S4 E24 - Dr. David Woods - Resilience and Complexity: Part One

In this episode, I sit down with Dr. David Woods, a leading expert in resilience engineering and complexity science, for the first of a two-part conversation. Together, we explore the interplay between resilience, complexity, and systems thinking, diving into how these principles intersect with W. Edwards Deming’s transformative ideas on quality and continual improvement.

Dr. Woods introduces the foundational concepts of resilience and highlights its critical importance in an era marked by rapid digital transformation. We discuss how organizations in IT, healthcare, and manufacturing can adapt to unforeseen challenges by fostering resilient systems capable of anticipating, absorbing, and recovering from disruptions. Connecting these insights to Deming’s framework, Dr. Woods emphasizes the need for leadership to shift focus from static efficiency to dynamic adaptability.

Key themes in this episode include:

• Resilience in Action: Real-world examples of how organizations have successfully implemented strategies to thrive under complexity.

• Complexity and Systems Thinking: The dangers of oversimplifying complex systems and how embracing uncertainty can drive innovation.

• The Role of Leadership: How leaders can cultivate an environment that values learning, experimentation, and Systemic Thinking, echoing Deming’s principles of profound knowledge.

Dr. Woods’ expertise provides a thought-provoking lens for understanding how organizations can prepare for the unexpected while staying true to the pursuit of quality and improvement. Part one lays a strong foundation for the continuation of our discussion in the next episode, where we’ll delve deeper into practical strategies and case studies.

Transcript

John Willis: [00:00:00] Hey, this is John Willis again, another profound podcast. And I'm really excited. This is going to be probably, this should be a fun one. So if you've been following the track with you know, Jabe. I know a lot of people have really been enjoying some of the couple I've done in JAIB and then we did Eric Lawson and to me this is sort of the like putting, I think we're going to do a lot more than just this one, but, but really putting a nice sort of like a cap on all of that with Dr.

Woods, Dr. Woods, you want to go ahead and introduce yourself.

Dr. David Woods: Well I'm a, a person who is in between, and it all started. Back with a big accident, three mile island and the shock that represented I was just a kid fresh out of with my doctorate. And there weren't many jobs in academics at the time and I ended up going off in Pittsburgh Research and Development Center to redesign to [00:01:00] study how to people handle the emergency situations and to study how to use computers to help people handle, you know, difficult evolving problems.

And, you know, the story goes, I walked into a room full of engineers and they turned around and said, here's our new cognitive guy, please tell us how people went wrong in this accident, and how we can use computers to do better in the future. And I looked at them, trying to hide the terror in my heart and said, Oh, I left that book back in my office.

I'll be right back. Knowing full well that there was no book at that time to talk about it. So that was 79, late 79. And we embarked on a very innovative period, which very quickly was the collision of multiple trends [00:02:00] going on. One was the safety of complex systems, complex system failures. But the other was AI.

And so I was there in Pittsburgh for the AI gold rush 1 and now here we

John Willis: go.

Dr. David Woods: AI gold rush 2 and what's a gold rush? We but the other side of this was a combination of, empirical work, drawing on a variety of, of of things that you've talked about in some of your podcasts, going back to William James, going back in the history of psychophysics, perception, how people perceive and find meaning in the world.

Some of this comes from a lot of German scientists, some of you, some of which were the Gestalt is doing insight problems that there was a wide range that I had explored in, in my graduate education. And some other thinkers like Pierce or Hearst, [00:03:00] depending on do it right. But I think at this stage he's so far in the distance and, and, you know, I'm just going to go with people will say

John Willis: Pierce, yeah,

Dr. David Woods: it's on the spelling.

And on a variety of other characters JJ Gibson some anticipations of Gibson from some Germans. And all of a sudden I had a big repertoire to help me start to sort the world. But it required going in a different direction. It wasn't running hypothesis testing experiments because the world was too complex and there were too many factors.

And then the other side was design. We had something different than the studies, which is we could design things. And so influences from Pierce and and Gibson and some other characters, Nelson Goodman led us to take advantage of the new medium. And saying, what can you do with the medium? I can now design symbol systems to help people understand what's going on in [00:04:00] dynamic worlds.

