Cognitive Neuroscience with Courtney Nash & Lindsay Holmwood

Systems of Technology, Systems of People

Trevor and Bridget talk with Lindsay Holmwood and Courtney Nash about how human brains shape the way we build and run systems. Lindsay moved from engineering into management a few years ago and has spent about three years reading everything he can about how people and groups work and about cognitive biases. He's three weeks into a role as infrastructure and platforms lead at the Australian government's Digital Transformation Office, whose remit is to build "clean, fast, simple, and humane services." The platform is the easy part, he says, and the hard part is helping teams learn to get the most out of it. Non-Australians can be hired with visa sponsorship.

Courtney, an editor at O'Reilly and Velocity conference chair, was a cognitive neuroscientist in training, a year from finishing her dissertation on learning and skill acquisition, when she left to work at Amazon. Velocity struck her as operations and therefore not for her, until she found people who realized that "when you have to start talking about systems of technology, you have to start talking about systems of people." She and Lindsay met at Mountain West RubyConf in 2013 after his talk on escalating complexity, which took 60 or 70 hours of research and was emotionally taxing because "people actually die."

Fast, Slow, and Priming

Courtney warns that if someone says their product is deep-learning AI, most of that is "just not true," and that the brain tricks us into thinking we make rational decisions. She explains the Stroop effect: it takes longer to say the word blue when it's displayed in green than in blue, because the brain systems for reading words and reading colors conflict. Priming can produce similar friction, for instance in how quickly you associate a woman with the word physician versus software developer. The interesting part, she says, is becoming aware of where those conflicts sit.

Lindsay has given his DevOps Field Guide to Cognitive Biases three times. He describes Kahneman and Tversky's two systems, a fast one and a slow one, with the brain optimizing for speed over accuracy. He notes that Thinking, Fast and Slow has one of the lowest completion rates among Kindle books because it's dense, and admits he stopped after a couple of chapters. Smell, he says, bypasses the fast system, and he tells of a New Zealand safety company that has workers mount a handkerchief with a loved one's perfume on their equipment to prime them to slow down. Courtney adds that smell takes a different route in the brain than the other senses. His gentler recommendation is Dave McRaney's You Are Not So Smart.

Biases Can Be Retrained

Courtney says the good news is that brains aren't fixed. She had "a wicked reaction time" on the Stroop test only because she practiced it to embarrass people. Lindsay says biases come from underlying heuristics and can be countered. In a blameless postmortem, he says, you can appoint a devil's advocate to argue for the old view of blaming a person, which gets people to spend more time justifying what actually happened.

His other example is the Dunning-Kruger effect: the less you know about something, the better you think you are at it, and you're also worse at recognizing genuine skill in others. He says it's what "basically drives DevOps," with ops thinking developers can't run things and developers sure they could keep it up if only they got in there. Even minimal training in a topic improves your ability to self-rate, and being embedded in another team is enough. He also claims the effect mostly shows up in the West and doesn't replicate in Southeast Asia.

Three Biases That Get in the Way of Empathy

Courtney says empathy has reached buzzword status, which she likes, and recommends Indi Young's work on it as a skill. Lindsay names the three biases he sees affecting it most. The fundamental attribution error is judging someone's action by their character, so the driver who cuts you off is an idiot, and not a person swerving for a possum or a wombat. The better-than-average effect: he says a study in Australia found 70% of people rate their driving above average. The halo effect is when a great impression of a person or brand colors your reading of everything they do, which is how you get a cult of personality.

Normalcy Bias, Alert Fatigue, and Anomaly Detection

Bridget asks about Lindsay's talk on the psychology of alert design. His main takeaway is the normalcy bias: before something bad happens we discount it because it never has, and during and after an event we're slow to recognize it. He cites research on tornado sirens finding people need confirmation from five to seven sources before acting. Courtney relates it to alert fatigue: the brain filters and finds patterns, so if alerts fire without anything bad happening, it stops reacting. Bridget adds the temptation to raise the threshold on an alert that's always going off.

Lindsay was working on an anomaly detection startup until a month earlier, chaining open source tools together to mimic newer findings on how the brain does pattern recognition, like "a basement full of college students that are just looking at a single graph all the time." He says statistical techniques built on normal distributions don't fit web operations data, and he plans to open source the work. Courtney says most so-called AI is really good pattern recognition, which isn't nothing.

Getting a Talk Accepted

Courtney asks Lindsay about his early Velocity rejections. He says his talks were procedural and purely technical, and got better once he related the work to bigger ideas from management, leadership and psychology books, which he reads instead of Hacker News. His Field Guide talk is, he says, "a blatant ripoff" of Sidney Dekker's Field Guide to Understanding Human Error, and it worked.

Courtney's advice, borrowed from Scott Berkun, is to give a talk about something you love, hate or are very good at, and to treat the CFP as the proving ground for the idea. Lindsay tries out 5 or 10 minute pieces at local meetups, taking notes the way comedians do. Bridget says speaking at meetups also gets you video, which program committees will watch for unknown speakers. Courtney agrees big stages aren't where to start: "it's terrifying."

Check-Outs

Lindsay recommends Patrick Lencioni's books, McRaney's two books, Kahneman's, and QF32, a Qantas pilot's account of landing an A380 and the team dynamics involved. Courtney recommends Ramez Naam's Nexus trilogy and Neal Stephenson's Interface. Bridget recommends the indie film Advantageous and two Velocity Santa Clara keynotes, by Laura Bell and Astrid Atkinson.