Selection as a design tool
First, I wish Steve Jones [an evolutionary biologist excerpted in the presentation] had been a bit more careful in his terminology, because the process he was talking about was not natural selection but artificial selection. There are important differences. One is that people can control and even design artificial selection. The people who created that soap nozzle [which Jones used as an example of evolutionary design] chose how many generations to create, how many progeny to create per generation, how to create the variation (which was certainly not perfectly random), and how to tie their selection to fitness for the task at hand. They also could choose to stop the selection at any point and veer off on a totally different tack. Because artificial selection is artificial, it can be incorporated into deliberate design in ways that makes it much more valuable to human ends than natural selection could ever be. So a "roomful of scientists" is exactly the group that can use and control and benefit from artificial selection.
It follows from this that selection is not the only good way to solve wicked problems; it is one tool in the designer's toolkit, and many of the tools work together. I've had some little experience with this myself. In a previous career making educational software, I designed an artificial selection process for "breeding" plants in simulated 3D space (to teach botany and to help artists populate 3D worlds). An essential element of the system was the ability to "reach in" and stop the evolutionary process at any point, make some changes, restart it, and essentially incorporate it into a larger design process. This relates to the point I've often made in my writings, that in human life self-organized complexity does not exist apart from human-made structure, and that it is pointless to pretend we can or should leave structure entirely behind. The greater utility is found in intelligently managing the two sources of order (organization and self-organization) to create something neither could create alone.
Getting to the main point about development and evolution, to my mind there are three main issues to consider in applying artificial selection to problems of development.
My fitness, your fitness
One issue is that selection must be strongly tied to actual fitness in order for the process to work. In human societies fitness depends on perspective, and there may be conflicting ideas of what deserves to reproduce and what deserves to die. Aid that makes donors want to give more money is not always aid that works; aid people will accept is not always aid that helps them; aid people will admit works is not always aid that works; aid that works once is not always aid that will work again; what helps in one place may harm in another; my solution may be your problem; and so on.
This is somewhat like sexual selection, in which male peacocks grow larger and larger tails and become more and more vulnerable to predation. They are more fit overall because they are also likely to sire more offspring. In natural selection (which envelops sexual selection) the balance between death and reproduction irons out all details with a heavy hand. But in human affairs there is no such equalizer. If what matters to you is vulnerability to predation, you may see peacock tails as a disaster. But what if I see things differently? Who decides what is most fit and should reproduce? What happens when people can't agree on that? Does it just go back to power and money again? If it does, what have we gained?
Even something as apparently binary as death is not so simple when you consider human affairs. If a cockroach is dead, it's dead. When you move into the plant world where organisms are modular things get more complicated. I have the stump of a maple tree in my yard that has been rotting away for more than ten years; but every year it reminds me that it is not yet dead by putting up new maple stems (which I ruthlessly "kill" wanting the sun for gardens around it). Is the maple tree dead? Yes and no. (And here I simply must quote the irreplaceable Billy Crystal in The Princess Bride: "It just so happens that your friend here is only mostly dead.") In the same way, it might be hard to say whether development projects are alive or dead. What constitutes death, and who decides that?
As with death, reproduction also gets difficult to determine when you are setting up an artificial selection process related to human endeavors. Reproduction in biological evolution seems simple from the outside -- either the place is swarming with babies or it isn't -- but even there things are sometimes difficult to parse. In many species "maiden aunts" take care of the children of siblings, or of their parents, and thus promote kin selection to the detriment of their own reproductive success. Is this reproduction? Yes and no. When I was in graduate school in the 80s the concept of "sneaky fuckers" was all the rage (yes, that was the technical term) and it led to what is now a widely accepted understanding that reproduction in many species has various complementary modes, some of which involve deception and counter-intuitive means to the popular end (such as looking like a female in order to get close to the females). So even in nature reproduction is not as simple as it may seem.
In the soap nozzle example of artificial selection, control over reproduction was easy: one nozzle was carried forward to the next generation, by human caveat. But in selecting development projects, what would reproduction mean? A project would get more money? Or an aid group would? Or an aid worker? Or an approach? If an approach got more money, how would the system enforce adoption of the winning approach? What if groups or approaches got more money by "sneaky" reproduction, meaning, getting money outside of official channels? Would this be considered valid reproduction? If not, how could it be stopped? If so, what effect might it have on what is selected and how development evolves?
