Antifragile Libraries pt 1: Antifragility and collections

25 November 2013

I recently read Nassim Nicholas Taleb's excellent book Antifragile: how to live in a world we don't understand. It's a rather sprawling, heavily footnoted opus - but with time to reflect I think Taleb has a great deal to teach librarians.

Given the word count Taleb assigns to railing against bureaucrats, corporations, universities and government institutions, he may be less than impressed by my application of his ideas to academic, corporate and public libraries. He does make an exception for municipal government, however, so perhaps he would let public librarians like myself off the hook.

Taleb is a big fan of the private library, one that allows its owner to read wherever her interests take her. A wonderful idea, of course, but one that many humans, due to a lack of means, simply cannot realise. The public (or, in some cases, academic) library can make up for this if it is managed well.


Firstly Taleb introduces the concept of 'anti-fragility'. The anti-fragile is not just robust, surviving shocks and instability. To be anti-fragile is to benefit from instability. A supermarket chain with a long just-in-time perishable food supply chain is fragile. A gun smuggling operation is anti-fragile. A publicly listed bank is fragile. A small Sharia bank is (probably) robust. A money-changer is anti-fragile.

Taleb further explains that anti-fragile systems or entities tend to be made up of a large number of fragile entities. So in biology a species or ecosystem may be fairly anti fragile, but this is because of variation between fragile individuals within it - the weaker individuals die faster and fail to pass on their weak or sub-optimal genes. This natural selection takes place more rigorously in unstable environments. For the fragile individuals this may lead to a quick death, but for the population at large it leads to greater strength and resilience.

Antifragility is all about risk (which is unsurprising given Taleb's long standing interest in risk). Instability means an increase in unquantifiable risks. If increased risks/chances benefit you (to a point), you are anti-fragile. If they damage you, you are fragile. If they make little difference (to a point) you are robust. Another way of thinking about this is in terms of of 'optionality' - the more options you have, the more anti-fragile you become.

Taleb gives a fair amount of space to the concept of iotragenics. This is a fancy word for the law of unintended consequences. An iotragenic medical treatment is one that is undertaken to fix some condition, but in the process causes or exacerbates another medical problem. So chemotherapy kills cancer cells, but it also kills all the white blood cells that fight off viruses, thereby weakening the patient in a potentially fatal way. Iotragenics is the reason the Hippocratic Oath instructs doctors to 'First, do no harm'.

Taleb argues that our modern (or, more accurately, Modernist) culture ignores the risk of iotragenics and is addicted to 'acting', rather than seeing the advantages of, well, not acting. Taleb calls this the 'via negativa' approach.

Finally, Taleb crystallises an ethics based on his idea of the fragile-robust-antifragile triad. This ethics is based on a simple proposition that whenever anyone makes a decision to take a particular action, they are making a decision about risk. The trick is, the risk is not always born by the person making the decision. Taleb's triad can be used to consider who carries the risk, and whether the person making the decision has 'skin in the game'. His triad then looks like: [no skin in the game]-[skin in the game]-[skin in the game for the sake of others]. Of course, the banking industry has a phrase for when the person taking the risk has 'no skin in the game' - moral hazard. Taleb keeps things simple - if you don't have skin in the game, any decisions you make that increase downside risk are unethical.

Antifragile libraries

So what does this all have to do with libraries?

It has become a cliche to write that these are uncertain times for librarians. As information professionals living through an information revolution, it is hardly surprising that much of the discourse between librarians concerns change, uncertainty, and how to deal with it. Traditional models for information exchange and distribution have proven to be rather more fragile than was assumed. Taleb's ideas provide us with a framework to guide ourselves, our libraries and our communities through the chaos - not just surviving, but thriving. By becoming anti-fragile, uncertainty becomes our friend.

Left leaning barbells

We can start with that most longstanding and traditional library activity - physical collection management. Library collection management has increasingly moved from 'just in case' to 'just in time' models following industrial production patterns. This seems more efficient, but it leaves us more at the whim of commercial publishers and more importantly at the whim of their continued existence. What we have done is increase our efficiency at the cost of optionality. In some ways legal deposit libraries are best placed (because they keep everything), but of course it is not necessarily sensible for resources to be spent storing thousands of old tomes at the expense of purchasing new ones or providing other services and facilities.