And this just grew, ended up switching to the university and now teaching and working problems and got invited in to deal with AI and space missions and automation, advanced automation and cockpits, again, often driven by accidents. This got us into the patient safety movement driven by accidents that shocked the public and moving that forward and, you know, many other things culminating in how we met and the context of, you know, critical digital services around 2015 and how as I said to you then John and all spawn invited me to talk about some of the findings from this work.

I've been referring to, and I came to the meeting velocity that and went, [00:05:00] you don't need what I've done. You represent the future. You've already kind of figured out, but you don't quite understand how and why it works, how the future of a highly autonomous high capability. world really works.

The things that all these technologists with all these waves of technology think they're going to create, you, it already exists in what you've created. Let's talk about understanding these rules. And by the way, we have some starting points because you're following certain rules, even if you don't quite know them yet, or you only have a partial version of them or specific version for your setting.

But in fact, they're general rules. And so we've spent the last nine years and, you know, using your world as a another natural laboratory. So I'm very different than the typical academic who, because we look at natural laboratories, and we have to deliver something in [00:06:00] terms of design and new directions for managing change and improving performance, particularly in the face of risk.

Not performance in the face of can I be a little more productive in the short run, faster, better, cheaper pressure, but how faster, better, cheaper pressure. This comes out of our studies of space mishaps, including the Columbia Space Shuttle accident. And faster, better, cheaper pressure turns out to act in a very, very different way than people thought.

John Willis: That's right.

Dr. David Woods: And and so, and has a different kinds of effects and that helped create resilience engineering circa 2000. And so the last 24 years have been about pushing that forward and. And I'll do a plug. We've just released the first set of videos and our foundations of resilience engineering and I will encourage people to go take a look at it.

And at some point, we'll put up the [00:07:00] the links for that. In in this podcast or make them available to anyone and I will probably refer to some of them for capturing some of the key ideas we're going to talk about.

John Willis: Yeah, the, the impact, you know, again, I, I will you know, John Allspaw is, is, you know, so iconic and he would hate, he'd yell at me if I said that, but and for DevOps, because, you know, not only would.

He introduced in terms of his original sort of 10 deploys of Flickr, but, but introducing, I think the first time I remember, and I knew John way back, you know, I think I sold him chef, which is back when he was at Etsy, but but he, you know, when he, he posted you know, Richard's you know why complex system fail.

And that just shook everybody's, you know, like, to your point, like, we had a system. We saw, like, it wasn't hard for us to, like, see the translation. And then I got to meet you at Velocity and, you know, and then I've been sort of a lowercase student of you guys work, lowercase s student of all the [00:08:00] work that you and John have been doing.

And it has created incredible results. In the DevOps community, you know you know, so I think that's pretty cool. So on to this, you know, so just a quick history. I was talking to John and just for people listening, you know, John, John, and indirectly, you were very helpful in my Deming book, right?

Because I, I really want to understand, you know, what they explain to people, what Deming thought about systems thinking, complex systems. And so I I tried to, you know, sort of in a reductive way, try to use the Sully, the Miracle, and the Hudson, and, and, and it got me into bounded rationality, long story short, you guys helped me, John sent me some of the transcripts of you know, the, you know, which were public information of the, the flight, and, and I was able to sort of, so I was talking to John about six months ago, five months ago, and, hey, John, I'm writing a book about, you know, AI, and he's like, you know, of course you are, you know, and then I told him, you know, it's funny how Herbert Simon, like, shows up again, And And he goes, [00:09:00] oh, you know, Dr.

Wood just kind of want to talk to you. And and that started a great conversation that, that really, we've had a couple of conversations now. You've given me some homework and it's just, it's really fascinating this whole, I think what we're going to get on. So, so I'll start off with, you know, like, sort of when we first talk, like, I think this idea of these cycles, you know, I, I mentioned You know, sort of an AI winter and you, you know, you sort of said it's probably a better to think about it as gold rushes and florences.

So I'll just hand it off to you at that point.

