Time to Design Our Networks
(Because Science)
“The world is complex. It is growing more so, perhaps exponentially.”
Has said possibly every generation in modern times. Yet, there is something tangible about that “exponential” part today, something concrete that sets this age apart from the others.
That something is digital. Digital creates an imperative for science and design to get together and address the challenge of complexity. Not just any science, though. Specifically, the sciences that have to do with network behavior.
Digital is a tool, just like a lathe or a screwdriver. We use it to manufacture things. What sort of things? Networks. Newly minted ones pour out of our smart phones like widgets from a factory. Our new imperative is to design them.
Networks are nothing new. Humans were networked from the get-go, our big brains and language capacity the byproduct of our social adaptation. In modern times, transportation and phone systems were both network innovations, albeit ones with limited interactivity. Then came the original Internet in the 90’s. It was richly interactive, but anchored to a large device. Digital unshackled networks from physical limits. It enabled practically anyone, anything, anywhere, to forge interactive links at little cost. The smartphone, born circa 2007, caused our network connections to escalate, and our types of networks to proliferate. Hence, the exponential growth in the feedback, and therefore complexity, we face.
We are in the Network Age (Danny Hillis calls it the Age of Entanglement).[1] Evidence abounds. Most obvious is the swarm of network-based start-ups: Uber, Facebook and the like. Then there are key emerging network technologies like the AI cloud, the Internet of Things and blockchain. It would be a mistake, though, to think this a Silicon Valley or Beach phenomenon. What is happening is independent of any innovation center. It is the world grappling with the challenge and opportunity that effortless network manufacture brings. Almost every decision maker confronts it: from the AI researcher to the marketing VP to the head of an NGO. In each case, the question is the same.
How do I design the behavior of my network(s)?
Therein lies the problem. Complexity is born of networks. All those connections, all of them interacting in some way. As the links grow, so the variables compound. It becomes harder to isolate cause and effect, to predict and control outcomes.
Our go-to adaptation for dealing with complexity is reductionism. We naturally gravitate towards breaking a problem into manageable bits. These, we probe and study, predict and control. The result, the whole, is but the sum of these parts. As Danny Hills wrote, this is the natural outcome of the Age of Enlightenment, the brief period in human history when we thought we lived in a clockwork universe.
With interactive networks, this Enlightenment view becomes ever harder to hold to. The world is more complex than reductionists realize. Or rather, our escalating network connections are making the world more complex.
Design thinking came of age right when interactivity did, and it recently exploded in use along with the number of network links. This is no accident. Design is fundamentally an adaptation to complexity. To understand why, it helps to explore how humans process complex information from their networked environment.
Herbert Simon is a forefather of design thinking. Simon’s view was that our minds are bounded by our imperfect awareness.[2] We are not omniscient, nor are we especially rational. From Simon’s prophetic work, two opposing sects emerged on decision making and human fallibility. Let’s call them Cartesians and Externalists.
Economists are the keepers of Cartesian thought. “I think therefore I am” implies that rationality is what makes us who we are. Economics holds that we make decisions by using that rationality to consciously weigh probabilities and payoffs. This “expected return” calculus is at the heart of Cartesian influence over our decision making framework, an influence that reaches every corner: business, government, education, non-profits, etc.
Where is complexity in Cartesian thought? Pre-Simon, it was almost wholly absent. Afterwards, Kahnemann and Thaler confirmed Simon’s views about our cognitive shortcomings.[3] [4] They found that our subconscious traps us in repeated error. This would seem a big blow to Cartesian thought, but, practically speaking, it morphed into just the opposite. Behavioral economics did not spawn a new focus on complexity among Cartesians. Instead, it caused them to double down. If we are not inherently rational in making decisions, then perhaps we can recognize this, and, with discipline, reach our full potential. Big data is really the apogee of this Cartesian absorption of behavioral econ. It promises that we can replace human rationality with machine rationality when the variables become too complex for our minds to handle.
