,
11 tweets,
2 min read
Read on Twitter
The debugging problem of research. Solving research challenges has much in common with debugging software that contains 2 or more bugs. 1/10
Fixing software with 2 simultaneous bugs is difficult because iterations may not direct you to the correct fix. Only if the iterations are simultaneously constructive *in each dimension* will things improve. 2/10
Same with research when 2 or more factors are missing. Adding only 1 factor may not improve a model of one's understanding, and may only make things worse (as with buggy software) 3/10
But if several factors are applied simultaneously, it may be enough to lead one to the right path. Yet, the number of combinatorial factors is immense, which is why research can still make progress. 4/10
That's why one should not be dissuaded by suggestions of "If it's such a good idea, why hasn't someone thought of it before?". Well, cause of combinatorics, that's why! The search space of co-factors is still immense. 5/10
For the ENSO problem, additional factors are period doubling, tidal forcing, and solving Laplace's tidal equations analytically. These all had to be applied simultaneously to see improvement in the model. 6/10
No wonder that no one had made any progress. Applying only one factor wouldn't make a big enough dent in matching a model to reality. 7/10
How does the choice of factors evolve? By solving related problems applying the same factors. For example, QBO, Chandler wobble, and LOD also involved the same factors. 8/10
Those are like the test cases for solving a software bug. Pick smaller behaviors that narrow the range of possible factors. 9/10
Thereafter the combination of factors solve the more challenging problem, while vastly improving the plausibility of the solution as it solves other related problems. 10/10
@threadreaderapp unroll
