How to do the next right thing

Chatbot speak, unreliable tools, and directions not targets.

Nadia Piet & Archival Images of AI + AIxDESIGN / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0/

📖 Here’s what I’m reading this week

• Will you forget what it means to be human? That’s the contention of a great essay arguing that the AI poses a real threat not because of rogue robots but because “it offers answers faster than humans can ask the questions that help them contemplate their existence.” Live the the existential questions, don’t passively accept the antiseptic answers of an algorithm. (More)

• Are you willing to rely on a technology that will always be wrong some of the time? The future of the web, commerce, service delivery, etc. are all being reoriented by unreliable chatbots. In making this shift, we’re building upon a new infrastructure with a lot of pot holes, requiring a kind of on-going vigilance and default doubt. As Matteo Wong writes, “AI products could settle into a liminal zone. They may not be wrong frequently enough to be jettisoned, but they also may not be wrong rarely enough to ever be fully trusted.” (More)

• Did a chatbot tell you to say that? According to new research, we’re starting to mimic the language of chatbots, morphing ourselves to sound more the machines. Along the way, we’re losing “verbal stumbles, regional idioms, and off-kilter phrases” that communicate our humanity. Take this (totally relatable) example from the piece: “I’m sorry you’re upset” versus “Hey sorry I freaked at dinner, I probably shouldn’t have skipped therapy this week.” Don’t skip out on therapy, and don’t give up on the authentic, messy communication that makes you human. (More)

• Taylor Lorenz
  spent hours talking to people who believe large-language models are an all-knowing god. Check out what she learned (More)

This week’s guide is about how to do the next right thing amidst uncertainty. As a reminder, Untangled Guides offer strategies to see your system clearly, anticipate its behavior, and ensure technology works for (not against!) it.

---

⚒️ How to do the next right thing

In complex systems, you don’t set targets, you set directional goals. Targets foreclose real-time learning and sense-making. As Dave Snowden puts it, “targets become an entrained pattern of expectation and people cease to question.” This is particularly problematic in complex systems that don’t behave how you might expect.

Once you’ve set a directional goal, you don’t build a long-term plan, you build adaptive strategies rooted in sense-making. In my guide, “How to determine if you’re dealign with a complex system,” I distinguished between complicated systems (e.g. a car engine) and complex systems (e.g. traffic). Well, in the former, it makes sense to ‘sense-analyze-respond’ to use Snowden’s terms. A car engine is based on well defined principles of physics and engineering, so if you understand how the components interact, you can predict the engine’s behavior and fix it if it breaks down. You simply need to collect the relevant facts, analyze them, and apply the appropriate procedure.

But complex systems don’t abide by cause-and-effect. Traffic changes its behavior based on information from its environment. Focusing on one car won’t teach you much about the entire system because what matters is the interactions between them, and the patterns they form. In this case, you would follow the protocol of ‘act-sense-respond,’ so that you can sense how the system responds to your intervention and then adapt accordingly.