Looks like another mini-blizzard is on its way this weekend. At least in CT it may be more rain than snow. In the edX course Quantum Mechanics and Quantum Computing, we are entering the second week and we are learning about “entanglement“. I am really fascinated by this stuff. The first assignment that was due last Monday required some inner product calculations for projecting vectors. And I had to resort to pen and paper! My wife could not believe what I was doing and wanted to take a picture of me! I am really enjoying this class. It is pretty hard math and but I am going to stick with it to the end. We shall see.

Talking about entanglement, one of the books that I read a while ago is one of my favorites – The Age of Entanglement: When Quantum Physics was Reborn by Louisa Gilder. It is so well written, with just enough scientific concepts, but more of a story about the debate about quantum mechanics and gives you a glimpse into the lives of some of the great scientists. I read it a second time recently.

When we announced extended hours for the library, there were several concerned community members who wrote to us to ask what we are doing about the safety of our Library Collections. It was really heartening to hear from so many people with genuine concerns. We have thought about this a lot and have tried to convey many of the measures we have taken, but I thought it is worth discussing some of them here. I want to caution that LTS staff are preparing a detailed plan and will communicate a response to these questions in short order. These are simply my ramblings on the subject.

In the library, we have very valuable items – books, journals, CDs, DVDs, VHS Tapes, Art, etc. Whereas generalizations are dangerous in this context, it is fair to say that depending on the library some of these items may no longer be available for circulation, and that some of them are on the decline. For example, science libraries have far less print journals than ever before because most of the journals are available electronically and the science faculty and students are fine with accessing them this way. It is the case that wherever we have valuables, we need to worry about their loss – this is the unfortunate truth.  In our case, the question before us is, what exactly are we doing in order to prevent/minimize collection losses, especially when we leave the library open late. In order to answer this question better, you need a baseline. You need to know what your current losses are and then you put in measures to prevent them and measure again to see if your measures are being effective.

We have several measures currently in place and we continue to refine them and add more measures, consistent with the best practices as they apply to a small liberal arts college. For example, ACRL has a set of guidelines, that calls for a Library Security Officer, which is probably not something we can afford, but we do have the responsibility for the security distributed amongst several of our staff. On the other hand, we are right in line with some of the other guidelines, such as the segregation of  valuable collections from the rest (Special Collections and Archives) or installation of video cameras.

Of course, collection loss comes in many ways – someone stealing a book, a mis-shelved book, a book from which certain pages are either missing or are unusable etc. The real question is – how do you estimate the loss? This is very hard to do unless you are committed to certain inventory control practices. This is time consuming and very resource incentive. Therefore one needs to look at return on investment. It is also the case that with the deterrents in place, one would like to think that the loss has gone down. The reason why I say this tentatively is that we really don’t know the answer, precisely because we cannot clearly estimate losses. But there are soft evidences. For eg. we know that the circulation statistics shows continuing decline, which means our reshelving efforts must be less (therefore, less errors and lost books) and that we should be seeing fuller shelves.

The authoritative source for our holdings, of course, is our catalog. For the academic year ending in June 2012, the total entries for various collections (books, maps, journals, maps, videos etc.) is about 1.116 million. Can we account for them accurately? The answer is no. We know how much we should have pretty accurately in a theoretical sense, we know how many are in circulation at any given point, but how do we know that the remaining items are physically in the libraries? Or that they are shelved in the right places? RFIDs on books can possibly solve this issue in that you can have scanners that do high speed scanning and you can get a pretty good estimate of what we physically have in the building. Even the mis-shelved items can be detected this way. But this is expensive!

This is where the return on investment comes in – not knowing the extent of loss, it is very hard to justify an RFID project, which, even when implemented in a small scale, can be prohibitively expensive.

We are launching a small scale inventory initiative to try to get a handle on the loss. It basically involves selecting through some criteria a subset to inventory. Then you go ahead and look at what are in the shelves and what have been checked out to derive at some estimates of what/how many are missing. Doing this over a period of time will reveal the necessary trends. This of course, will only provide an estimate, but better than nothing.

I think it is important for us to reinforce to the patrons that the library collections are there for their collective benefit. If they do any of the things that results in loss of collections, it is an irresponsible behavior that affects a large number of patrons. With the increased security measures we have put in place, we also should develop tighter policies in dealing with collections loss.

I would think that this lends itself to an excellent modeling study and I wonder why there have not been many already done. I traced one publication in 1976, Estimating Lost Volumes in a University Library Collection. I will attempt to get the full text version of it to just understand the methodology. I can see umpteen different ways to model this problem. If you think about this, this is an inventory controls issue and the term that is used in the industry for the loss is “shrinkage“. One could build on the best practices there (which are very similar to what we are thinking of doing).

But a more interesting approach to this may be an artificial intelligence approach. How about a neural nets? As you can see in this reference, neural nets have been used to study the general inventory control problems. One could adopt a similar approach to learning about the changes to inventory to develop a predictive model which can help answer all sorts of things – are there predictable patterns to discipline specific collection loss? how does the loss relate to the population of borrowers (for example, does the losses change predictably with every new cohort of first years?), can we predict when the losses are likely to peak during a year (so we can fidn ways to reduce the loss)?, is the gate count a predictor of loss (for example,  is the loss greatest when the traffic is the least or vice versa)? OK, I will stop here!

Whereas these are possible, neural nets are only as good as the data we feed. And discussed earlier, the data needed to anser some of these questions are needed at such level of detail that no one has the resources or willingness to collect them. Finally, even if we can do all of this, we need to be EXTREMELY careful to formulate the questions at the aggregate level and not beyond that.