- 3.

Teachers are great sharers. It is a human-centred, collaborative profession after all. For a long time teachers everywhere have shared curriculum scope and sequence documents, unit and lesson plans, assessments, teaching resources, and student work samples online. Student teachers, usually referred to as pre-service teachers, have a vast repository of exemplars and resources to draw on, modify and use for assessment at university and at practicum. Plagiarism checkers could and can still identify if a student has directly copied something from the internet and not cited the source or tried to pass it off as their own.

Questioning the quality of the AI output

Gen AI, drawing on all the teaching curriculum resources available online, can almost instantly produce scope and sequence, units of work, lesson plans and resources such as work sheets, by predicting what the user wants according to the prompt and synthesising or summarising what is available online. It is then up to the pre-service teacher or teacher to make judgements about the quality of the AI output in relation to the task or the appropriateness for the learning context.

Teachers who have gone through traditional method courses at university – learning to first read a syllabus for structure and meaning, and then translating this into a lesson plan, a unit of work and a scope and sequence through a carefully scaffolded developmental arrangement of courses across a degree – are mostly well equipped to make professional judgements about automated outputs from gen AI. However, we are entering a new era where it may be possible to produce work for discipline, method and learning courses without having to think critically or authentically about what is submitted for assessment.

There will be a sizeable proportion of students graduating from universities who would have relied on Gen AI outputs in an expedient or shallow way to get through their degree having been exposed to limited opportunities that “test” depth of understanding, application and transfer and creative or innovative thinking. Universities won’t want to talk about this for a long time – just as they were slow to address the impact of essay mills. But it will be a phenomenon which will shape trust in higher education institutions and ultimately professions.

In teacher education this could mean a heavier burden for teachers supervising students on practicum. In the world of Gen AI these supervising teachers are well placed to evaluate whether a student has developed competency through their application of discipline, curriculum and pedagogy, and learner knowledge.

There are many, many teachers using Gen AI to generate curriculum material, school reports, newsletters and other artefacts considered ripe for an efficiency overall in their time-poor day. If Gen AI were to cease tomorrow, I would hazard a guess that the vast majority could still create these texts as they have gone through the sequenced training prior to and in-service, and have experience to draw on, including the experiences of other teachers.

However, we may be entering an era where there will be the first cohorts of teachers who have come to rely on Gen AI to a point that they did not develop these skills or the necessary judgement vital in designing curriculum to suit context. Gen AI raises a lot of questions related to professional knowledge and standards.

Will pre-service and practising teachers develop AI dependency? Will this erode the unique combination of professional skills teachers have? Does this matter? Should we augment our competencies and intelligence and redefine the fundamentals of professional knowledge?

AI: it’s about what exists, not what’s possible

Finally, what will happen to innovation in curriculum design if pre-service and in-service teachers slowly stop drawing on their vast cognitive resources to create and share new unit plans or teaching resources, instead relying on the quick Gen AI fix? We need to remember that Gen AI is a summarising and synthesising tool, predicting a response from a prompt to communicate what already exists not what is possible.

Let’s start having a more serious and sustained conversation in teacher education and the teaching profession about what we gain as educators in using Gen AI and what we potentially erode, lose or irrevocably change, and will it matter for our students?

To return to my original question but orienting it towards the training of pre-service teachers – what exactly do pre-service teachers need to know and be able to do to demonstrate competency with and without Gen AI? This question surely goes to the heart of teaching standards.

Erica Southgate is an associate professor in the School of Education, University of Newcastle. She makes computer games for literacy and is an education technology ethicist and an immersive learning researcher.

Six reasons Artificial Intelligence technology will never take over from human teachers

The next twenty years will see teachers under increasing pressure to convincingly justify their existence. Advances in artificial intelligence (AI) technologies are already prompting calls for teaching to be automated, learner-driven and ‘teacher-proof’. While these technologies might still require non-specialised classroom facilitators and technicians, the role of the highly trained expert teacher is coming under increasing threat. There is a growing sense that “we don’t really need teachers in the same way anymore”.

Put bluntly, the entire premise of ‘the teaching profession’ faces an impending challenge. In a future where education can be reliably provided by machines, why continue to invest millions of dollars in training human experts to do the job? Given the likely trajectory of technological developments over the next few decades, is there anything that an expert teacher does that machines will never be able to do? As an education researcher and teacher, I would like to think that there is! Here, then, are six aspects of expert human teaching which are getting overlooked in the current rush toward automating the classroom:

1. Human teachers have learned what they know

There is clear benefit from being with someone who can pass on knowledge, especially someone who has previously been in the position of having to learn that knowledge. This latter qualification is a uniquely human characteristic. When a learner learns with an expert teacher, they are not simply gaining access to the teachers’ knowledge but also benefiting from the teachers’ memories of learning it themselves. Technology can be pre-loaded with content of what is to be learned. Yet, no AI technology is going to ‘learn’ something exactly the way a human learns it, and then help another human learn accordingly.

