I have much respect for Nassim Taleb. Eg, I don't understand all he explains in his Dynamic Hedging, but I assume I am the one to blame Nonetheless I don't really get what is new in Taleb's concept of antifragility.Eg, excerpt from the CFA Institute's website:"Taleb?s antifragility may be distilled down to the idea that all systems need exercise. That is, systems need to suffer, just a little bit, in order to remain fit and ready for the uncertainties of the world."Is that so different from what Bertalanffy explained on systems?And it seems to me that later on Economists of Innovation eg Nelson and Winter (back in the 80s) have grounded their works on such ideas.I will be glad to hear your comments.

QuoteOriginally posted by: edouard I have much respect for Nassim Taleb. You do? Try this one: Antifragility

I don't know that any idea is ever really "new" although some ideas create more change than do others. I'm sure a great many people have looked at the behaviour and dynamics of systems under exogenous and endogenous loads or perturbations. There's plenty of interesting work in this area in systems theory, biology, economics, control theory, etc.What is interesting about Taleb's view is the reframing of risk and risk-affected systems. Rather than strive for systems that never create risk or volatility, he's advocating that we create systems that benefit from risk and volatility. Although the recent crisis has brought a repudiation of risk as if the creation of risk caused the problem, I think he's arguing that risk aversion is no solution. Flight-to-safety is a false refuge. In fact, risk aversion is actually a very dangerous solution.And he's right that the English language has no word for things that are the opposite of fragile. Loosely speaking, we might define:Fragile: ∂Performance(t)/∂Volatility << 0Robust: ∂Performance(t)/∂Volatility = 0Anti-Fragile: ∂Performance(t)/∂Volatility >> 0And if we are to be strictly correct, a fragile system (e.g., a crystal champagne flute) is one that performs to perfection up to some threshold level of stress and then abruptly shatters. Fragile systems do not degrade gracefully. That's what makes them so dangerous. They seem to function extremely well under some day-to-day modicum of stress and then fail catastrophically under some infinitesimal increase in that stress. Moreover this is NOT an issue of rigidity (although fragile things do tend to be rigid). An aluminum champagne flute and a crystal champagne flute are equally rigid yet the aluminum one fails by bending. Similarly, one can build glass structures that are extremely flexible and yet when they reach some critical level of stress, they still fail catastrophically. Its the hidden nonlinearity that make fragility so dangerous.The root cause of fragility of a material (e.g., glass) or a system (e.g., a bank or an economy) lies in how it handles minor imperfections with incremental loads (defects + stress). In a material such as glass, a defect + stress creates a self-amplifying growth of the defect so that local failure almost instantly leads to global failure. Although avoiding risk (e.g., "handling with care") might seem advisable, that's no solution both because defects inevitably occur and stresses inevitably overstep the threshold of failure.An anti-fragile system would have the opposite response to local defects and local extrema in stress -- it would grow stronger, not shatter. From an economic, regulatory, and organizational perspective, an anti-fragile system must have implicit or explicit mechanisms for handling (rather than avoiding) local defects and local concentrations of stress in ways that increase the system's tolerance to said defects and stress. But even more so than "tolerating" defects and stress (which is simple robustness), the anti-fragile system improves with exposure to defects and incremental stress. In other words, it learns from each defect or incremental stressor to be able to handle an every greater amount of that defect or stressor. The result is a system in which the optimum defect rate and stress volatility is greater than zero -- the system constantly pushes the limits in small local ways, handles the resultant local failures, and increases it's performance and operating range.

pfffff nice explanation T4A. thank you very much indeed.

QuoteOriginally posted by: edouardpfffff nice explanation T4A. thank you very much indeed.You are most welcome. I'd been thinking about posting on this topic for some time to help get the ideas sorted out in my head. Your question motivated me to do so.