Mourning the soap nozzles
The second issue I see is that natural selection embraces death but people do not. If you are designing an artificial selection process to create a better soap nozzle, nobody mourns the lost soap nozzles. So it's easy to make strong decisions about what lives and what dies. But when you are trying to help people, it's harder to work in that critical death element. Everywhere you put it, it bounces off or morphs into something else.
I would suggest that to plan a successful artificial selection process for development policy, what dies and what reproduces has to be carefully chosen and agreed upon, and you need to be prepared for such definitions to change over time. This issue was mentioned by other commenters [to Owen's post] who said evolution was "unforgiving to the weak" and that "you have to work against the political forces that resist calling a failure a failure". It hinges on values. You need to find some soap nozzles you can throw away without anyone rushing to defend them or mourning their loss. Artificial selection in development won't work if projects or funding sources or field units or institutions must die, because people will fight to keep all of those things alive. You might say approaches can die, but people get very attached to approaches. You might say ideas can die, but again, people fight for ideas. What can die? What can be selected out? It's a hard question, but you can't proceed without getting past it.
What you select selects you
The third issue is the issue of scale and awareness. If mice in a particular valley are falling prey to a particularly clever cat with exceptional night vision, mice halfway around the world don't hear about it and tell each other stories about night cats. But people do. This is both an enabler and an obstacle in planning artificial selection processes. Say you design a system where only projects that meet five carefully chosen criteria receive more funding. You have set up a system of artificial selection, with fitness tied to particular characteristics of individual elements in the system. Fine. What is the probability that in a few years' time every single project meets those criteria? What is the probability that some of those projects will fail to produce positive outcomes for those they mean to help, or even hurt those they mean to help, even though they meet the criteria on paper? What happened? People were aware of the selection and reacted.
This change-it-and-it-changes-you-back situation reminds me of the folk tale where the devil gives somebody three wishes, and it seems wonderful until the person gets the wishes and realizes how the devil can deliver them in a devilish way. In one movie that repeated this old story (Bedazzled), the main character wanted to be very rich, and he became very rich -- and a drug lord about to be attacked by rivals. Unlike mice, people become aware of global patterns and change their behavior, appearance, tactics, even sometimes their self-definition to suit the new criteria. So I would say that artificial selection, when it applies to people, has to include a recursive element that operates on itself. Which forms of variation and selection deserve to die, and which should reproduce? And again, who decides?
Watch out for those sharp edges!
I'm not saying you can't apply artificial selection to development. I'm saying it's a tool whose power and danger have to be equally respected. This is partly because it will be operating on top of several layers of biological and sociocultural evolution that can't stopped to create an "all other things being equal" experiment. (Some argue that biological and cultural evolution have become so intermingled in human life that they cannot be considered separately.)
Artificial selection has been responsible for many of the best things humans have done, but it has also been responsible for some of the worst. So it's not a solution without its own dangers. Feedback loops can be positive as well as negative, and even with negative feedback the wrong things can feed back. Both natural and artificial selection are replete with stories of events going "off the rails" with disastrous consequences. As I've mentioned on this blog before, I'm concerned about people believing that complexity science and evolutionary theory present panaceaic solutions to human problems. If you think complexity is a uniformly benevolent force, Google the term "ant mill". There are some videos of this phenomenon on YouTube. They make my hair stand up. It's smart to be aware of complexity and evolution, but it would be as much of a mistake to look for simple or easy solutions in them as it is to look for simple and easy solutions in rigid centralized planning.
Selection and narrative
Now of course I have to make a plug for the complex solution I've been working on for the past ten years: stories. Stories are unique vehicles of human communication, packages of thought and belief and value, tiny simulations of life itself, that we use to make sense of our lives together. I've helped lots of organizations and communities work with raw, personal stories to create new feedback loops that bring difficult-to-articulate values, beliefs and experiences where they most need to be heard. I've helped people use narrative techniques to pump up diversity of thought through exposure to new perspectives; to facilitate selection through self-organizing participation; and to connect variation and selection to fitness functions relevant to the community through the crystallizing lens of sensemaking.
I sound like an evangelist, and I suppose I am one! My bias is that narrative work is particularly well suited to enabling well-informed, self-reflective creation of artificial selection processes in human societies. If there is any chance of people meeting the goals you set out in your presentation, I humbly submit that stories are likely to be involved. The reason I wrote (and am now expanding) my free book on this subject is to help people use narrative methods for exactly these purposes.
Readers, I recommend Owen's excellent presentation, and I wish him the best in his important work!