Dr. David Woods: Yeah. So you make reference to this thing that today's AI explosion is not the first. And so there are historical cycles of advanced technologies, new breakthroughs in technological capabilities, and they influence cycles of growth. And stalls or stagnation comes about [00:10:00] or, you know, we, you know, so growth and change happens and then a plateau is reached or sometimes things stagnate.

Sometimes things fall back that the technology doesn't deliver as much as the initial hopes and dreams indicated. And this applies to many kinds of technologies and it certainly applies to AI gold rush one in the 80s. And I was there in Pittsburgh and I worked with some play some of the players and some of the, especially the transition from the university labs into applications at that time in different settings.

Because everyone thought the solution to whatever problem they had was a AI system of the time. And I think it's important to remember that people are participating. In this process of growth, complexification, it's a process of fluorescence. Right. A fluorescence. It [00:11:00] refers to a blooming a flowering and certainly the expectations hopes.

The envisioning is that we're going to see all kinds of new stuff happening because of the new power. We're about to unleash, but it actually plays out quite differently than people project. And there's some hard constraints about why that projection that envisioning doesn't work out. And in fact, we.

started studying not just the technology, how to design, you know joint systems of people and intelligent machines and different settings, but we were also studying the people pushing this and the processes of technology change. This process is a fluorescence install of a fluorescence that kept seeming to happen.

And then didn't happen. My virtual reality might be more of an example of one that kept keeps trying to launch and then falls back. Because something is some ingredient [00:12:00] is missing. We can contrast a little bit gold rush one and gold rush two and the characteristics of gold rushes. And some of those are about fluorescence and some of those are about people's attitudes towards these periods of change, because they're very exciting.

But that excitement can also be a little blinding to more humble, thorough, you know, what really happens more empirically science grounded. How does this really work that there are rules, there are laws, there are even theorems now in the last 20 years that say how these processes work because these are processes of adaptation.

In an environment where cycles of adaptation go on. So a gold rush is properties of some of these cycles of adaptation. So this is where, you know, the example in the 80s was, oh, we have the gold rush and then AI [00:13:00] winter. AI spring, AI winter. And that's how AIers talk to it. And in hanging around with them.

What did we really witness? Well, we really witnessed what's more typical of these fluorescents, which is, right, new capabilities came along. People projected the way they would be used. That turned out not to really be the important impact. And very quickly, things evolved in a new direction. Powerful. And, And producing a great deal of change, change that reverberates still to today, though.

It's not recognizable as AI. AI wasn't because of the technology or its impact, though it was different than what it was envisioned. But rather that AIers all of a sudden didn't have the gold rush. It was people were just throwing money at them. The assumption that if you did more R and D on the, on the systems, if [00:14:00] you package them up a little more for increase the accessibility to non specialists, we would have these huge leaps in capability, huge leaps in performance for in different application areas.

And when all of a sudden people started realizing that that sort of automatic performance leaps in the real world didn't happen. Weren't happening because the ingredients were missing of many kinds. They all kind of the money dried up and it was like, what do you mean? I have to work hard and compete for money.

People don't just throw money at me. Cause I said, I've got an AI solution. And, you know in around 2010, 2012, I was on an advisory committee about the impact of new autonomous technologies and the the first line. Of the of the reports and briefings was, you know, autonomous technologies are a capability, not a [00:15:00] solution, and that wasn't controversial to, you know, high level mission oriented strategic thinkers or to the technology people.

But that's not the way it was laid out in the gold rush. It's automatically going to come to this stuff. And so and so, well, what was the real reverberation? Well, I like to explain it in the, remember the startup companies that were going to change the world, the equivalence of open AI, et cetera, here today happened spinning off from Stanford, CMU, MIT, et cetera.

And only a few years later, after they spun up, only one of those companies was still active. And that company had transformed to not be a selling AI. What it was selling was advanced software capabilities. It was selling advanced software engineering. It was [00:16:00] selling new capabilities for building more effective systems that integrated a wide variety of human roles.

