Attraction vs. Repulsion

Written by  on February 9, 2016

While criticizing this article at the CSS reading group, I made the comment that I thought it was a terrible (yet great in one way) article because it spanned the gamut of topics without digging deep enough into any one topic. I paid homage to the method, adopted by another member of the group, to the concept of studying “seminal” (paternalism, anyone?) or otherwise canonically “good” articles, rather than studying bad ones. The other person then made the statement that (paraphrasing) “attraction is a better gradient to follow than repulsion because the number of outcomes from attraction is lower than that of repulsion. With repulsion, you can end up anywhere.” I didn’t pick nits with him at the time because, in these meetings, I prefer to stay close to the topic. But his position exhibits the very same thing the article cautions against, the concept that there is an objective reality and we can approach it by targeting. So, it would have been on topic … albeit argumentative.

I much prefer the concept of constraint-based reasoning, which is founded on repulsion. Any solution within the solution space bounded by the constraints is reasonable. And perhaps the optimal solution is located in the “middle” of the bounded space. So, the only way attraction is a better method than repulsion is if the space has a large ratio of unbounded to bounded sides.

To boot, when a space is relatively bounded, relying on attraction is mind-numbingly restricted. And this is the core problem with both PhD programs, big science funding, and peer review. The extent to which one is free to bounce around in solution space is severely limited if one tries to mimic previous results. By contrast, if you can clearly explicate your boundaries, then you have the freedom to find any plausible solution/invention within those boundaries and it will likely be more novel than solutions found by mimicry/attraction.