The Gardener’s Dilemma

Notes on the practice of innovation and technology commercialization

A colleague at T2VC posed this question:

  1. We know that if everyone is an entrepreneur, a society will not function. 
  2. We know that if everyone is a producer, a society will wither. 
  3. Thus, is it possible to determine the exact proportion of innovators versus producers to maximize the productivity of an ecosystem? 

This question is, I think, part of a broader one: in an innovation ecosystem is it possible to adjust all the various elements so that the ecosystem optimize its outcomes (measured in some way)? These elements may be universities, technology commercialization systems, public sector financing funds, and so forth – more about this later. You may remember that in the February 2013 Imperfect Works blog in this series we talked about optimizing which means “make something as good as possible within a whole system” although each individual element may not be operating in the best way it can.

In their paper Chaos prevailing on every continent: Towards a new theory of decentralized decision-making in complex systems, researchers David G. Post and David R. Johnson present a method of finding optimal configurations of elements in complex adaptive systems. The authors pose the Gardener’s Dilemma:  how can a gardener find the best or at least a good configuration of a collection of plants whose overall “fitness” (for example total yield) is dependent upon the behavior of all the other plants? You may begin to see how the garden is an analogy for an innovation ecosystem. We shall see shortly if this analogy is helpful.

In this imaginary garden there are plants of different species. The gardener would like to obtain the most luxuriant overall growth. The gardener must decide for each individual plant: should it be pruned or not? How can the best combination of pruned and un-pruned plants be created that will produce the greatest yield for the whole garden?

As always, we need to make some assumptions, which for this garden are:

(1)   The relationship between an individual plant’s pruned or unpruned state and its growth is different for each plant. For some plants growth will be increased by pruning, for other plants pruning will reduce their growth.

(2)   Each individual plant’s growth can be affected by the growth of other plants, for example as one plant grows it might block sunlight reaching another plant that needs it. Kauffman calls these spillover effects although I prefer interactions.

Even with such simplifying assumptions it turns out that this problem, and many similar ones, are “computationally intractable” or in other words incapable of true solutions by any known analytical methods. A little thought may convince us why this is the case for our garden. Suppose there are only three plants each of which may be pruned (we will call this state 0) or unpruned (we will call this state 1). We can use the tree diagram below to help figure that 8 possible configurations exist, namely 2 x 2 x 2 = 23.

ImageA little further arithmetic will show for 4 plants there are 16 possible configurations, namely 2 x 2 x 2 x 2 = 24 and so on. To generalize this pattern we can say that in any system with N elements, each of which can take one of S possible states, there are SN different system configurations. This SN can rapidly become a very large number. For 10 plants the number of configurations to be tried is already 1,024. Hard work for the gardener!

Is this gardening knowledge of any practical use for those of us developing optimal innovation ecosystems? If this this ecosystem analogy is computationally intractable, why am I wasting your time discussing it?  Stay with me for just another moment and I will demonstrate its practical use – but first one final research result.   

When a problem cannot be solved by mathematical analysis computer modeling may help, even greatly oversimplified models can be used as long as their simplifying assumptions are not forgotten when applying the results to the real world. Stuart Kauffman and his colleagues have developed a family of computer models and problem-solving algorithms for complex interconnected systems, known as “NK models,” for studying various forms of the Gardener’s Dilemma in, for example, evolutionary biology and cyberspace law. We don’t have space for details (which are in Johnson and Post’s publication) but essentially the method consists of modelling interactions between ecosystem elements. The elements in the garden ecosystem are sub-divided into any number of non-overlapping but interacting self-optimizing parts called patches, like a patchwork quilt. 

ImageTo quote Post and Johnson “The result is a fairly remarkable one: It is by no means obvious that the highest aggregate fitness of the system will be achieved if it is broken into quilt patches, each of which tries to maximize its own fitness regardless of the effects on surrounding patches. Yet this is true. It can be a very good idea, if a problem is complex and full of conflicting constraints, to break it into patches, and let each patch try to optimize, such that all patches co-evolve with one another.”

A typical innovation ecosystem has elements (S) such as universities and research institutes with  their technology transfer offices (TTOs), new business incubators and accelerators, some form of central support organization to assist TTOs with issues such as market intelligence and so forth, financing programs such as early-stage R&D grants and seed and venture funds, economic development organizations, science and technology park, a contract research organization, and sometimes miscellaneous organizations which were formed for different times but are still functioning.  All these can certainly exist in many states so N will be much larger than 2 as in our garden.

Applying what we have learned, it is good practice to divide these ecosystems into patches, for example four possible patches could be:

Universities and research institutes with  their technology transfer offices (TTOs)

Central support organization to assist TTOs

Early-stage R&D grants and seed and venture funds

New business incubators and accelerators

Contract research organization

Science and technology park

Economic development organizations

Miscellaneous organizations which were formed for different time

Having the elements trying to maximize their fitness within their patch improves trust and communications between them – and as a result – decreases transaction costs.  Furthermore, this should break down the rigid hierarchical structure from which some ecosystems suffer. Other practical applications of the Gardener’s Dilemma can be found, such as explaining the familiar S-shaped curves of technology growth and disruption. Climbing around the fitness landscape searching for peaks and valleys will be discussed in future blogs.

Your response to all this may be “but this is obvious, I don’t need computer models.” In complex adaptive systems what appears to be obvious is not always so (occasionally disastrously not so). Personally I’m comforted by having a foundation theory.

Next time: Games of chance? Cause and effect in innovation ecosystems.

3 Comments on “The Gardener’s Dilemma”

  1. Ray MacNeil says:

    Interesting piece on the computational issues of complex systems. Its not completely clear why this solves anything however.

    For example, what are the ‘optimal’ size of the patches?

    Are they thematically generated?

    If they are, who decides on the themes?

    If the patches are self-organizing, what is the relationship across the patches? Ordered?

    To extend your analogy a little further, cross-pollination is an essential part of the gardener’s dilemma in that some plant’s require cross-pollination. So perhaps the metaphor breaks down a little there?

    I liked where the piece was going, but perhaps you could extend your thinking here sometime in another piece with more specifics on this idea of patches and why its useful.

    Just came across your blog and will certainly keep an eye on it in the future.



    • Alistair Brett says:

      Thank you Ray for your comments which help me to think more and research further. One purpose of this blog series is to raise questions for which I don’t necessarily have all the answers. As I mentioned, I was trying to come up with an answer to the question from a colleague. Your questions will help me refine my response in a future blog. I need to think about your “cross pollination” point. We don’t want isolated patches. The research to which I refers discusses optimal patch sizes in computer models, but for application to innovation ecosystems more work is needed.

      As to the usefulness of patching, I will say more about this later and try and address your other questions. In a current T2VC project to improve an innovation ecosystem in a Latin American nation we are collecting experience and results from patching. Our team has selected themes based on both similarities of needs and an addition of conflict (see my December 2013 blog “let’s all have a good argument”).

      Please keep commenting.

  2. […] and ‘fitness’ of the innovation landscape  (see February 2014 blog The Gardener’s Dilemma for the concept of […]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s