Interfaces, new visualizations, more powerful software, some, including some, quote, intelligence from the AI world, but it was really a software engineering power that everyone started to take and run with, and it represented the early stage of something that continued to evolve into, you know, critical digital services eventually.

John Willis: Right.

Dr. David Woods: So there was a shift, which is often common in these periods of fluorescence where it starts out. There's a period of there's a fallow period. And so you can look back even into the fifties, late fifties, but in the sixties of things that were going on to start to lead up to the fluorescence in the late seventies, early eighties.

And the same thing happened with [00:17:00] today's inductive AI. Where we had a much longer fallow period that actually had its stalls as we started to understand more about how the brain worked and use that. And I can remember, you know, 45 plus years ago you know, people were denigrating the neural network stuff that was emerging.

Because it sounded like analog computing and we're in the age of digital computing.

John Willis: Oh, wow. Yeah, yeah, yeah.

Dr. David Woods: Why would we think about this as analog computing? And there were other, they were political fights over limited resources and other things that went on. And so there was a fallow period and then it came back and it developed further as we look back and give credit to people like Hinton.

And then, then a few things combined and all of a sudden we have the ingredients for one of these fluorescents that takes off. And now we have the properties of the gold rush as a human endeavor, [00:18:00] not as a technical endeavor, but as a human endeavor that has the power to change what people do, why they do it, what they accomplish, what their goals are.

And in that process creates conflicts between different human groups with different purposes. Right. And different trade offs and different risks. And so we see these conflicts and trade offs come to the fore because they're universal. The you've asked me the question is, does the past in these past cycles of technology change, tell us about the one we're in now.

John Willis: Right.

Dr. David Woods: I think the answer is easily yes. And what's interesting in a gold rush is one of the properties of it is the people who are powering and excited to be part of the gold rush. It's a gold rush, right? They're not, they're not [00:19:00] they're not really, they don't really want to pause and stop. By the side of the path and say, oh yeah, let's have an interesting debate about the past and how this affects this.

And, you know, some of the safeguards or false false trails, we're going to run down, they're like, Hey, there, these other guys are running ahead on these other, I got to get running. I don't have time to talk to you. This is brand new. It has reset everything. I like to refer to in 96, Alan Greenspan, made a comment.

And this was at a period of the. com boom. And he asked as a question, not as a statement, which is important. All right. But he used the phrase irrational exuberance. And the issue was the gold rush says we have exuberance and it will be irrational. We don't have the time to fully think through because we're in a rush [00:20:00] to the new capability to happen.

We're in a rush to say, look what it might do, and others are moving forward. And so inevitably it's irrational and it is a Greenspan asked it as a question as a policymaker saying, what am I supposed to do? Is this a new fluorescence? He didn't use that word. Is this a new fluorescence that's changing the financial landscape and the tech, the or organization or organizational and technological landscape that we're in, or is this a bubble?

And, you know, my example of like calling, I call them fluorescence is it's both. It is the bubble. It is an advance. It is a change. It is important and right. It Purdue, it doesn't eliminate trade offs. It doesn't eliminate limits. It doesn't eliminate messiness. It changes the systems, the reverberation spread wide.[00:21:00]

And so people do new things in new ways, in new relationships at new scales. And this was part of the. Founding of resilience and engineering around 2000, where a variety of episodes of technological change in the context of changing human relationships, changing environments, changing goals, changing conflicts were going on.

And technology was seen as often a magic bullet. And then there was an after action report that went, wait a minute. The deployment didn't didn't lead people to have lower workload, right? To have an easier job, right? And the technology led effective leaders to ask people to do more, do it faster and do it in more complex ways.

This ability to be faster, better, cheaper, this acute [00:22:00] pressure to do more now was what we saw in organizations prior to a major accident. And we saw this in 1999 when NASA had some space operation mishaps, space exploration mission mishaps, and an independent review because NASA did independent.

Investigations said, wait a minute, this wasn't some simple mistakes on this piece or that piece, this component, that sub component, this person, this role. It was a, it was about the system and how we were under faster, better, cheaper pressure. And in the process, we were getting rewarded for cutting corners without recognizing we were cutting corn.

