Simanaitis Says
On cars, old, new and future; science & technology; vintage airplanes, computer flight simulation of them; Sherlockiana; our English language; travel; and other stuff
OVERDOING IT
IS THERE BETTER LIVING THROUGH OPTIMIZATION? Not necessarily, posits mathematical modeler Coco Krumme in her new book Optimal Illusions. In their review of the book in Science, September 14, 2023, Dov Greenbaum and Mark Gerstein term it “Resisting Efficiency’s Overreach.”
Optimal Illusion: The False Promise of Optimization, by Coco Krumme, Riverhead Books, 2023.
IndieBound observes that Coco Krumme is “an applied mathematician and writer. After completing her doctorate at MIT and working in technology, she founded a scientific consultancy and moved into a cabin on a remote island in the Pacific Northwest, where she now resides.”
I’d say she’s doing just fine, thank you.
Krumme’s Thesis. As reviewers Greenbaum and Gerstein observe, she asks “why no one questions whether we should be optimizing in the first place…. At the crux of her thesis is the notion that our unchecked pursuit of optimization has precipitated the erosion of three fundamental pillars within systems: ‘slack’ [the ability to adjust to setbacks], place, and scale.”
Slack. “Incremental failures,” the reviewers note, “accumulate stealthily in optimized systems until an unexpected breakdown ensues,” bringing entire industries seemingly to their knees.
Image from CLEANPNG.
I recall the Covid pandemic as an example. The reviewers write, “The consequences of losing slack are most visible in just-in-time supply chains with systems tailored for lean efficiency rather than resilience.”
Scale. “Meanwhile,” Greenbaum and Gerstein observe, “our relentless pursuit of optimization has inflicted a profound loss of scale, argues Krumme, as characterized by the rigidity of systems that are hamstrung in their capacity to change course or evolve when the necessity arises.”
Image from FREEP!K.
Large system are evidently more difficult to adjust than local ones. There’s an inherent inertia in complex algorithms.
Place. Greenbaum and Gerstein continue, “Optimization also erodes ‘place,’ diminishing the distinct wisdom that fosters a diverse range of practices and distancing individuals from the underlying data context that underpins the algorithms.”
They note, “This predicament becomes evident, Krumme maintains, when those entrusted with an optimized system ‘lose track both of its design and its vulnerabilities’—a paradoxical outcome of obsessive refinement.”
Image from HiClipArt.
Klumme is a Silicon Valley alumna and recalls the mantra “What gets measured gets managed.”
But not necessarily to society’s betterment.
A Proposal. “In the book’s final section,” the reviewers say, “Klumme advocates optimizing locally, toward simpler-to-describe optima as an effective strategy for deliberate deoptimization. This approach addresses the pitfalls associated with optimization ‘while maintaining its many virtues.’ ”
A thoughtful review of a somewhat counter-cultural book. ds
© Dennis Simanaitis, SimanaitisSays.com, 2023
Simanaitis Says 
I question the objective function of the optimizer described above, and therefore the definition of “optimization” that appears to be accepted.
I’ve worked in mathematical optimization for nearly my whole career. Early on we moved away from optimization only on price, for multiple reasons.
You can add whatever factors you like to your objective function, as long as you can mathematically model them and normalize them with the other objective function factors. This could include maximizing the happiness of the delivery drivers or minimizing the total emissions of the supply chain.
The problem to my mind is that we have socialized so many of the costs (e.g. emissions of international transport) that those optimizing can ignore those factors and still be a viable business.