Nassim Nicholas Taleb and Antifragility

At the Royal Society on Thursday I saw Nassim Nicholas Taleb (Distinguished Professor of Risk Engineering at Polytechnic Institute of New York University) in conversation with Rohan Silva (Senior Policy Advisor to the Prime Minister) and Matthew Taylor (Chief Executive of the RSA).

The subject was Taleb’s latest book – Antifragility – which he hopes isn’t as simplified as his previous book (he seemed to dislike the fact that The Black Swan was bought and read in airports by people who subsequently misinterpreted it). I think he has succeeded, as at times I struggled to grasp his arguments.

Nevertheless, I’ll have a go at summarising it. The images below are taken from his slides, available alongside the audio of the event, at the RSA website.

I believe Antifragility draws on Prospect Theory (incidentally, Michael Lewis has recently written about Daniel Kahneman – one of the main proponents of the theory – in Vanity Fair). The thesis is that organic entities are fragile, and that potential harm is non-linear to the size of the event (see below). In fragile systems, potential losses outweigh gains.

Antifragility is the opposite of fragility – where gains outweigh losses (robust is something different). The parallel Taleb drew was:

• Fragile – Sword of Damocles (risk of fall)
• Anti-fragile – Heidegger Hydra (cut one head off and get two back)
• Robust – Phoenix (stays stable)

To give an example, it is more harmful to jump once from 100m than 100 times from 1m.

This means that the unpredictability of Black Swan effects costs more when the size of the event is large.

This means (I think) that risk is concave, and there is more pain than gain in an event. For instance, traffic is concave – it is fine to the point of over-leverage, when the cost of travel will suddenly massively increase.

Antifragile events are convex. Taleb believes that bottom-up structures are convex, while top-down dictated structures are concave. We cannot effectively design systems as we cannot fully understand them, and so there will always be more harm than good.

The result of this theory is that large events, large organisations and large governments should be avoided. “Too big to fail” is catastrophic, and Taleb is firmly in favour of smaller organisations – on a moral level (it smaller societies you are close to the result of your decisions and so may feel shame) and an economic level (the cost of liquidating one large bank is far more than the cost of liquidating ten banks one tenth of the size of the large bank).

Uncertainty makes mistakes costly, and thus both businesses and governments should remain small. Taleb suggested that if Tesco suddenly ran into difficulties then the government would have to bail out a supermarket. He feels the government should only intervene in things that can’t organise organically, and is thus advising the government on how to make its institutions smaller (such as splitting the NHS into localised, autonomous units).

Taleb also talked a bit about risk management, and said that the models were flawed as they don’t combine risk and growth –  to get rich you must survive. He feels slow and steady growth is preferable to high growth that cuts corners, since that will ultimately crash. He pointed out that perfect growth is Madoff, and that Greenspan and Brown’s attempts to end boom and bust have led to where we are now (he claimed that Brown was more dangerous than Blair as Brown thought he understood the world).

He also mentioned that within small systems, stresses can be positive. For instance, we fast to kill cancer calls. In exercise, we should combine slow and steady exercise (such as walking) with the occasional intense burst (such as lifting weights at the gym). Up to a point, what doesn’t kill you makes you stronger.

Taleb doesn’t see big as permanent – Google could quite easily disappear tomorrow. To get around organisations growing in size, he recommends a contract saying that if a business wants to have the option to be bailed out then they can never pay out bonuses. Since people will want bonuses, they will intentionally keep their business small in size to not require bail-outs.

He closed by saying organic organisations are perishable and that each day they exist their life expectancy drops by a day. Whereas non-perishable entities – such as ideas – increase their life expectancy each day. For instance, if an idea has existed for 25 years then it will exist for 25 more.