And that's when I first proposed that resilience was a proactive way to look ahead. All right. And and balance faster, better, cheaper pressure [00:23:00] with the ability to have adaptive capacity oriented to the future that new things will produce change. And that change will will not eliminate risk will not eliminate surprise will not eliminate.

Overload, it will move overload around, it will move risks and vulnerabilities around.

John Willis: Yeah.

Dr. David Woods: And everything will, and because it's successful, because it's valuable, will change the scale at which things happen. And so you have to look for these reverberations. You have to figure out how to help systems work.

In other words, it's a trend. It, the technologists are correct, but in the wrong way, their power is transformative. The way it's transformative is not what the technologist can see. He's, they're too busy creating the power. Yeah, I know. And view is too narrow relative to all the factors going on.

John Willis: I think a lot [00:24:00] about this right now, you know, I've been thinking about, like, I've been starting to talk about this concept of dear CIO, you know, I mean, there's, there's a little bit, you know, I think it fits right in because I, you know, one of the trends I'm seeing a lot right now is part of this fluorescence and this faster, better, not better, actually faster, cheaper you know, and, you know, like CEOs are hiring chief AI officers, chief AI officers are hiring, you know, pure, you know, AI experts, and then they're building 90 percent of the things they're building are just plain infrastructure, and they're cut, you know, and they don't even know they're cutting corners.

They don't even know what the corners are, right? They're just, you know, they're putting in infrastructure that people have been trained. And the thing I keep thinking about is like, in some ways, this is nothing new, right? Like, I'm old, like, you know, I can go back to when mainframes went to distributed computing when, you know, and then sort of jump through a whole bunch of things that happened, the fluorescence of nothing to do with AI, you know, or well actually did have something to do with it.

But and then we get the cloud and we get the big [00:25:00] data and and they all tend to be that sort of there's this sort of goal rush now and and and. The thing I think a lot about and a lot of what I've learned from you and John and Richard is like, how do, like, why do we always miss the resilience piece, you know, like the, the you know, I do something really simple like the Abraham wall missing bullets, right?

Like, can we just as a sort of, you know, a higher level way to think about why do we never think about where the missing bullets are? In, in, in a sort of a resilience approach. And we're doing it clearly now with ai, but the only other thing I wanted to add to that is it, it's something about the name in ai, because we, we, we had gold rushes with computer, but we, like, we, it didn't catch the zeitgeist of the, the winter we, well, like I was just, as you're explaining, I'm like, I, I've never heard anybody say there was a computer winter and a computer spring and, you know you know, so I don't know.

I don't wanna go too far on the, like, [00:26:00] the bad name.

Dr. David Woods: But, well, the, the, the point you're raising about why certain ones. Take off. Well, we always joke, even when back when I was in grad school that if you invented a clever oxymoron, you would be a big hit in some technical area. So artificial intelligence, virtual reality, AI safety on and on.

These are all totally But that joke aside. I think one of the factors in a gold rush one and two that's different from other major shifts in technological power. Is accessibility to non specialists, right? The promise here is you don't have to know a lot about the underlying technology and how to make it run.

You don't have to understand the machine and you know, what's inside the black box. It offers you a kind of power you can now just run with. Another one is the societal myth that we can, [00:27:00] that we solve Problems that they are closed done and dusted. And then we're done and we move on and it means that we this idea that we can that we each improvement reduces the potential for surprise, for anomalies to arise, for problem, you know, for crises to be there.

And and so as I talk about the, you know, the metaphor of the dragons of surprise, are lurking out there and the idea is I improve capabilities and deploy them and the dragons get less frequent and they get less fierce and they're, they don't, they don't breathe as much fire and, and the world gets quiet on the boundaries.

They get rare, they're corner cases, you know, black swan. Oh, no, no, no, no, no, no. That's not what goes on. Actually, [00:28:00] like the surprise is inevitable and immutable. In fact, not only is the surprise not rare, it's going on right now in every organization. The issue is. There is a source for resilient performance.

There's an ad hoc source of resilient performance. Sometimes there's some of it's designed in, in some way, but there's a special source of resilient performance which handles the inevitable gaps, right? The regular run of smaller surprises that are inevitable in any system. In other words, systems are inherently messy.

