Explaining Reggio, Montessori, and “Academic” Philosophies

If you are choosing preschools for your child, you are probably bombarded with names of different philosophies. Reggio Emilia, Montessori, Waldorf, “academic”, “progressive”, Bank Street, “play-based”, “child-led” – the list goes on and on. As a quantitatively-minded person, you’re probably wondering – what are the decomposable principal components that define them and make them different? These philosophies can overlap and diverge, but here is the question that underpins differing educational thought:

How much do you want to optimize for efficiency vs self-discovery?

And following from that one main question, the bases emerge:

1. What is the objective? How much does efficiency of skill acquisition vs the generalizability of self-discovery matter?
2. What is the reward function?
3. How closely should the training set match the test set?
4. What role should the child play in the setting up of the training?

Academic preschool is akin to supervised training. There is a desired target output, and the training data is intended to produce that desired target output. An example of an “academic” activity is: the goal is to learn how to count. The teacher introduces 1–10, the children practice counting 1–10, writing 1–10, remembering 1–10. The teacher corrects any errors, and drills provide repeated practice. At the end of this exercise, the child should be able to count from 1–10.

Montessori is akin to RL in an engineered gym. The task is to count from 1–10, but the tools that are given are number rods: 10 wooden rods that are length 1–10. Each rod consists of alternating colors, and the objective is to build an ordered staircase from 1 through 10. The child should, upon correct completion, be able to count from 1–10, but also the rod lengths could help them understand incidental things like comparison and addition / subtraction, and the alternating colors could help them understand odd / even numbers.

Reggio Emilia is akin to putting an agent into the real world, and discovering useful skills incidentally through other motivations. For example, if the children are playing outside and collecting stones, the teacher may ask them: “how many do you have”? “Do you have more or less than your friend”? “If every child here needs their own stone, how many do we need”? The reward function is a child’s natural curiosity, desire to understand the teacher and cooperate with their friends, and other individualized motivations.

There are advantages and disadvantages to each, depending on how you weight your objective function. You might believe skill acquisition is the most important thing to achieve at young ages, because they need a solid foundation before they can engage with the real world meaningfully with an individualized motivation. In that case, academic philosophies can be extremely sample-efficient in imparting knowledge, but maybe less generalizable to a different out-sample task. You might believe creativity and unconstrained idea generation is the most important thing to preserve in the age of AI. In that case, maybe Reggio Emilia philosophies are better at encouraging extremely self-motivated, broad, creative children. Or, you might believe in the Montessori middle-ground, which is self-motivated, but constrained by a particular task at hand.

As we all know, empirics with education is hard, because it’s all confounded by nature, nurture, and exogenous influences. That’s why each of these philosophies has a home. People must have priors and opinions about which of these methods is best for their child. And even more difficult is the fact that a different learning method may be optimal for each age, or each skill level. An LLM can go through many different training regimes: pre-training, supervised fine-tuning, reinforcement learning, and when deployed, RL with new online data. It makes sense that a child would similarly benefit from more than one learning method for each developmental stage and subject.

At Presto, we serve children ages 0–5. We heavily believe verifiable skill-acquisition is important, as a precursor to real-world engagement or true creativity. We believe competence expands the space that creativity can explore. A child who can count can ask quantitative questions about the world. A child with more language can index more complicated ideas. That kind of creativity is more grounded in the real world, and has a higher hit-rate to produce a real, tangible impact. That’s why our focus is measurable foundational skills, just accelerated. This is our belief in the optimal sequencing of education, but this is and will remain an open question, perhaps the hardest question there is to answer.