I like the concept of Taleb but it's like the snake which eats itself.For "fragility" model risk you need a model, then you need a model for your model risk of your model risk and so on In some sense I even think that these concept is dangerous because it induces a sense of "security" (why did I go bankrupt my model risk was supposed to be "antifragile" with respect to my model?!!). But this is a nice starting point anyway.

Reminds me of two concepts: Spontaneous symmetry breaking and crystallisation.

QuoteOriginally posted by: frenchXI like the concept of Taleb but it's like the snake which eats itself.For "fragility" model risk you need a model, then you need a model for your model risk of your model risk and so on In some sense I even think that these concept is dangerous because it induces a sense of "security" (why did I go bankrupt my model risk was supposed to be "antifragile" with respect to my model?!!). But this is a nice starting point anyway.You only need a risk model if you want to estimate a probability of failure. A fragility model is more of an "if-event-then-impact" model, not a P(event) or P(loss) model. I think Taleb would be the first to agree with you that constructing P(event) or P(loss) is a fools errand and a dangerous one at that.A fragility model is more of a ("if-event-then-impact1" AND "if-impact1-then-catastrophe") model In antifragility, we seek to: 1) understand these if-thens, their coupling, and the potential chains of causality that lead to catastrophic failure; 2) change the conditional causal structure to reduce the cardinality of event states and impact states that lead to serious failure states. That is, we want to characterize the conditional nonlinearities in the system and then remove or limit those nonlinearities. Events and losses will still occur, but they are less likely to lead to nonlinear cascades of losses. The system is more likely to bend than break.Although it might be tempting to try to prevent all events that create an impact (i.e., risk aversion), that strategy has two significant draw-backs. First, one can't prevent all events both because some exogenous causes (natural disasters) lie mostly outside the spheres of human influence and because an open system is bound to create new unforeseen classes of events (e.g., IceSave). Second, even if many events were prevented, the result would be a false-sense of safety and under-investment in the kinds of systems need to respond to disaster.What's interesting is that the anti-fragility approach directly addresses the Ouroboros problem of modeling the model risk of model risk ... of a model. It does this by intentionally exercising the system with some level of stresses and defects. The empirical treatment of the system with risks helps map the if-then causality. Each time someone pushes the limits a little too hard, an anti-fragile system suffers some minor consequences that teach the system and make it stronger for the future.I'll also note that even bankruptcy is not a terminal failure. The US Chapter 11 bankruptcy process (I'd imagine the EU has some analog for reorganizing a stricken company) could be seen as a powerful anti-fragility feature because it lets the organization continue to function (maintaining both employment and the value of tangible and intangible assets) while legally decoupling the organization from a destabilizing influence (e.g., unsustainable debts or liabilities).

we already have a thread dedicated to all things Talebian

QuoteOriginally posted by: daveangelwe already have a thread dedicated to all things TalebianA place for everything and everything in its place!Methinks that's a fragile strategy! Besides, Nassim seems like a multithreaded kind of guy.

QuoteOriginally posted by: Traden4AlphaQuoteOriginally posted by: daveangelwe already have a thread dedicated to all things TalebianA place for everything and everything in its place!Methinks that's a fragile strategy! Besides, Nassim seems like a multithreaded kind of guy.i liked this review in the GrauniadQuote So what we get are lots of personal reminiscences buttressed by the ideas of the few thinkers he respects, almost all of whom happen to be his friends. The result is both solipsistic and ultimately dispiriting. Reading this book is the intellectual equivalent of having to sit patiently while someone shows you their holiday snaps.

QuoteOriginally posted by: daveangelQuoteOriginally posted by: Traden4AlphaQuoteOriginally posted by: daveangelwe already have a thread dedicated to all things TalebianA place for everything and everything in its place!Methinks that's a fragile strategy! Besides, Nassim seems like a multithreaded kind of guy.i liked this review in the GrauniadQuote So what we get are lots of personal reminiscences buttressed by the ideas of the few thinkers he respects, almost all of whom happen to be his friends. The result is both solipsistic and ultimately dispiriting. Reading this book is the intellectual equivalent of having to sit patiently while someone shows you their holiday snaps.LOL! That's not an entirely unfair assessment.Yet despite this, one must still answer the question: what is the opposite of fragility and how do we make economic systems that are the opposite of fragile? (Or is it impossible?)