They are inherent. There are, there will always be gaps. Between the system and it's fit into the world,

John Willis: right?

Dr. David Woods: It's not getting more and more fit. Therefore, reducing the problems on the edges. Yes, some are still important and rare things can happen. Oh, no, no, no. The problems on the edges are are omnipresent, [00:29:00] but you don't see them anymore.

You don't see them because there's a source of adaptation that resolves and bridges the gaps. And those are some people, some of the time, who are overcoming the brittleness of systems. So let me point this out historically because the AI Gold Rush 2 people don't want to hear this because they don't want to get slowed down in producing power.

They want to have the fun of making the next big thing and being part of the next big organization. So one of the one of the issues is 75 years old, and it was noted by a professor at MIT, who was part of a gold rush that of creating right, the mechanisms in engineering for control systems that permeated everything we went from manual and separated control [00:30:00] to more integrated control of processes.

More automation scaling up what we could build and how we could control it. And 75 years ago, he started warning that his, the machines, right? We're literal minded. They had the danger of doing the right thing in the wrong world. There was a limit to their model of the world they were in. And the warnings about the dangers of literal minded machines echoed.

And then what happened in new gold rushes? Even the smaller ones, it was, Oh, the new gold rush takes care of that. We don't have to worry about that anymore. Cause look at how much more powerful it is. So that, and that warning of doing the right thing according to its model, but not being able to tell you're really in a different world than what was set up still lingers till today and applies as a limit on gold rush [00:31:00] too.

The second one, 40 years ago, in our reaction to AI Gold Rush 1, my colleagues and I did some studies of the new kinds of AI systems and what happened if you use them the way the AI handbook said to use them. And what did we note? We noted they were brittle. They were brittle. In other words, brittleness is, hey, there's a limit, there's a boundary and performance falls off rapidly at the boundary.

Remember, brittleness is a boundary effect, doesn't say anything about how good or how much improvement you are when you operate away from your limits, but at your limits, you fall off quickly. And in our study of deploying AI systems of the era into even not that complicated worlds, we quickly found that messiness was normal, that snafus arose.

The machine offer had value, [00:32:00] but by itself quickly ran into barriers. It was brittle, right? And the edges were approached regularly. I didn't have to design special edge and corner cases to stumble into combinations of factors, which made the knowledge of the AI systems incomplete in the inference, unable to handle the situations.

And so what we, you know, what we started in 83 was the idea that, Hey, this stuff needs to be a joint system. We're building a multi agent human and machine system, which very quickly becomes multi human and multi machine with different kinds of autonomous capabilities and roles, and this joint and distributed cognitive system system doing cognitive work was what was important to design.

So that's 40 years old. Well, guess what happened, you know 25 years ago, 25 years ago, we start to recognize empirically and formally with [00:33:00] proven theorems, the robust yet fragile theorem and the empirical findings from studying how people handle problems in dynamic risky worlds that brittleness is universal.

It's not incidental. It doesn't just happen in special situations. It isn't amenable by the next round of computer power, the next round of autonomous machine capabilities. It's, it applies universally. And this is a formal as well as empirical result that people don't want to, they don't like it. It doesn't fit their mindset, what you're slowing down the gold rush, man.

But no, this is the way this universe works. There are rules in this universe, right? And the fact you're on a gold rush doesn't change the rules, the rules operate for each gold rush. And [00:34:00] yes, there is fluorescence and new capabilities. Yes, that those new capabilities are used by effective leaders, human leaders, right?

To do new things in new ways and new relationships at new scales. But that doesn't make the limits go away. It moves the limits around. It changes the dragons of surprise, but they're still there. And you have to still have the adaptive capacity, right? At the boundaries. Well, with the improvements, when you operate far from the boundary, that these are two parallel regimes and physics has parallel regimes all over the place.

We handle it in physics. I mean, it's pretty esoteric physics, but we handle it just

John Willis: fine. Yeah.

