Making teachers more productive?
Unpacking the Productivity Commission's recommendations on lesson planning, AI and edtech - featuring a little lesson planning experiment with Claude
I’ve now been back in education policy land for two and a half years after starting my career at the CIS in 2013. In between, I’ve oscillated between teaching at high school and doing policy and advisory work, including working as a political adviser.
This history has given me a slightly unusual outlook on education-related matters, as I entered the teaching profession asking questions about things like effectiveness and efficiency, which teaching colleagues sometimes found quirky (at best) or off-putting at worst.
But my coalface experience also means I’m often sceptical about how much governments and government policy can really do to change the way the profession operates. This approach does not always sit well in policy land, particularly in political advising. Why use finite political capital and policy energy to get contentious high-level reforms across the line if you can’t ensure the benefits of those reforms will be able to be felt within the timeframe of the political cycle?
So I have a certain amount of sympathy for those in the policy space trying to make recommendations about education, particularly when they have been tasked with couching these in the context of Australia’s productivity problems.
When educators can’t even agree on what the desired ‘outputs’ of education even are are (or, worse, imply it’s morally suspect to talk about the noble endeavour of education in economic language) it becomes difficult for the policy boffins at organisations like the Productivity Commission to come up with something that fulfils their brief but is also realistic for the sector.
Now that the final recommendations relating to lesson planning, AI and edtech in education have been published by the Productivity Commission in their Building a skilled and adaptable workforce report, it’s worth reflecting on these matters further.
First, some context
I’m going to make an assumption here that many of the people reading this are teachers, or teaching-adjacent (e.g. school or system leaders), so it may be news that the Productivity Commission is even making recommendations on education matters.
The genesis of the report was the desire by the Albanese government and Treasurer Jim Chalmers to address concerns about productivity growth, and Chalmers flagged human capital development - and the role of education and training in that - as a key direction:
Lifting productivity is about empowering workers and making the most of our human capital.
Teaching and training them to adapt and adopt technology.
Unlocking innovation, investment and dynamism to lift the potential of our people and their economy.
Hence, one of five reports focuses on workforce. Working backwards means what happens in schools matters.
Now, ‘making education better’ is a recipe for a very broad and diluted inquiry, so as best as I can understand it the PC’s approach was to narrow it down to look at the productivity of teachers.1 In this frame, what teachers ‘produce’ is student academic outcomes, and we want these to increase/improve.
At this point, it’s critical to clarify the oft-misunderstood idea of ‘productivity’, which most people just associate with ‘working harder’. Imagine a big red X through the image below.
Productivity is not about increasing inputs (such as time and money), it’s about finding ways to increase outputs (what students learn) without increasing inputs. In other words, what can help teachers teach more effectively and efficiently, by working smarter - not harder?
If that’s the frame, then it makes sense that the PC chose to focus on lesson planning (where the Grattan Institute’s Lesson Lottery work was very effective at making the case for collective lesson planning leading to better use of teachers’ time) and AI/EdTech, as adoption of tools and tech in most other sectors is typically associated with smarter use of inputs.
The interim report and the CIS response
Typically the PC produces an interim report based on one round of submissions and consultations, and then opens that interim report to further submissions, after which they produce a final report. It was in response to the interim report and its draft recommendations that the CIS produced a submission.
For context, these were the initial recommendations in the interim report to which the submission responded:
Draft recommendation 1.1 Invest in a single national platform for all teachers to access lesson planning materials
Draft recommendation 1.2 Lead national efforts to ensure equitable access to educational technology (edtech) and artificial intelligence (AI)
The whole submission is available to read here, but the main arguments are captured below.
Written curriculum quality is foundational
A knowledge-rich, coherent, well-sequenced written curriculum is a necessary precondition for long-term improvement in student outcomes and equity, and any enacted curriculum resources can only be effective if they are aligned to a high-quality written curriculum. In an environment where the existing Australian Curriculum leaves much to be desired in this respect, and divergence in curricular approaches across jurisdictions, attempts to layer a national bank of quality-assured resources on top of inconsistent written curricula are unlikely to succeed.