The Externalists are smaller in number and much less influential. Their intellectual leader is Gerd Gigerenzer, and, in a sense, the lineage from Herbert Simon to design thinking proceeds through him. Gigerenzer heretically holds that our irrational subconscious is a great adaptation.[5] It is tailor-made for taking on complex network behavior. This networked subconscious processing, not consciousness, is what makes us who we are. It is a processing that is inherently social: the few infants that tragically grow up with no social contact suffer physiological damage to their brains. “I” is the result of interaction between our subconscious and the external environment (hence the label Externalist).
Evolution links Externalism and complexity. Our subconscious adaptation to complexity is at least partly the key to our species’ success. It enables us to process the enormous amounts of information needed to operate as part of a social network. Think about the array of signals broadcast by a group of friends at a party: a tiny shift in the arch of an eyebrow, a sideways glance, a higher-pitched voice… Our subconscious minds crunch this ‘big data’ at supercomputer speeds. They stitch the pieces together, weave them through our emotions and intuition, and then output a plan for dealing with the network and shaping its behavior. Our subconscious is not an outmoded organ, a reptilian brain that stands in the way of Cartesian infallibility. It is, instead, our innate and powerful tool for network design.
Just as economists are Cartesian, designers are Externalist. Design harnesses the designer’s subconscious, and it recognizes the need to empathize with the user’s. Of course, it isn’t as black and white as all that, but rather a matter of orientation. Design largely bypasses the Cartesian payoffs-and-probabilities calculus. It embraces human fallibility and the chance to learn from error. It goes where complexity takes it: to cultural anthropology and ethnography for insight on social interaction, for instance.
Wherever complexity takes it, including science? Unfortunately, the answer has been “no”.
This shouldn’t come as too much of a surprise. Science is a natural ally of Cartesian reductionism, not Externalist design. The philosophy of most scientists is Cartesian reductionism. It holds that if we use our rational minds and keep searching long enough, and at a small enough a scale, we will finally discover the laws that determine the behavior of universe. As physicist Steven Weinberg put it, “The explanatory arrows always point downward.”[6]
The alliance, though, is fraying. Outposts of science are rejecting the reductionist worldview. Cosmologist George Ellis, for instance, talks about causality in physics as driven by as much by the top-down emergent behavior of a system as by bottom-up rules.[7] Similarly, biologist Stuart Kauffman and his Santa Fe Institute (SFI) brethren argue that there are no “entailing laws” that drive the behavior of life.[8] Rather, this behavior emerges from interaction in a way that is hard to predict. This philosophy of science is a radical departure. Labeled as “complex systems”, it is interdisciplinary and focused on the whole rather than the parts. It seeks to discover the properties of emergent network behavior.
If Ellis, or the Santa Fe Institute, were the only disciples of this “network science” view, there would be little to talk about. The Institute has been around for thirty years now, and despite its impressive output, it has not managed to derail science from its reductionist track.
Lately, though, various scientific disciplines have started to gravitate towards studying networks. Call this a coincidence, a function of SFI’s sown seeds, or part of network behavior (aka zeitgeist). Whichever, the fact is that network thinking is coming into focus in discipline after discipline. There are the obvious ones: ecology, for instance, and network science itself. Beyond those, many others have emerged. Evo Devo researchers like Jamie Davies argue that DNA is not a ‘blueprint’ but part of a networked chemical interaction between the environment, cells and the genome.[9] Biologists like Robert Sapolsky see our organs (and microbiome) as joined in a network that exchanges feedback in response to stress and diet, and what emerges is diseases like diabetes.[10] Jeff Clune at the University of Wyoming AI Lab wants to design learning algorithms that interact and co-evolve, and so create innovation through serendipitous network emergence.[11] Collective intelligence researchers like Nicholas Christakis discover how to improve the health of pregnant women in Honduras by targeting the influencers in their social network.[12] As these researchers seek to collaborate, interdisciplinary institutes multiply. It used to be that SFI had cornered the market on complexity. Now dozens of similar institutes exist within universities, including, notably, the MIT Media Lab.
Science, in short, is broadcasting a signal to design:
“We find ourselves caring about networks a lot these days, and even learning some useful insights about their behavior.”