Dr. David Woods: And physics has potential, right? We have potential energy, a potential for how things will work in the future. Well, this adaptive capacity about brittleness is a future oriented adaptive capacity. It's why I say, I summarize one of the 20 year old [00:35:00] findings in resilience engineering.

One of the first things is resilience is a verb in the future tense. Oh, wow. Wow. Out your capacity to act in the future because There's a guarantee in this natural universe that however good you are now, however better you just got the world will change like what you did to get better still is limited by finite resources.

So there's trade offs you had to make. Still, even though it's better than before. And oh, by the way, remember what we said already. Other people are adapting to take advantage of your new capabilities. So, others are adapting, the world has finite resources, change keeps going on. What does that mean? The dragons of surprise are never going to go away.

John Willis: Right. So, the, so the constant, so I was wondering, I'm pretty certain most of the people listening to my podcast are smarter than me, but for me to sort of summarize this, based on my sort of observation, which [00:36:00] was, There are always boundaries and there's the edge of the boundary is the brittleness And the thing is we don't normally see the boundary shifting as we're going through these fluorescence and therefore we feel this overconfidence Is that, would that be a,

Dr. David Woods: that's right.

That's right. So here's the way you think about it. And actually we introduced this idea of a confidence envelope back in 1987 relative to brittleness in the AI gold rush one. And so what happens is the competence envelope change, but changes. Because you have new power, you can do new things, but that doesn't mean you absorb more and more of the universe of variability and surprise.

No, instead, you have shifted, you have produced growth. Remember, this fluorescence process is a process of growth, right? So what's happened with our software infrastructure? over the last 40 [00:37:00] years, right? Our software infrastructure now underlies, underpins, often, by the way, some interesting stories of fluorescence there as well, underpinning dramatic changes in society.

What we do, how we do it. Right. We are, the connectivity goes up, our ability to share information. Notice these all have human connections, right?

John Willis: I can share more.

Dr. David Woods: I can get access to more data. I can see more parts of the world. I can sample over here, right? I can think about those things more. I can do more.

And of course, what happens is we end up All of us are working 24 seven. All of us are everyone. We expect everyone else to be accessible almost 24 seven. You know we're, we're engaged. The idea that we are not operating near overload is, is ridiculous. Everything in the [00:38:00] faster, better, cheaper world fueled by growth and capability is to push people closer to saturation.

Closer to boundaries closer to brittle brittle. Okay.

John Willis: You're not only the boundary shifting. We are with this gets us closer to the boundaries than we were because we are more interactive.

Dr. David Woods: So we don't always operate close to the boundaries, but we are regularly experiencing pressures that push us near it.

And it goes on in our everyday conversations. I'm sorry to bother you. I know you're really busy, but I'm, I'm trapped. I really need some help to get this done. Cause I'm, I'm overloaded and I'm under pressure to get something done. And all of us are, have adapted to the pressures of, of being near saturation regularly.

And we demonstrate this with certain special worlds where we can go study them. Emergency rooms when they're overloaded. I mean, what's their surge capacity? And what's interesting is [00:39:00] they have a surge capacity to, right, through what? To stretch at the boundaries. What's their surge capacity? That's this adaptive capacity, but this adaptive capacity has limits.

So the ER surges. Well, what happens if the, if the oh, and by the way, The emergency rooms operate regularly above their nominal capacity, surge capacity, their nominal capacity to handle patients. In other words, there's a nominal capacity to handle patients. They have a nominal way to surge, but their real way to surge is different.

And their real load is different. And their real capacity, surge capacity is different, bigger, and they still can get overloaded in certain circumstances. Remember the simple example is we went from half a dozen people being injured in Giffords Congresswoman's shooting, thinking that would overload the emergency medical system,

John Willis: to

Dr. David Woods: over a hundred people being injured in Las Vegas shootings.

John Willis: Right.

Dr. David Woods: And so, all of a sudden, our emergency [00:40:00] medical system has, has shifted in its capability to handle events that arise from various societal trends. The so this process of change and trade offs and finite resources is demonstrated. So what happens when the ER gets under a mass casualty or a surge of things because they're, they're happening simultaneously that goes beyond their normal surge capacity, they have to mobilize extra capacity, and they have to generate an extra capacity and what happens everybody starts adapting.

