By connecting practitioners to learn from each other, peer learning facilitates collaborative development.

How does it compare to expert-led coaching and mentoring “fellowships” that are seen as the ‘gold standard’ for professional development in global health?

Scalability in global health matters. (See this article for a comparison of other aspects.)

Simplified mathematical modeling can compare the scalability of expert coaching (“fellowships”) and peer learning

Let *N* be the total number of learners and *M* be the number of experts available. Assuming that each expert can coach *K* learners effectively:

For N>>M×KN>>M×K, it is evident that expert coaching is costly and difficult to scale.

Expert coaching “fellowships” require the availability of experts, which is often optimistic in highly specialized fields.

The number of learners (N) greatly exceeds the product of the number of experts (M) and the capacity per expert (K).

## Scalability of one-to-one peer learning

By comparison, peer learning turns the conventional model on its head by transforming each learner into a potential coach who can provide peer feedback.

This has significant advantages in scalability.

Let *N* be the total number of learners. Assuming a peer-to-peer model, where each learner can learn from any other learner:

In this context, the number of learning interactions scales quadratically with the number of learners. This means that if the number of learners doubles, the total number of learning interactions increases by a factor of four. This quadratic relationship highlights the significant increase in interactions (and potential scalability challenges) as more learners participate in the model.

However, this one-to-one model is difficult to implement: not every learner is going to interact with every other learner in meaningful ways.

## A more practical ‘triangular’ peer learning model with no upper limit to scalability

In The Geneva Learning Foundation’s peer learning model, learners give feedback to three peers, and receive feedback from three peers. This is a structured, time-bound process of peer review, guided by an expert-designed rubric.

When each learner gives feedback to 3 different learners and receives feedback from 3 different learners, the model changes significantly from the one-to-one model where every learner could potentially interact with every other learner. In this specific configuration, the total number of interactions can be calculated based on the number of learners *N*, with each learner being involved in 6 interactions (3 given + 3 received).

The total number of interactions per learner is six. However, since each interaction involves two learners (the giver and the receiver of feedback), we do not need to double-count these interactions for the total count in the system. Hence, the total number of interactions for each learner is directly 6, without further adjustments for double-counting.

Therefore, the total number of learning interactions in the system can be represented as:

Given this setup, the complexity or scalability of the system in terms of learning interactions relative to the number of participants N is linear. This is because the total number of interactions increases directly in proportion to the number of learners. Thus, the Big O notation would be:

This indicates that the total number of learning interactions scales linearly with the number of learners. In this configuration, as the number of learners increases, the total number of interactions increases at a linear rate, which is more scalable and manageable than the quadratic rate seen in the peer-to-peer model where every learner interacts with every other learner. Learn more: There is no scale.

Illustration: The Geneva Learning Foundation © 2024