One solution has been attempted by academic libraries in Australia with the creation of CAVAL and the CARM shared resource centre. The CARM collection is essentially a lending library for libraries, so material used less often is stored offsite in a shared repository. This is a great idea - except that everything has now been centralised into a single point of failure. Whilst the CARM Centre is climate-controlled and secure, it can't mitigate against all risks, only known ones that are considered plausible.

From an optionality, anti-fragility point of view, it is better to have multiple redundancy by not relying too much on centralised storage systems and the assumption that 'someone else will keep it'. Whilst more expensive, a system of multiple off-site repositories holding the same works is a safer long term bet. This is essentially how the National Archives of Australia (NAA) manage their electronic records. The NAA has built multiple electronic archive sites physically separated from each other, and runs two completely separate networks for digital archive storage.

Digital information technology simultaneously increases both fragility and anti-fragility. On the one hand, it is a lot cheaper and easier to store documents and information in digital format rather than on paper. This fact by itself means that much more information is stored, thereby becoming discoverable by scholars and citizens. Combined with increasingly sophisticated software, the 'long tail' of documents provides researchers with optionality on a scale never known before in human history. We call this information overload. On the other hand, digital storage media and techniques are highly fragile. Every person reading this has probably experienced data loss due to a corrupt file, failed hard-drive, or superseded file format. Digital storage requires huge amounts of reliable electricity. Computing equipment increasingly relies on precious metals and 'rare earths' in its construction. The name itself reveals the fragility of any strategy relying on such materials remaining readily available. Digital media can also be altered invisibly, 1984 style. An archive where records can be altered without anyone ever realising is fragile indeed.

So managing physical collections with constrained budgets and storage requires us to constrain our options, tending towards what is popular at the moment, whilst digital collections decrease the fragility of scholarship and casual research, but are themselves highly fragile. What to do?

An anti fragile approach suggested by Taleb is the 'barbell strategy'. Taleb wrote about this strategy in his earlier book The black swan. There, he described an investment strategy whereby 90% of funds are invested in very conservative assets, such as US Treasury bonds or gold, and 10% is invested across a wide range of highly risky assets with possible but unlikely massive returns, such as tech stocks or junior mining companies. Taleb expands this idea in Antifragile to consider famous authors who combine risky (from a financial point of view) artistic creation with very secure and intellectually tedious day jobs. Essentially, what Taleb is pointing to is that the traditional approach to risk reduction is to aim for the average, based on calculations of risk. But the risk of 'black swan' events by their very nature can't be calculated, so simply taking the 'average' is a flawed long term approach. The average day for the turkey involves generous feeding and not much stress. Right up until Thanksgiving, when the 'average day' suddenly becomes a poor indicator of what to expect.

Librarians might consider whether a 'barbell' approach to collection management could be a winning strategy of long term success - particularly in volatile times like these. For example digitisation is not necessarily 'better' - it might be worse in the long term due to fast-changing technical standards and the need to secure the data in increasingly complex ways. A 'barbell shaped' approach to digitisation might involve saving it to high quality microfilm but also making it available online in today's digital formats. This gives the benefits of a useable and convenient way of accessing the information in the short term, and combines it with the long term stability (up to 500 years) and human-readability of microfilm (all it really takes is an appropriate magnifying lens and a light source). Some major collecting organisations already do this.

In my own workplace we are considering this barbell approach to our Reference collection. Most public libraries have significantly reduced the size of their hardcopy reference collections in the last decade. Some have disappeared altogether - the remains absorbed into the lending collection. This has been a response to the rise of online resources for generalist reference sources and the decreasing availability of traditional hardcopy reference resources as they are migrated online or simply discontinued.

