“The first answer you choose on a multiple-choice test is best.” This advice I heard in school feels right. Trust your gut, it implies. But a closed-book, multiple-choice test is a pretty artificial scenario: you can’t research additional data, collaborate with other people who hold different views, or otherwise adjust or calibrate your decision.
In real life, information is everything—not just what you already know, but also what you can learn from the information you can access. For example: Are you asking the right questions? How do you assess the answers you receive—or the lack thereof?
Seeking out and wading through information can be tedious; it introduces friction into the decision process. And there is a point when “doing research” morphs into analysis paralysis. But adding the right degree of friction to a process can be a lifesaver (sometimes literally).
The upside of slowing down
Another, more counterintuitive benefit of slowing your reaction time is that too-rapid reactions can cause unnecessary oscillations and destabilization of systems.
Managing delays and reaction times is a tough balancing act. As I wrote in an essay called “All About Oscillations”, if a car dealer observes a sales increase that lowers inventory:
The dealer wants to wait long enough to make sure the sales increase isn’t just a blip. Then she starts ordering more cars to meet demand. But there’s a delay before the inventory becomes available, during which inventory drops even more as sales continue. So orders end up increasing again.
But on the back side of this dynamic, as the system catches up with earlier actions, inventory arrives… and arrives... and arrives… ultimately resulting in a glut of cars. Whoops. So orders will eventually be cut… and cut… and the oscillation will oscillate in the opposite direction.
It might seem like shortening the delays in the system, or shortening the dealer’s reaction time, could smooth out these oscillations. But, perhaps counterintuitively, Meadows writes that shortening delays may not have much effect. And shortening the dealer’s reaction time—allowing the dealer to act based on even less information—actually amplifies and increases the magnitude of oscillations! Instead, increasing reaction time and gaining more information before acting can smooth out oscillations and make them more manageable.
Of course, there are times when reaction time is of the essence. But as a wonderful doctor once told me, “This isn’t a five-alarm fire.” Most things aren’t.
Automated contracts are terrifying
Delays and friction can also benefit counterparties in business contracts. An ideal contract establishes boundaries for a business relationship and provisions for worst-case scenarios while leaving room for human intervention to prevent those worst-case scenarios. In many cases, it would be a disaster if contract provisions were automatically activated the instant something went awry.
That’s one reason why automated smart contracts are a little scary: they don’t necessarily hit the brakes in situations where humans would recognize the need to slow down and provide some flexibility. If a smart contract says, “When X happens, this counterparty is in default and will be liquidated,” that’s what will happen when the contract executes, even if giving the counterparty another hour or day to provide funds could prevent a market dislocation affecting thousands of other participants.
An AI warning
The benefit of human flexibility in business processes is especially important to keep in mind as 2023 dawns, since this year will likely bring more AI breakthroughs and integration into everyday tasks. It will be tempting to minimize human decision-making to increase efficiency and cut costs.
But the truth is that humans are great at managing ambiguity, and machines still suck at it. How many millions of people have been reduced to jellied piles of fury shouting “AGENT! REPRESENTATIVE!” into their phone because their issue falls somewhere between “Option 3” and “Option 4” or just isn’t listed at all?
I’m pretty sure this is not the amazing future anyone envisions. To avoid outcomes like this in an AI-powered world, developers and business leaders would do well to remember the value of leaving space for delays, friction, flexibility, and humanity.
Counterintuitively, they can make systems work better.
The person who codes the contract could build in logic to provide extra time, but how much time? What if a counterparty needs three days instead of one? What are the effects, both positive and negative, on indirect counterparties? There’s no iron-clad way to predict these kinds of specifics in advance, yet they can be vitally important when SHTF. “The algorithm” is not the best way to deal with these kinds of situations.
I understand what you mean, a concern about the minutiae inside a machine of an agreement, primed for scenarios that we may not have accounted for, calling for buffers of friction/time, which through trial and error will have to be incorporated.