Also, never mind the fact that the jurisdictions disagree! It would be almost impossible to get agreement on what good curriculum looks like from the teaching bodies and subject associations which would undoubtedly hold significant sway over what the parameters are for any quality-assured lesson bank. Instead, we recommended the government undertake a comprehensive benchmarking of the Australian Curriculum and its state variants against evidence and high-performing systems, and use this to inform the next curriculum review.
We need a deeper market for curriculum
The move towards more centralised curriculum provision makes sense when school-level responsibility for developing enacted curriculum increases teacher workload and contributes to inequity. In this context, system-produced lesson resources can help, because this would enable teachers to focus on intellectual preparation for teaching, not on assembling activities.
On the other hand, whether this helps with teaching quality (and not just workload reduction, or redirection) depends on whether curriculum resources and lesson plans are embedded in coherent scope-and-sequence frameworks based on what students need to know and when, rather than being isolated activities. The Victorian Lesson Plans project for instance, was undertaken first as a workload reduction measure and has led to the odd situation where lessons were produced prior to curriculum sequencing and provision of appropriate scope and sequence documents.
Against this backdrop, national quality assurance of whole-class curriculum resources is a nice idea, but its ability to impact teaching quality is contingent on stronger written curricula, a deeper resource market, and agreed evidence-based criteria (see my point above about the difficulties of such a thing). When we still have teachers defending to the death their rights to ‘autonomy’ then conversations about quality assurance seem premature.2
Learning interventions are where quality assurance approaches can still play a role
Learning intervention decisions are described as high-stakes for schools: they involve significant cost to deliver, opportunity cost for schools and students (including student learning time), and the bar for success is set high, as they require students to make accelerated progress relative to their peers.
As research and evaluations from CIS and many others have found, teachers currently receive little reliable guidance when selecting interventions, despite national policy emphasis on tiered and targeted supports. Past large-scale intervention efforts have produced limited evidence of impact, reinforcing the need for better evaluation and decision-making support for schools.
Given interventions have narrower criteria for effectiveness, there is a greater likelihood of achieving the consensus necessary to develop effective criteria for quality assurance, especially when interventions can be trialled in the field and compared with each other. Any future QA body for teaching resource should initially focus on interventions, ideally sitting within a future Teaching and Learning Commission and be linked to monitoring national attainment targets.
EdTech and AI adoption require clarity of educational purpose first
The PC’s recommendations are based on a borderline fanciful assumption that technology will lead to better resource allocation in teaching, when there is little evidence to support the existence of a shared understanding of effective instruction across the teaching profession.
In my reckoning, there is no way that pursuing AI/LLM-generated lesson plans, or training teachers to generate these, or developing AI tools to help teachers generate these is something where the juice is worth the squeeze.
LLMs are just predictive models that vary based on the quality on the input. If there are established ways of doing things out on the interwebs, then an LLM is just going to reflect that framing back in terms of what it produces. I’m sure it has its uses, but a profession where lots of people still believe in learning styles (as my recent research has found) is probably not in a great position to apply quality professional judgement to evaluate what these tools produce.
A little experiment in LLM lesson planning
To round off the submission, I attached an appendix where I asked Claude to generate a lesson for me for Year 8 Humanities on feudalism using “evidence-based practices to promote student learning” and to explain why specific activities had been used. Have a look at the whole thing (at the end of the submission) because I include some brief commentary about why it, frankly, stinks.
Besides some obvious problems (like the curriculum links being a hallucination and the time allocations being deeply unrealistic), the thing that jumped out to me the most was how many terms and phrases that have some basis in good practice were used: activating prior knowledge, gradual release of responsibility, working memory, formative assessment were all in there.
But the specific activities which were being discussed had no real relationship with deeper ideas about the science of learning, the most prominent of which is that the development of a lesson needs to be orientated around a knowledge objective, not activities.
