Why Causal Decision Diagrams are the Most Important New Development in Solving the Most Complex Problems (and what that has to do with dogs)
Lately, I’m surrounded by dogs: mostly service dogs, who are able to learn sophisticated behaviors and help their handlers in miraculous ways. And so I’ve been reading a bit of dog training theory. Surprisingly, it turns out that the causal decision diagrams (CDDs) we’ve been using at Quantellia for years are analogous to a model used in dog training and behavioral psychology in general: more evidence that, as I describe in Link based on our market research interviews, they are a universal archetype.
Dog training is about creating interactive environments that help them to navigate chains of cause-and-effect that are longer than they would otherwise learn, So we get complex behaviors like this one: [embedyt] https://www.youtube.com/watch?v=80NQLXXQwRM&t=60[/embedyt]
But this goes WAY beyond dogs. Decision intelligence, too, is about creating interactive environments that help us to navigate chains of cause-and-effect that are longer than we are able to learn on our own.
And, as I’ve learned from my new friends at Emmy-award winners The Underground Engine, computers allow these simulations to happen at scales—through both space and time—that we’re not able to understand without this kind of help.
When Mark Zangari and I invented decision diagrams back around 2010, we did so because they captured a pattern we’d observed in dozens of interviews with decision makers: decisions were described as a thought about some action, which would lead through some chain of events, under some circumstances, to some outcome.
We created a template for this pattern, like this:
As it turns out, this is a template for operant conditioning as well. I’ve learned a simple “ABC” acronym that applies both to training my puppy Bowie to “sit” as well as to the most complex analysis I perform for my client organizations. What I’ve been calling Externals are Antecedents, Actions based on decision levers are Behaviors, and outcomes are Consequents. Check out this excellent podcast on these fundamentals from Hannah Branigan (little does she know her dog training expertise has such wide implications on the most complex problems we face)
Over the years, my company Quantellia has built lots of decision models based on this archetype. At the same time, the principles used in animal training are starting to be used to train humans faster and more effectively than ever before, in domains from sports to surgery.
But there are domains that are less concrete: where, for instance, money interacts with psychology and social movements interact with climate. These can’t be trained in a physical space. Instead, we must use AI to learn the individual links using massive amounts of data, in ways that humans can’t. And we need interactive computer simulations to allow us to reason through chains of multiple links in ways that are otherwise impossible for us.
Getting this right, and solving these existential problems, means that humans must work hand-in-hand with computers. That means we must use a representation that is as natural for us as possible, to invite the broadest range of diverse expertise into the mix. Given the convergence between the results of our market research and decades-old knowledge about how humans and other animals think through the path from actions to outcomes, I’m pretty sure we’re on the right track with the CDD.