Traditional approaches to library collection development use systems like CREW/MUSTIE but ultimately often come down to considerations of frequency of use. In an average suburban lending collection this is generally appropriate, although it can go a bit pear-shaped if you blindly weed-by-numbers.

These traditional approaches rely on some contestable assumptions. The first is that since we serve the general public, our collections should be broadly generalist and steer away from material (or collections of material) that are highly specialised. Secondly, that historical patterns of use should be the main factor in determining whether they deserve shelf space.

From the perspective of a librarian attempting to build an anti-fragile collection, both assumptions are problematic. If all public libraries collect the same 'average' material, the 'local' library simply reflects a physical location rather than the focus or your services and collection. There is a vast difference between the needs an desires of a library patron in, say, Mosman and one in Alice Springs or even Macquarie Fields. Aiming for the average means two things - firstly, your hardcopy collection is essentially competing with material on the open web via the world's favourite search engine; secondly, any truly interesting or complex enquiry will be unanswerable using the material you provide. Likewise, relying heavily on historical patterns of use in a time of rapid change seems rather foolhardy.

Dan Cohen has written about this on his blog coining the term 'Wilkin Profile' to describe the uniqueness of a particular library collection. A collection containing mostly items not held anywhere else has a 'left leaning' Wilkin Profile whilst a collection made up mostly of items also held in other collections has a 'right leaning' profile.

Arguably then, what public libraries should be considering, particularly for hardcopy reference and non-fiction collections, is a left-leaning barbell shaped Wilkin Profile. Let's see what that looks like in practice.

Building a left-leaning barbell-shaped Wilkin Profile

To build a collection with a left-leaning barbell-shaped Wilkin Profile, what we're looking for is a number of very high-use items combined with a large number of unique and unusual items that will be accessed rarely.

The first type might include hardcopy dictionaries, atlases, a local street directory and similar 'straight facts' general material where hardcopy provides a better user experience than online and the information quality and accuracy degrades slowly. In my own library this would include a number of foreign language dictionaries (e.g. Chinese-English, Korean-English) due to our community demographics, which include many first-generation migrants and 'overseas' university students.

The bulk of the collection would then consist of more niche and esoteric information. This could include guides to makers' marks on silverware, guides to coins, obscure sports statistics, subject dictionaries, and books like the Machinery's Handbook

What would not be included is the 'middling' material. No encyclopaedias, almanacs and the like. Anything that can be found on the open web within 10 minutes has little place in a public library's reference collection.

When we combine this hardcopy collection with a complementary digital collection strategy, the whole information service becomes more anti-fragile. Things like shipping and Births, Deaths and Marriages records should be stored on long-life film, but also made available through easy to use digital interfaces. Valuable and rare local heritage collections can be managed in the same way. The missing middle in our hardcopy collection can be filled through a curated collection of online resources. Areas of knowledge with a short half-life such as computer programming, genetics research and so on are best dealt with in this way. This could take the form of paid-for database products or material freely available on the open web.

Suspended collections

Another anti-fragile approach to collection management could be a back-to-the-future 'social library' approach. If the storage and sunk costs are forcing us to fragilise our collections, one solution is to distribute the potential collection. The extreme example of the recent attempted destruction of the Timbuktu Library collection highlights how distributing a collection throughout a community can be highly effective at reducing fragility and increasing options.

Phil Minchin also explored this in his paper on two-way libraries (a sister document to his In the Library with the Lead Pipe article). Minchin simply calls this 'sharing copies' but I can't help thinking of 'cafè sospeso'. This 'suspended' or 'pending' library collection is physically distributed, thus both reducing or removing the cost of storage and mitigating against the risk of catastrophic destruction.

In effect, this is similar to what share traders and executives do when they purchase 'options'. A library would have the right but not the obligation to use items, and no obligation to store them.

I will continue to explore these ideas, and I encourage you to share your thoughts in the comments. on your own blog or Twitter, with a link back.

In Part 2 of Antifragile Libraries I'll explore what Taleb's idea of anti fragility can mean for librarianship more generally, looking at how we approach metadata, literacy and our role in our communities.

Discussed by

Friday varia and quick hits