Let’s look at a few examples, starting with the ‘hook’/starter activity (first 10 minutes of a 60 minute lesson):
1. Hook Activity (10 minutes) Activity: "Medieval Status Scramble"
• Students receive cards with medieval occupations (king, knight, serf, lord, merchant, priest)
• Working in small groups, they arrange these in order of importance/power
• Groups share their arrangements and justify their choices
• Teacher records different arrangements on the board
Teaching Strategy: Activate prior knowledge and create cognitive conflict when groups disagree
This was included because:
This activates students' prior knowledge and creates cognitive dissonance when groups disagree about social hierarchies. Research shows that beginning lessons with what students think they know, then challenging those assumptions, leads to deeper learning. The activity also taps into students' natural curiosity about social status and power.
Firstly, research shows no such thing about how allowing students to dwell on misconceptions before correcting those misconceptions supposedly leads to deeper learning, and there is no evidence students have a ‘natural curiosity’ about status and power.
Secondly, for an activity that is supposedly about activating prior knowledge, this product does not care what students should already know - it just assumes students will have awareness of those terms, when a word like ‘serf’ is practically unheard of outside of a handful of topics, none of which we could expect a Year 8 student to be familiar with unless they have been taught this in some way. You can just imagine an average kid putting the king at the top and saying “I dunno” to the rest and continuing to chat about whatever brainrot they’ve seen on TikTok for the remaining 9 minutes and 30 seconds.
But the lesson looks like it covers some of the main ideas about feudalism, and there’s lots of room for engagement and students doing different things, and the teacher does some assessment at the end via a thumbs activity and exit ticket, and doesn’t it mention using a graphic organiser to assist working memory?
Used like this, off the shelf LLMs are likely to just multiply the quantum of edu-bullshit making its way into Australian classrooms, not reduce it. Bureaucratic buzzwords like ‘quality assurance’ and ‘alignment’ are unlikely to be terribly helpful because the real problem is the absence of deep professional knowledge across the whole profession - but that’s a whole other soapbox.
The final recommendations
Recommendation 1.1 Establish national frameworks to assess the quality of teaching resources
Recommendation 1.2 Improve student outcomes by providing equal access to high-quality instructional materials
Recommendation 1.3 Improve student outcomes by providing better access to educational technology
Recommendation 1.4 Provide professional development to support teachers to implement best practice [in edtech and with instructional materials]
As we can see, the final recommendations are a little less ambitious in some ways than those in the interim, and arguably more ambitious in others.
It’s positive they recommended that frameworks need to be established for quality before a body is set up to provide quality assurance, and making the mission of high quality instructional materials clear and direct rather than implied is positive. So too is the acknowledgement that professional development is needed to support best practice.
But what sits underneath these headlines also matters. What’s the point in the Australian Government producing a full suite of AC-aligned resources when a) they would be better off fixing the AC and b) this pretty much already exists thanks to Ochre and other bodies.
As for the Australian Government doing more on edtech, give us a break. Hasty adoption of edtech has caused far more problems than it has solved and the last thing that’s needed is the imprimatur of the federal government - so far removed from the realities of classroom teaching - providing more encouragement in that direction. The same goes for professional development - if state governments, who run teacher registration and accreditation, can barely ensure everyone is getting quality PD then how will the Australian Government?
The PC’s recommendations seem so out of kilter to me because the fact is there are several things the Australian Government can do to make teachers more productive, and many of those policies are already on Jason Clare’s agenda. I’d have liked to be a fly on the wall for the conversation between the Education Minister and the Treasurer after this final report was published, to be honest.
The most important thing for Clare at this juncture is to hold the line on key policy reforms that are completely within the remit and power of the Australian Government, namely, the quality assurance around initial teacher education programs and holding states and territories to account for their progress towards BAFSA (Better and Fairer funding agreement) targets, particularly in improving the performance of low-achieving students through quality teaching.
But I suppose ‘keep doing what you’re doing, but properly’ probably isn’t the stuff that the Treasurer’s big-talking productivity reform process had in mind.
Other work by the PC has tried to look at the productivity of other elements, such as spending, in the review of the National School Reform Agreement
This recent post from one of my former colleagues Shaun Brien is a great reflection on how the devotion to autonomy instead of low-variance curriculum causes myriad problems in schools