Maybe my questions are:

What kinds of aggregate patron data can identify individual patrons? (If this sounds like a paradox, think of the AOL search log fiasco) When does aggregate data disrupt a patron’s freedom to read?

I want libraries to be pro-active in this because in the short term anyway, most of us are going to use the social software tools that ILS vendors provide.

I’m excited to see libraries be centers of online communities instead of just nodes, so thanks for setting the bar so high.

I work in a fairly big public library service in Canberra, Australia, which is only just recognising the Library2.0 opportunity and moving very slowly on it. I think Library2.0 will offer and demand a lot more from peak bodies in organising disparate public libraries to move in similar directions.

Also, are there specific borrowing trends where time is a crucial factor — we’ve played around with something we’ve dubbed a “lending path” where we try and predict the book that’s most likely to be borrowed next.

If you look at the path for Harry Potter and the goblet of fire, the previous three books are in the correct order of the series. In other words, if you borrowed “Goblet”, then the most likely book you borrowed before that was “Azkaban”, and before that “Chamber of secrets”, etc.

Whilst it might seem obvious to a Potter fan that the books should be read in that order, most OPACs don’t explicitly state what the series order is (unless a cataloguer has added notes). Wouldn’t it be great if we could tap into the borrowing trends to help organise books or make more personalised suggestions?

The question this brings up to me is whether suggestions based on checkout patterns are really analogous to Amazon’s suggestions based on purchase. I check out lots of books from the library that I end up not liking, but I’m more careful when I purchase to make sure it’s something I’m going to like (or the giftee will like) because I’m spending money. It seems that some form of a rating system would be a great partner to this for the suggestions to be truly meaningful. IOW, maybe we should be looking more to Netflix than Amazon.

To clarify where this data comes from: each of our users are given the option to turn on check-out history. It is not, in any way, pushed upon them and is an opt-in system. So by opting-in, users are aware that their circulation data is stored on our system unless they specifically delete it. So we are able to generate recommendations on the fly, without compromising privacy.

Caleb, we did have some internal discussion about the privacy issues behind this particular feature and decided that because there is no way for the general public to track a particular recommendation back to a particular user, we would proceed.

But what I think you’re really asking about, Caleb, is whether we’ve had a larger discussion about the potential privacy implications. And the answer is yes. You mention the term “selling”, but that’s really not what we’re trying to do. We simply want to provide a more enriching experience for our users that is based on search and discovery, not just the search.

I’ve always stated that I believe that social software in libraries would not compromise privacy because we can develop these systems in parallel without touching the existing ILS.

In the UK we have the Data Protection Act which places restrictions on data that can “identify a living person”. Like John, we aggregate essentially anonymised data — in fact, one of the advantages of generating the suggestions off-line (rather than on the fly) is that the table of suggestions used by the OPAC simply contains lists of bib numbers and nothing else.

When we implemented it, the only concerns that were raised by staff related to how our students would perceive the suggestions — would they realise that they were based on peer borrowing, or would they think these were suggestions being made by librarians or tutors?

Ultimately, I think it does raise the question of whether or not there is a divide between what our patrons want and what we think they want.

Or am I really the only one who sees the potential for one being a barrier to the other?

I take it you’re generating the suggestions in real-time? We’ve got nearly 3 million circ records, so I’m having to re-generate our suggestions once a week.

Have you looked at borrowing trends through time yet? In other words, “is book A more likely to be borrowed after book B” — if so, what’s the average amount of time that elapses between. I love the idea that you could predict exactly what someone is going to borrow 6 months from now just by looking at the books they’re borrowing today!

Caleb — we’ve not had a single query about privacy from our borrowers since we added suggestions to our OPAC in Nov 2005… in fact, the suggestions are easily the most popular and well used enhancement we’ve ever made to the OPAC.

My suspicion is that our users are so used to seeing suggestions elsewhere (e.g. Amazon) that it simply doesn’t surprise them to see them in the OPAC.

Libraries’ attitudes towards privacy and intellectual freedom are challenged by some aspects of providing social software, and I’m glad that the library community is starting to have some discussions about it.

For example, recommendations built into the library catalog can erode the perception of neutrality that many libraries try to cultivate (for better or for worse and probably both).

I appreciate the opt-in policy and that books don’t get recommended unless at least a few people are in accordance, though 3 readers seems like a pretty low threshold, I think that setting it too high would lead to other problems.

I’m curious to know, what discussions about privacy and intellectual freedom did AA have on the way to this feature? Were there barriers, and how did you get over them?