Science is not ready to quit Cartesian reductionism, not by any means. There is, however, a big enough network-based revolt to matter now. We are nearing, in network behavior terms, a tipping point towards a critical mass.
So, a big change is underway, a disruptive shift. Once science fought only for the Cartesians. Now, a rebel group, intellectual Ronin brandishing network weapons, wander about, looking for a side to give them broader relevance. That side is Externalist design. Design should send out a recon squad to find them. That squad would ask some basic questions. What can we learn from information theory, from ecology and evolutionary biology, from collective intelligence? How do weave those insights about network behavior into the design process?
It is just a recon squad. The questions won’t be answered completely. Rather, there will be glimpses of the potential, like an ephemeral snowy peak on cloudy day. Still, there is much worthwhile in this early look. Like in many endeavors, just changing the frame can release untapped value.
The right question for the Network Age is, what do we know, collectively, that will help us to design networks?
There is a particularly helpful generic answer at this point. We want to search for a sharper understanding of how networks behave. In all that complexity, patterns exist. These patterns are what scientists and designers both search for. This is a common mission and a worthy one, a practical one, one that can help us solve intractable challenges like the diabetes epidemic and climate change; but also improve our lives by helping us find a date or surprising us with music we didn’t know we liked.
Networks have common patterns. The most important one is self-reinforcing behavior, aka the positive feedback loop. This is the stuff of social networks, financial crises and epidemic diseases. With self-reinforcing feedback, the network does the work. One thing leads to the next and then to more of the first, pulling in energy and using it to drive to the limit of the dynamic. The start-ups we see, like Uber, use this “work” to produce outsized outcomes from small interventions. More drivers lead to lower wait times and more passengers, which pulls in more drivers, and so on. The network acts like a lever that, with the right design, can be used to move a mountain. Feedback leading to self-reinforcement. It sounds a bit mundane, but evolutionary biologist Jamie Davies argues that it is the very stuff that life is made of.[13] What if we could learn more about it from science? What if we could learn how to bring it about from design?
There are many other designable network behaviors, ones with more exotic names like explosive percolation, panarchic cycles and hypercriticality. At first glance, these might seem less than inviting as a set of tools for d-school grads. Like anything, though, all that is needed is to design the interaction between science and designers: to make it accessible to the user, to make it meaningful so that they are able to connect with it at that subconscious level.
That brings up another early possible win in the match-up between science and design: using media to create immersive experiences for the user. Hitch network science to the networked subconscious. Make it intuitive, bring it out of research papers and into a video feast of network metaphors: ants and their colonies, traffic, social networks, rainforests, each uncovering the underlying network behavior. A generation of d-school grads emerges from this network immersion. They walk with the ability to instantly recognize basic network dynamics. They design with the knowledge that network levers move mountains.
The philosopher Gregory Bateson argued that language itself relegates us to linear grooves, to an idea that cause and effect proceed, like first graders at a field trip, in single-file.[14] Complexity, he said, is better grasped at a level below conscious language. If we recognize that language can fail us in this particular case, we can make much progress in bringing networks to everyday design. This is a job for media content, a task that artistic, creative people excel at. Anders Hoff is an unheralded Norwegian artist, and one whose art combines human creativity and machine algorithms to generate beautiful, natural-looking network shapes.[15] Grasping, intuitively, how man and machine interact to produce those networks is a powerful way to learn network dynamics. Think how much more difficult it would be to describe those dynamics to designers in a Powerpoint presentation or paper. Bateson was right: causality sometimes acts as orderly as a swarm of bees, and art is one way we can communicate the patterns in their seemingly chaotic behavior.
New sects build on the temples of existing ones; technology progresses mostly through reassembly of existing components.[16] The design thinking process itself is adaptable enough to serve as a foundation for network design. First, understand not the user experience but the network experience: How has the network behaved of late? What is blocking it from going in the right direction? What could be done to coax it towards a tipping point? Second, what are the set of possible network futures? Which one leads to the best network experience? Third, how do we design a path to push the network into behaviors that lead to that new network experience? Whom should we influence first, and what effect will that have on distant nodes, and will that spark a self-reinforcing dynamic? And fourth, how do we track the network’s resulting behavior and adapt to it?
