Thursday, August 21, 2014

Non and multi model reality, decisions, failures

I read an interesting article today about the "The problem with risk models". It triggered a variety of thoughts in me.
Humans experience the world through conceptual models of it. The Buddha is said to have described how to experience the world as-it-is (i.e. without models I assume). But I don't even understand what that means. So for practical purposes, human experience is all about models of the world. Science and Mathematics both build and use models (structures/frameworks). Science builds theories (models) to explain the world and scientific revolutions (a la Thomas Kuhn) update these models towards higher accuracy. The models to explain different aspects of human experience are different from each other and are essentially "local" to that area of experience. There is no one theory/model explaining how to bake the best bread as well as explain gravitation in the universe. This is the nature of the game. Humans build models based on experiences determined by location and time.
Human actions depend on the mental models by which humans evaluate their options and choose among them. These mental models are based on the individuals experience or his learning from the experience of others (past/present) and contain multiple assumptions (not all validated/validatable). A common feature among them is that they are biased by human experience.But reality is not constrained by the presence or absence of a human to experience it and there is always enough reality never experienced by any one human or those that he/she has had occassion to learn from. Essentially the models are based on limited experience and are speculations that they are representative enough for that scale, time and type of experience. Essentially models have gaps by default. They are useful ways of dealing with reality, but one needs to stay aware of the potential realities that the model does not help in thinking about, since there almost always are such potential realities. Essentially one needs to understand the limitations of one's models at all times, else one tends to fool oneself far too easily with dire consequences.
Now let us consider using a model. Most models are useful "most of the time". It is the remaining time where the model is not useful, but the case in question has high impact that really matters. Not knowing the "region" of experience in which a particular model is useful as different from the "region" of experience where it is not, is a key part of knowing the model and using it correctly. Using multiple models to manage a complex area of experience where one model is better for one area and not for other is sensible, but requires more skill and care. Till knowledge advances and one model applicable to the entire complex area of experience becomes available, such multi-model approaches are unavoidable. Oversimplifying reality assumed by choosing just one model is risky.
Now consider models driving actions of humans. Actions of one human triggers same/different actions by different humans. Sometimes such action chains are consciously designed (and constitute "systems" in a loose sense) and may have feedback loops too triggering same/different actions by humans who originated some of the actions. Such "systems" cause correlations between actions of different humans (some of which may be related to the reasons explored in the article referrred above). Most of the time feedback loops in such "systems" sort themselves out, but sometimes they do not causing "system failures". There are real correlations and spurious correlations between actions of various humans and only some "systems" really exist. Different "systems" of humanity have different drivers for the correlations between constituents. But most of the time these "systems" are consciously or unconsciously designed to work "most-of-the-time" and little attention is paid to identifying and preventing failure scenarios. Humanity is still learning to understand how to identify and prevent potential failures in time in most of its key systems.
Previously most public systems evolved and were hardly designed. Now many public systems are designed and their evolution is governed consciously. But humanity's understanding of the process of design and management of such systems is in its infancy. As Internet pervades more and more living and non-living things on this earth, the number of drivers of correlations between human actions is set to increase due to increasing number of conscious or unconsciously designed human "systems". As this happens, unless humanity learns to identify, understand and manage these "systems" better, we are sure to have multiple crisis which we do not understand. We increasingly inhabit a world which is becoming more and more complex and unless our understanding of it evolves with newer concepts and tools to identify, understand and manage "systems" at social and global level, we are set to undergo a period of unprecendented turmoil.
I talk about some of these issues in my other article but unfortunately there is little consciousness among humans of the changed nature of our common existence as humanity continues in its reverie. But I contend that this reverie will soon be broken and a frantic struggle to make sense of an increasingly complex world will start soon.
Pratap Tambay
2nd August 2014
Post a Comment