2. Human teachers make cognitive connections

A human is uniquely placed to sense what another human is cognitively experiencing at any moment, and respond accordingly. In this sense, face-to-face contact with a teacher offer learners a valuable opportunity to engage in the process of thinking withanother human brain. On one hand, there is something thrilling about witnessing an expert who is modelling the process of thinking things through. Conversely, a human teacher is also able to make a personal ‘cognitive connection’ with another individual who is attempting to learn. As David Cohen puts it, teachers are uniquely able to “put themselves into learners’ mental shoes”. Despite the best efforts of computer science, many aspects of thinking cannotbe detected and modelled by machines in this way.

3. Human teachers make social connections

Teaching is a mutual obligation between teachers and learners. No teacher can stimulate the learning process without the cooperation of those who are learning. Good teachers make personal connections with their students, helping them gauge what might work best at any particular time. Before attempting to intellectually engage with a group, teachers will “take a mental pulse of students’ demeanours”. Teachers work hard to establish this mutual commitment to learning, as well as sustaining engagement through motivating, cajoling and enthusing individuals. All of these are interpersonal skills that come naturally to people rather than machines.

4. Human teachers talk out loud

There is something transformative about being in the presence of an expert teacher talking about their subject of expertise. Listening to an expert talk can provide a real-time, unfolding connection with knowledge. A good speaker does not stick rigidly to a written text, but refines, augments and alters their argument according to the audience reactions. A teacher speaking to a group of learners therefore engages in a form of spontaneous revelation. Key to this is the teacher’s role in leading and supporting learners to engage in active listening. As Gert Biestareasons, being addressed by another person interrupts one’s ego-centricism – drawing an individual out of themselves and forcing them into sense-making.

5. Human teachers perform with their bodies

The bodies of human teachers are an invaluable resource when engaging learners in abstract thought. Teachers use their bodies to energize, orchestrate and anchor the performance of teaching. Many subtleties of teaching take place through movement – pacing around a room, pointing and gesturing. Teachers make use of their ‘expressive body’ – lowering their voice, raising an eyebrow or directing their gaze. Crucially, a human will respond to the living biological body of another human in a completely different way to even the most realisticsimulation. Being looked in the eye by another person is a qualitatively different experience than being looked at by a 3D humanoid robot, let alone a 2D cartoon agent on a screen.

6. Human teachers improvise and ‘make do’

A key part of good teaching is the human capacity to improvise. Rather than sticking tightly to a pre-planned script, teachers will adjust what they do according to the circumstances. Like most performative events, teachers approach a session with a rough plan or structure. However, thereafter they improvise their way around these aims and objectives. Teaching requires acts of creativity, innovation and spontaneity – akin to dancing or playing jazz. Teachers and learners feel each other out, find common ground and build upon it. Teaching also demands a tolerance for imprecision, messiness and not knowing. Most human actions involve a degree of guesswork, bluff and willingness to ‘make do’. These are processes that computer systems are largely incapable of.

As these examples illustrate, an expert human teacheris able to support learning in ways that can never be fully replicated through technology. Unfortunately, these qualities remain largely unrecognised, even by teachers themselves. Many educators consider teaching to be an ‘unconscious’ act that is difficult to pin down and articulate. Yet such coyness does little to dispel the technology-driven arguments currently being made against the teaching profession. Teachers need to speak up and make an irrefutable case for the continued presence of expert professionals at the forefront of classrooms.

So how can we rehabilitate human teachers in the minds of their detractors? The uphill battle in countries like Australia is to revitalise schools and classrooms to allow teachers to work in the ways just outlined. These are all characteristics that a good teacher should have, but are considerably restricted in an era of ‘teaching out-of-field’, templated lesson plans and rigid standardised testing.

A first step in this direction might be to alter the ways that people think and talk about teaching. Teachers need to speak forcibly about these qualities – amongst themselves, within their professional associations, withparents, politicians, pundits and anyone else with influence. Teachers also need to argue directly against the tech industry and corporate reformers looking to replace them with machines. There is obvious value in the human expert teacher. Yet unless teachers are able to make a convincing case, they may well lose the argument before they even realise that there was one.

Neil Selwyn is a professor in the Faculty of Education, Monash University (Australia). He previously worked in the UCL Institute of Education, and Cardiff School of Social Science (UK). Neil is currently writing a book on the topic of robots, AI and the automation of teaching. Over the next six months he will be posting writing on the topic, hopefully resulting in: Selwyn, N. (2019) Should Robots Replace Teachers? Cambridge, Polity.

Neil can be found on Twitter @neil_selwyn