Ultimately, the talk was very thought-provoking but I’m not sure whether I agree with it or not. On the one hand, I can see the diseconomies of scale of large organisations, but on the other hand businesses will have become large through success. That success might have been random and unwarranted, but they nevertheless survived and so must have done a few things correct. Taleb appears to advocate limiting them in order to protect against the possibility of them failing and causing problems for the economy. This risk aversion and diversification may be prudent, but I’m not sure how fragile this system really is (particularly if it made up of bottom-up organisations).

sk

Advertisement

The Science of Lifelong Learning

On Tuesday evening I attended the lecture / debate “The Science of Lifelong Learning“, co-hosted by the RSA and the NIACE. The evening was centred on neuroscience and how understanding of its mechanics can impact upon education. The event was recorded, and I believe that feed will soon be uploaded here.

It was an enjoyable evening, though with over-running speeches the debate was sadly lacking and limited to a couple of rounds of questions (incidentally, this was the first panel where I’d seen questions asked in groups, rather than one at a time. Is this standard?).

If there was one overriding theme, it was an echo of Public Enemy – Don’t Believe The Hype. Neuroscience may aid an advancement in our understanding of how the brain works, but it does not work in isolation and cannot alone answer all questions.

The five speakers were a mixture of scientists and educators. Sarah-Jayne Blakemore (a neuroscientist at UCL) opened things up by looking at the “seductive allure” of neuroscience. Aids such as the Brain Gym are, apparently, sold on “facts that aren’t factual”.

She quoted a great study from Weisenberg et al (2008), which indicated that people became more satisfied with a “bad” explanation of brain activity once extraneous technical words were included. This was true for both neuroscience novices and students. Neuroscience experts were immune to this effect; amusingly, they instead felt the “good” explanation was less satisfactory when the neuroscience terms were included.

The takeout for the advertising industry is that products designed to improve intelligence will sell better if brain words are included in the description.

Despite the seductive connection between the mind and brain, Blackmore concluded that neuroscience can still be used effectively – e.g. we have discovered that the brain still develops within teenagers.

Usha Goswami, a neuroscientist from Cambridge, followed up by positing that creating optimal conditions for early learning makes our later learning more efficient, and increases the complexity of what can be learned. However, the ongoing environment has a bigger effect on development trajectory. This is a reason why continual learning is important (it can also help fight Alzheimer’s disease).

Paul Howard-Jones (University of Bristol) then presented evidence that training can improve targeted cognitive functions. While there is less evidence that this can transfer to non-targeted functioning, a recent study (Willis et al, 2006) suggested that reasoning tests in 74 year olds resulted in less self-reported functional decline in everyday life. Paul also had some very snazzy glasses, where the visual shapes projected within the lenses have been shown to enhance cognitive functions.

Andrew Pollard (ESRC Institute of Education) contended that there are four ways in which we can learn – formal workshops/training, personal support, a learning culture, and opportunities for self-reflection. He placed particular emphasis on the final point, saying that biography and history shouldn’t be overlooked. Learning needs to recognise the nature of phenomena; not just the brain but meaning and identity.

Matthew Taylor (Chief Executive, RSA) rounded off the speeches by agreeing that neuroscience is important but shouldn’t be overhyped. He also spoke about the extended mind and distributed cognition. The former refers to how external devices (e.g. a mobile phone) can extend our cognitive capacity while the latter relates to intelligence embedded in our social networks. He argued that learning takes place in groups, and that “collaboration is a meta-skill for lifelong learning”. Something our friends in social media will surely agree with.

A couple of interesting points also came out of the questions. Sarah-Jayne Blackmore said that different learnings have different sensitive periods e.g. social learning can continue for decades. Matthew Taylor noted that within schools, differing innate abilities meant that few could ever fall within a sweet spot of motivated learning and that assessment systems only added to the problem. Andrew Pollard named it the tragedy of the education system – feedback is not in a loop and so people cannot take it on board.

For further reading, see the blogs of Matthew Taylor and Tom Schuller, who chaired the evening.

sk

Image credit: http://www.flickr.com/photos/fornal/