David Ng
David Ng
Aug 22, 2017 · 2 min read

Another way to look at this is through the lens of optimization. Imagine we are trying to identify inputs to maximize a survival function. As individuals, we experiment by choosing different inputs: some survive and some don’t. Over time, the human race identifies a local maximum via hill-climbing. But it’s only a local maximum—there are higher maxima out there. We just can’t find those higher maxima because those inputs are beyond us.

If we find a local maximum that’s high enough to ensure the survival of the population and there is enough ‘surplus’—then the population still survives even if some individuals choose to experiment with inputs that are farther afield. Many of those individuals won’t survive, but if one happens to find a higher local maximum, the human race migrates to the new maximum over time. As the process repeats, we continue exploring a wider range of inputs and migrating to ever higher local maxima. In essence, we never know what the global maximum is; all we know is that we have found the highest point within the domain we’ve sampled.

This actually relates to vertical learning. I believe we perform a cost-benefit analysis when we choose between assimilation (making the data fit into our mental model) or accommodation (revising the model to fit the data). Revising a mental model is costly: we have to exert resources to resolve any cognitive dissonance and, during the revision process, we experience a level of uncertainty that some find debilitating (self-doubt, anxiety, etc). I believe many people end up choosing to assimilate even when, to outside observers, it seems more ‘rational’ to accommodate because, for those people, the costs appear to outweigh the benefits. They are making the rational choice by not revising their mental model even though it doesn’t fit the data.

But that rational choice is based on limited sampling. In their experience, accommodation is debilitating and lasts a long time, and the revised model is usually not that much better than the original. So why go through that? According to vertical learning theory, the only way to encourage people to revise their mental models is by helping them sample different inputs in the accommodation function and learn that accommodation doesn’t have to be debilitating, it can be quick, and it can lead to revised mental models that are substantially ‘better’ (in survival terms).

)

David Ng

Written by

David Ng

Founder and Chief Learning Officer of Vertical Learning Labs