You don't follow the normal rules for procurement. You don't follow the most efficient methods to get some to get various kinds of resources. You but you also have new challenges. How do you insert new personnel? How do you reorganize? Around certain forms of expertise, because the injuries are all similar and you only had a couple experts in that kind of how to treat [00:41:00] that kind of injury burns, for example.

So, who's the burn specialist? How many burn specialists do we have? Well, quickly, they become right leaders teaching others. Okay. This is what you do. This is what you do in order to handle, right? A large number of burn patients coming in relatively simultaneously from an industrial accident. Okay. So we have concrete demonstrations.

I keep on emphasizing this. First, it is self evident in the run of shock events that happen, that go beyond our normal capacity to adapt at the boundaries to prevent brittle collapse, right? We're good We have mechanisms. Resilience is, as Richard Cook would put it, it's always there

John Willis: because

Dr. David Woods: otherwise our system would be too brittle and we would have brittle collapses or nothing would work right.

Right. And so, these things only work because of these sources of resilient performance is [00:42:00] extra adaptive capacity. And so we only see when that extra adaptive capacity is insufficient to handle a bigger kind of surprise or a combination of events that come together. And of course, all of this is eminently visible in the incidents and how the incidents are handled and only a few go so far as to produce major outages in our critical digital services.

Right, right. So the stream of events is happening all the time. Now, let me give you one of the laws. Why don't we see this all the time? Why are we blind to this? It's called the law of fluency. Fluency. What is fluency? Well adapted activity hides the difficulties, handles, and the dilemmas resolved.

That's a law, right? This operates all the time. And think about it. It happens with control systems. I, it absorbs variability. It's what Ashby said in 1956. About, [00:43:00] about what we now treat as normal, old fashioned traditional automation, it absorbs, it destroys variability, it hides it, you can't see it anymore in an effective control, dynamic control system.

But that control system still has limits and variability outside its bounds can happen. And I need something extra to come into play. I need to deploy or mobilize extra capacity when things approach saturation and saturation is in the it's in the bloody engineering textbooks. What control what undermines dynamic control at any level at any stage for any kind of system, human technological biological right lag and saturation.

And lag is really saturation in time. That's what we call decompensation, right? I can't keep pace, right? Events are moving forward faster. Trouble is building and [00:44:00] compounding faster than I, than the team, than the response system can decide on and deploy actions to affect. So I can't keep pace. And this leads us to come around because there's three kinds of ways adaptive systems break.

They break down, they saturate, right? And what do we do in resilience engineering? Try to design systems to behave intelligently, right? Near the boundaries. How do you, how do you behave effectively as you approach saturation? Well, how do you break down as you approach saturation? Well, one way is you're too slow.

You can't keep pace. Well, guess what? Another way is you can't coordinate and synchronize the different players. So in that emergency room scenario, when you're overloaded, you bring extra personnel in. Notice the [00:45:00] paradox. I need help. I'm approaching saturation, but it takes work. An event, yeah. That's a, that's in, it takes capacity that's in scarce supply to integrate the help into the flow of handling the load, which is building and moving over time.

It's not standing still. Hey, hey world, pause for a minute while I bring in this extra help and explain to them what's going on and, and figure out how we're going to synchronize our activities to handle all these patients and their needs. No, the patients are there and their needs are growing. If you don't treat them, they're getting worse.

And you now have to bring in and get these people integrated into the response at the same time. So this is an example of when I'm running out of capacity to get extra capacity to handle the load takes capacity. So how do I manage this in time to pull it off? Well, what's interesting [00:46:00] is worlds that experience this all the time, figure out in an ad hoc way, how to do this pretty well, not perfectly, not the best we could do, but pretty well.

Why? Because they experience saturation regularly and they see the risks. A brittle collapse. Our emergency responders are better at this. That's why our critical care physicians are better at this. Now they don't understand what they're better at though. Right, right. We have the principles now. We have the science now that says this is how it works and this is what you do to make it work better.