For design and science, it is a two-way street. Design can not only borrow from science, it also adds value to it. Loads of value.
The network sciences lack engineers. One reason is that complexity studies phenomena that resist prediction, which after all is the point of engineering. A second is that engineers tend to specialize, and complexity is inherently multi-disciplinary. Complexity has had some successes in down-streaming discoveries to the engineering layer, mainly things like agent-based models used in simulations. That said, the field’s enormous potential rises behind a dam, it’s insights prevented from flowing into solutions that benefit society.
What complexity needs to release that potential is not engineers but designers. Designers thrive in ambiguity. They are naturally anti-disciplinary. The design process has humble aims: not predictive certainty and control, but the best achievable solution in a complex, changing environment. Often that solution need only be, “better than before”, or just, “better than the competition”, or even, “a failure we can learn from”.
Designers set the bar low, and this is a good thing for complexity science. Rigor, when it comes to networks, is not our friend. At least not now, when the pressing challenges we face are of a networked nature: climate change, the diabetes epidemic, the need to create a human-friendly AI. Science can inform network design, and network design can help science break the dam and unleash its potential, attract more people and funds to the complexity field, and generate yet more advances. This is self-reinforcing behavior: the more design, the more science, the more design…
Stuart Kauffman wrote, “most complex things will not never exist.”[17] He meant that the appearance of a complex organ like the heart is so unlikely. It cannot just result from a role of the dice. There was too much network dependency, too many links that relied on other links, and those on yet more links. And so, Kauffman argues, there must be a process that ‘designed’ the networked whole: not a higher being, but evolution. Likewise, most complex network solutions will never exist. There must be a process that designs them. One is the natural evolution of the economy, the networked emergence that generates innovations like Uber and Bitcoin. There is, though, a better way, a process that deliberately imagines networks into being, that allows people to use their imperfect awareness and agency to create them. Science belongs in this process; it informs it and propels it. Science is on the side of network design.
[1] Danny Hillis, The Enlightenment is Dead, Long Live the Entanglement, http://jods.mitpress.mit.edu/pub/enlightenment-to-entanglement
[2] Herbert A. Simon, The Sciences of the Artificial, MIT Press (1996).
[3] Daniel Kahneman, Thinking Fast and Slow, Farrar (2011).
[4] Richard Thaler, Cass Sustein, Nudge: Improving Decisions About Health, Wealth and Happiness, Yale University Press (2008)
[5] Gerd Gigerenzer (ed.), Bounded Rationality: The Adaptive Toolbox, MIT Press (2002).
[6] Stuart A. Kauffman, Reinventing the Sacred, Basic Books, 2008.
[7] George F.R. Ellis, On the Nature of Causality in Complex Systems, Talk at the Copernicus Center For Interdisciplinary Studies (2012)
https://www.youtube.com/watch?v=nEhTkF3eG8Q
[8] Stuart A Kauffman, At Home in the Universe, Oxford University Press (1995).
[9] Jamie Davies, A Closed Loop, Aeon Magazine, https://aeon.co/essays/the-feedback-loop-is-a-better-symbol-of-life-than-the-helix
[10] Robert Sapolsky, Why Zebras Don’t Get Ulcers, W.H. Freeman (1994).
[11] Jeff Clune, Podcast Interview, Design4Emergence.com (upcoming), March (2016).
[12] Nicholas Chritakis et al, Social Network Targeting to Maximise Population Behaviour Change, The Lancet http://www.thelancet.com/pdfs/journals/lancet/PIIS0140-6736%2815%2960095-2.pdf
[13] Davies…https://aeon.co/essays/the-feedback-loop-is-a-better-symbol-of-life-than-the-helix
[14] Gregory Bateson, Steps To An Ecology of Mind, University of Chicago Press (1972).
[15] Anders Hoff, Inconvergent: A Study of Generative Algorithms
[16] Brian Arthur, The Nature of Technology, Free Press (2011)
[17] Kauffman et al, No Entailing Laws, But Enablement in the Evolution of the Biosphere, http://arxiv.org/abs/1201.2069