QuoteOriginally posted by: Traden4AlphaQuoteOriginally posted by: daveangelQuoteOriginally posted by: Traden4AlphaQuoteOriginally posted by: daveangelwe already have a thread dedicated to all things TalebianA place for everything and everything in its place!Methinks that's a fragile strategy! Besides, Nassim seems like a multithreaded kind of guy.i liked this review in the GrauniadQuote So what we get are lots of personal reminiscences buttressed by the ideas of the few thinkers he respects, almost all of whom happen to be his friends. The result is both solipsistic and ultimately dispiriting. Reading this book is the intellectual equivalent of having to sit patiently while someone shows you their holiday snaps.LOL! That's not an entirely unfair assessment.Yet despite this, one must still answer the question: what is the opposite of fragility and how do we make economic systems that are the opposite of fragile? (Or is it impossible?)How would evolution have evolved (sounds like a bad English) if the world was not fragile ?

QuoteOriginally posted by: daveangelQuoteOriginally posted by: Traden4AlphaQuoteOriginally posted by: daveangelQuoteOriginally posted by: Traden4AlphaQuoteOriginally posted by: daveangelwe already have a thread dedicated to all things TalebianA place for everything and everything in its place!Methinks that's a fragile strategy! Besides, Nassim seems like a multithreaded kind of guy.i liked this review in the GrauniadQuote So what we get are lots of personal reminiscences buttressed by the ideas of the few thinkers he respects, almost all of whom happen to be his friends. The result is both solipsistic and ultimately dispiriting. Reading this book is the intellectual equivalent of having to sit patiently while someone shows you their holiday snaps.LOL! That's not an entirely unfair assessment.Yet despite this, one must still answer the question: what is the opposite of fragility and how do we make economic systems that are the opposite of fragile? (Or is it impossible?)How would evolution have evolved (sounds like a bad English) if the world was not fragile ?Now that is a really good question!Personally, I would say that the biological world is not fragile regardless of all the environmentalists' talk of "delicate balances". That individuals, populations, species and even entire ecosystems disappear is only fragility if one assesses performance as MIN( { E(performance(subunit_i)) } ) rather than E( {MIN(performance(subunit_i)) } ) or some other aggregate central tendancy. In fact, some might say that extinction is a key necessity of antifragility -- culling defective subelements is good (unless one is that subelement).But that leads to some unpleasant ethical questions for proponents of antifragility. A key element of antifragility must be some form of firewalling -- ensuring that losses in one part of the system don't cascade into the larger system. Yet a firewall inevitably means that some people are on one side and some are on the other and if one is on the wrong side then one is toast. For example, maybe we must let some banks go extinct rather than let all banks think they will be rescued regardless of behaviour. In other words, antifragility must involve a certain amount of tough love in which government is far less likely to step in a save people from themselves in order to motivate more people to be prudent and ensure that society is less likely to be left holding a large bag of collateralized poo. Thus, I'd say that there may be a negative correlation between empathy and antifragility.

I'm still not convinced that "antifragility" is really sufficiently usefully different from "robustness" to be worth a new word. It still seems to just be a matter of the degree and scope of the stresses that are survivable. It seems impossible, in a logical, almost Goedelian sense, that a system can be so adaptable that it can survive/adapt to any as yet unconceived set of stresses. Particularly in a forward looking economy with opportunistic agents, as soon as the system is devised, people will start trying to subvert it. And subvert it they will. In a sense, to try and formalise the idea of antifragility embodies exactly the sort of risk-management hubris that the idea seems to want to avoid!